/drivers/video/Intel-2D/gen4_source.h |
---|
File deleted |
/drivers/video/Intel-2D/intel.h |
---|
File deleted |
/drivers/video/Intel-2D/sna_stream.c |
---|
File deleted |
/drivers/video/Intel-2D/sna_render.h |
---|
File deleted |
/drivers/video/Intel-2D/compiler.h |
---|
File deleted |
/drivers/video/Intel-2D/brw/brw_eu.h |
---|
File deleted |
/drivers/video/Intel-2D/brw/brw_wm.c |
---|
File deleted |
/drivers/video/Intel-2D/brw/brw.h |
---|
File deleted |
/drivers/video/Intel-2D/brw/brw_eu_emit.c |
---|
File deleted |
/drivers/video/Intel-2D/brw/brw_sf.c |
---|
File deleted |
/drivers/video/Intel-2D/brw/brw_eu.c |
---|
File deleted |
/drivers/video/Intel-2D/utils.c |
---|
File deleted |
/drivers/video/Intel-2D/sna_transform.c |
---|
File deleted |
/drivers/video/Intel-2D/kgem_debug_gen5.c |
---|
File deleted |
/drivers/video/Intel-2D/kgem_debug_gen4.c |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_sf.g5b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_yuv_rgb.g4b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_yuv_rgb.g6b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_sf_mask.g5b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_src_projective.g4b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_noca.g5b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_ca.g4b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_src_projective.g6b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_src_sample_argb.g5b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_ca.g6b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_src_sample_planar.g4b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_mask_projective.g5b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_src_sample_argb.g7b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_src_sample_planar.g6b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_mask_sample_argb.g4b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_mask_projective.g7b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_src_sample_a.g5b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_mask_sample_argb.g6b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_write.g4b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_src_sample_a.g7b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_mask_sample_a.g4b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_write.g6b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_mask_sample_a.g6b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_src_affine.g4b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_xy.g4b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_ca_srcalpha.g5b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_src_affine.g6b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_mask_affine.g5b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_sf.g4b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_mask_affine.g7b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_yuv_rgb.g5b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_noca.g4b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_yuv_rgb.g7b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_sf_mask.g4b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_noca.g6b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_src_projective.g5b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_src_sample_argb.g4b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_ca.g5b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_src_projective.g7b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_src_sample_argb.g6b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_mask_projective.g4b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_src_sample_planar.g5b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_mask_projective.g6b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_mask_sample_argb.g5b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_src_sample_a.g4b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_src_sample_planar.g7b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_mask_sample_argb.g7b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_src_sample_a.g6b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_write.g5b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_mask_sample_a.g5b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_write.g7b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_mask_sample_a.g7b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_ca_srcalpha.g4b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_src_affine.g5b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_xy.g5b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_ca_srcalpha.g6b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_src_affine.g7b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_mask_affine.g4b |
---|
File deleted |
/drivers/video/Intel-2D/render/exa_wm_mask_affine.g6b |
---|
File deleted |
/drivers/video/Intel-2D/pciaccess.h |
---|
File deleted |
/drivers/video/Intel-2D/gen4_vertex.c |
---|
File deleted |
/drivers/video/Intel-2D/sna_cpu.c |
---|
File deleted |
/drivers/video/Intel-2D/sna.c |
---|
File deleted |
/drivers/video/Intel-2D/gen4_vertex.h |
---|
File deleted |
/drivers/video/Intel-2D/sna.h |
---|
File deleted |
/drivers/video/Intel-2D/kgem.c |
---|
File deleted |
/drivers/video/Intel-2D/gen3_render.c |
---|
File deleted |
/drivers/video/Intel-2D/gen4_render.c |
---|
File deleted |
/drivers/video/Intel-2D/gen5_render.c |
---|
File deleted |
/drivers/video/Intel-2D/gen6_render.c |
---|
File deleted |
/drivers/video/Intel-2D/gen7_render.c |
---|
File deleted |
/drivers/video/Intel-2D/sna_reg.h |
---|
File deleted |
/drivers/video/Intel-2D/kgem_debug_gen3.c |
---|
File deleted |
/drivers/video/Intel-2D/kgem.h |
---|
File deleted |
/drivers/video/Intel-2D/gen3_render.h |
---|
File deleted |
/drivers/video/Intel-2D/gen4_render.h |
---|
File deleted |
/drivers/video/Intel-2D/gen5_render.h |
---|
File deleted |
/drivers/video/Intel-2D/gen6_render.h |
---|
File deleted |
/drivers/video/Intel-2D/gen7_render.h |
---|
File deleted |
/drivers/video/Intel-2D/kgem_debug_gen7.c |
---|
File deleted |
/drivers/video/Intel-2D/kgem_debug_gen6.c |
---|
File deleted |
/drivers/video/Intel-2D/kgem_debug.h |
---|
File deleted |
/drivers/video/Intel-2D/kgem_debug.c |
---|
File deleted |
/drivers/video/Intel-2D/sna_cpuid.h |
---|
File deleted |
/drivers/video/Intel-2D/sna_render_inline.h |
---|
File deleted |
/drivers/video/Intel-2D/sna/brw/brw.h |
---|
0,0 → 1,17 |
#include "brw_eu.h" |
bool brw_sf_kernel__nomask(struct brw_compile *p); |
bool brw_sf_kernel__mask(struct brw_compile *p); |
bool brw_wm_kernel__affine(struct brw_compile *p, int dispatch_width); |
bool brw_wm_kernel__affine_mask(struct brw_compile *p, int dispatch_width); |
bool brw_wm_kernel__affine_mask_ca(struct brw_compile *p, int dispatch_width); |
bool brw_wm_kernel__affine_mask_sa(struct brw_compile *p, int dispatch_width); |
bool brw_wm_kernel__projective(struct brw_compile *p, int dispatch_width); |
bool brw_wm_kernel__projective_mask(struct brw_compile *p, int dispatch_width); |
bool brw_wm_kernel__projective_mask_ca(struct brw_compile *p, int dispatch_width); |
bool brw_wm_kernel__projective_mask_sa(struct brw_compile *p, int dispatch_width); |
bool brw_wm_kernel__affine_opacity(struct brw_compile *p, int dispatch_width); |
bool brw_wm_kernel__projective_opacity(struct brw_compile *p, int dispatch_width); |
/drivers/video/Intel-2D/sna/brw/brw_eu.c |
---|
0,0 → 1,150 |
/* |
Copyright (C) Intel Corp. 2006. All Rights Reserved. |
Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to |
develop this 3D driver. |
Permission is hereby granted, free of charge, to any person obtaining |
a copy of this software and associated documentation files (the |
"Software"), to deal in the Software without restriction, including |
without limitation the rights to use, copy, modify, merge, publish, |
distribute, sublicense, and/or sell copies of the Software, and to |
permit persons to whom the Software is furnished to do so, subject to |
the following conditions: |
The above copyright notice and this permission notice (including the |
next paragraph) shall be included in all copies or substantial |
portions of the Software. |
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, |
EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF |
MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. |
IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE |
LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION |
OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION |
WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. |
**********************************************************************/ |
/* |
* Authors: |
* Keith Whitwell <keith@tungstengraphics.com> |
*/ |
#include "brw_eu.h" |
#include <string.h> |
#include <stdlib.h> |
/* Returns the corresponding conditional mod for swapping src0 and |
* src1 in e.g. CMP. |
*/ |
uint32_t |
brw_swap_cmod(uint32_t cmod) |
{ |
switch (cmod) { |
case BRW_CONDITIONAL_Z: |
case BRW_CONDITIONAL_NZ: |
return cmod; |
case BRW_CONDITIONAL_G: |
return BRW_CONDITIONAL_LE; |
case BRW_CONDITIONAL_GE: |
return BRW_CONDITIONAL_L; |
case BRW_CONDITIONAL_L: |
return BRW_CONDITIONAL_GE; |
case BRW_CONDITIONAL_LE: |
return BRW_CONDITIONAL_G; |
default: |
return ~0; |
} |
} |
/* How does predicate control work when execution_size != 8? Do I |
* need to test/set for 0xffff when execution_size is 16? |
*/ |
void brw_set_predicate_control_flag_value( struct brw_compile *p, unsigned value ) |
{ |
p->current->header.predicate_control = BRW_PREDICATE_NONE; |
if (value != 0xff) { |
if (value != p->flag_value) { |
brw_MOV(p, brw_flag_reg(), brw_imm_uw(value)); |
p->flag_value = value; |
} |
p->current->header.predicate_control = BRW_PREDICATE_NORMAL; |
} |
} |
void brw_set_compression_control(struct brw_compile *p, |
enum brw_compression compression_control) |
{ |
p->compressed = (compression_control == BRW_COMPRESSION_COMPRESSED); |
if (p->gen >= 060) { |
/* Since we don't use the 32-wide support in gen6, we translate |
* the pre-gen6 compression control here. |
*/ |
switch (compression_control) { |
case BRW_COMPRESSION_NONE: |
/* This is the "use the first set of bits of dmask/vmask/arf |
* according to execsize" option. |
*/ |
p->current->header.compression_control = GEN6_COMPRESSION_1Q; |
break; |
case BRW_COMPRESSION_2NDHALF: |
/* For 8-wide, this is "use the second set of 8 bits." */ |
p->current->header.compression_control = GEN6_COMPRESSION_2Q; |
break; |
case BRW_COMPRESSION_COMPRESSED: |
/* For 16-wide instruction compression, use the first set of 16 bits |
* since we don't do 32-wide dispatch. |
*/ |
p->current->header.compression_control = GEN6_COMPRESSION_1H; |
break; |
default: |
assert(!"not reached"); |
p->current->header.compression_control = GEN6_COMPRESSION_1H; |
break; |
} |
} else { |
p->current->header.compression_control = compression_control; |
} |
} |
void brw_push_insn_state( struct brw_compile *p ) |
{ |
assert(p->current != &p->stack[BRW_EU_MAX_INSN_STACK-1]); |
memcpy(p->current+1, p->current, sizeof(struct brw_instruction)); |
p->compressed_stack[p->current - p->stack] = p->compressed; |
p->current++; |
} |
void brw_pop_insn_state( struct brw_compile *p ) |
{ |
assert(p->current != p->stack); |
p->current--; |
p->compressed = p->compressed_stack[p->current - p->stack]; |
} |
void brw_compile_init(struct brw_compile *p, int gen, void *store) |
{ |
assert(gen); |
p->gen = gen; |
p->store = store; |
p->nr_insn = 0; |
p->current = p->stack; |
p->compressed = false; |
memset(p->current, 0, sizeof(p->current[0])); |
/* Some defaults? |
*/ |
brw_set_mask_control(p, BRW_MASK_ENABLE); /* what does this do? */ |
brw_set_saturate(p, 0); |
brw_set_compression_control(p, BRW_COMPRESSION_NONE); |
brw_set_predicate_control_flag_value(p, 0xff); |
p->if_stack_depth = 0; |
p->if_stack_array_size = 0; |
p->if_stack = NULL; |
} |
/drivers/video/Intel-2D/sna/brw/brw_eu.h |
---|
0,0 → 1,2264 |
/* |
Copyright (C) Intel Corp. 2006. All Rights Reserved. |
Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to |
develop this 3D driver. |
Permission is hereby granted, free of charge, to any person obtaining |
a copy of this software and associated documentation files (the |
"Software"), to deal in the Software without restriction, including |
without limitation the rights to use, copy, modify, merge, publish, |
distribute, sublicense, and/or sell copies of the Software, and to |
permit persons to whom the Software is furnished to do so, subject to |
the following conditions: |
The above copyright notice and this permission notice (including the |
next paragraph) shall be included in all copies or substantial |
portions of the Software. |
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, |
EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF |
MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. |
IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE |
LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION |
OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION |
WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. |
**********************************************************************/ |
/* |
* Authors: |
* Keith Whitwell <keith@tungstengraphics.com> |
*/ |
#ifndef BRW_EU_H |
#define BRW_EU_H |
#include <stdbool.h> |
#include <stdint.h> |
#include <stdio.h> |
#include <assert.h> |
#define BRW_SWIZZLE4(a,b,c,d) (((a)<<0) | ((b)<<2) | ((c)<<4) | ((d)<<6)) |
#define BRW_GET_SWZ(swz, idx) (((swz) >> ((idx)*2)) & 0x3) |
#define BRW_SWIZZLE_NOOP BRW_SWIZZLE4(0,1,2,3) |
#define BRW_SWIZZLE_XYZW BRW_SWIZZLE4(0,1,2,3) |
#define BRW_SWIZZLE_XXXX BRW_SWIZZLE4(0,0,0,0) |
#define BRW_SWIZZLE_YYYY BRW_SWIZZLE4(1,1,1,1) |
#define BRW_SWIZZLE_ZZZZ BRW_SWIZZLE4(2,2,2,2) |
#define BRW_SWIZZLE_WWWW BRW_SWIZZLE4(3,3,3,3) |
#define BRW_SWIZZLE_XYXY BRW_SWIZZLE4(0,1,0,1) |
#define WRITEMASK_X 0x1 |
#define WRITEMASK_Y 0x2 |
#define WRITEMASK_Z 0x4 |
#define WRITEMASK_W 0x8 |
#define WRITEMASK_XY (WRITEMASK_X | WRITEMASK_Y) |
#define WRITEMASK_XYZ (WRITEMASK_X | WRITEMASK_Y | WRITEMASK_Z) |
#define WRITEMASK_XYZW (WRITEMASK_X | WRITEMASK_Y | WRITEMASK_Z | WRITEMASK_W) |
/** Number of general purpose registers (VS, WM, etc) */ |
#define BRW_MAX_GRF 128 |
/** Number of message register file registers */ |
#define BRW_MAX_MRF 16 |
#define BRW_ALIGN_1 0 |
#define BRW_ALIGN_16 1 |
#define BRW_ADDRESS_DIRECT 0 |
#define BRW_ADDRESS_REGISTER_INDIRECT_REGISTER 1 |
#define BRW_CHANNEL_X 0 |
#define BRW_CHANNEL_Y 1 |
#define BRW_CHANNEL_Z 2 |
#define BRW_CHANNEL_W 3 |
enum brw_compression { |
BRW_COMPRESSION_NONE, |
BRW_COMPRESSION_2NDHALF, |
BRW_COMPRESSION_COMPRESSED, |
}; |
#define GEN6_COMPRESSION_1Q 0 |
#define GEN6_COMPRESSION_2Q 1 |
#define GEN6_COMPRESSION_3Q 2 |
#define GEN6_COMPRESSION_4Q 3 |
#define GEN6_COMPRESSION_1H 0 |
#define GEN6_COMPRESSION_2H 2 |
#define BRW_CONDITIONAL_NONE 0 |
#define BRW_CONDITIONAL_Z 1 |
#define BRW_CONDITIONAL_NZ 2 |
#define BRW_CONDITIONAL_EQ 1 /* Z */ |
#define BRW_CONDITIONAL_NEQ 2 /* NZ */ |
#define BRW_CONDITIONAL_G 3 |
#define BRW_CONDITIONAL_GE 4 |
#define BRW_CONDITIONAL_L 5 |
#define BRW_CONDITIONAL_LE 6 |
#define BRW_CONDITIONAL_R 7 |
#define BRW_CONDITIONAL_O 8 |
#define BRW_CONDITIONAL_U 9 |
#define BRW_DEBUG_NONE 0 |
#define BRW_DEBUG_BREAKPOINT 1 |
#define BRW_DEPENDENCY_NORMAL 0 |
#define BRW_DEPENDENCY_NOTCLEARED 1 |
#define BRW_DEPENDENCY_NOTCHECKED 2 |
#define BRW_DEPENDENCY_DISABLE 3 |
#define BRW_EXECUTE_1 0 |
#define BRW_EXECUTE_2 1 |
#define BRW_EXECUTE_4 2 |
#define BRW_EXECUTE_8 3 |
#define BRW_EXECUTE_16 4 |
#define BRW_EXECUTE_32 5 |
#define BRW_HORIZONTAL_STRIDE_0 0 |
#define BRW_HORIZONTAL_STRIDE_1 1 |
#define BRW_HORIZONTAL_STRIDE_2 2 |
#define BRW_HORIZONTAL_STRIDE_4 3 |
#define BRW_INSTRUCTION_NORMAL 0 |
#define BRW_INSTRUCTION_SATURATE 1 |
#define BRW_MASK_ENABLE 0 |
#define BRW_MASK_DISABLE 1 |
/** @{ |
* |
* Gen6 has replaced "mask enable/disable" with WECtrl, which is |
* effectively the same but much simpler to think about. Now, there |
* are two contributors ANDed together to whether channels are |
* executed: The predication on the instruction, and the channel write |
* enable. |
*/ |
/** |
* This is the default value. It means that a channel's write enable is set |
* if the per-channel IP is pointing at this instruction. |
*/ |
#define BRW_WE_NORMAL 0 |
/** |
* This is used like BRW_MASK_DISABLE, and causes all channels to have |
* their write enable set. Note that predication still contributes to |
* whether the channel actually gets written. |
*/ |
#define BRW_WE_ALL 1 |
/** @} */ |
enum opcode { |
/* These are the actual hardware opcodes. */ |
BRW_OPCODE_MOV = 1, |
BRW_OPCODE_SEL = 2, |
BRW_OPCODE_NOT = 4, |
BRW_OPCODE_AND = 5, |
BRW_OPCODE_OR = 6, |
BRW_OPCODE_XOR = 7, |
BRW_OPCODE_SHR = 8, |
BRW_OPCODE_SHL = 9, |
BRW_OPCODE_RSR = 10, |
BRW_OPCODE_RSL = 11, |
BRW_OPCODE_ASR = 12, |
BRW_OPCODE_CMP = 16, |
BRW_OPCODE_CMPN = 17, |
BRW_OPCODE_JMPI = 32, |
BRW_OPCODE_IF = 34, |
BRW_OPCODE_IFF = 35, |
BRW_OPCODE_ELSE = 36, |
BRW_OPCODE_ENDIF = 37, |
BRW_OPCODE_DO = 38, |
BRW_OPCODE_WHILE = 39, |
BRW_OPCODE_BREAK = 40, |
BRW_OPCODE_CONTINUE = 41, |
BRW_OPCODE_HALT = 42, |
BRW_OPCODE_MSAVE = 44, |
BRW_OPCODE_MRESTORE = 45, |
BRW_OPCODE_PUSH = 46, |
BRW_OPCODE_POP = 47, |
BRW_OPCODE_WAIT = 48, |
BRW_OPCODE_SEND = 49, |
BRW_OPCODE_SENDC = 50, |
BRW_OPCODE_MATH = 56, |
BRW_OPCODE_ADD = 64, |
BRW_OPCODE_MUL = 65, |
BRW_OPCODE_AVG = 66, |
BRW_OPCODE_FRC = 67, |
BRW_OPCODE_RNDU = 68, |
BRW_OPCODE_RNDD = 69, |
BRW_OPCODE_RNDE = 70, |
BRW_OPCODE_RNDZ = 71, |
BRW_OPCODE_MAC = 72, |
BRW_OPCODE_MACH = 73, |
BRW_OPCODE_LZD = 74, |
BRW_OPCODE_SAD2 = 80, |
BRW_OPCODE_SADA2 = 81, |
BRW_OPCODE_DP4 = 84, |
BRW_OPCODE_DPH = 85, |
BRW_OPCODE_DP3 = 86, |
BRW_OPCODE_DP2 = 87, |
BRW_OPCODE_DPA2 = 88, |
BRW_OPCODE_LINE = 89, |
BRW_OPCODE_PLN = 90, |
BRW_OPCODE_NOP = 126, |
/* These are compiler backend opcodes that get translated into other |
* instructions. |
*/ |
FS_OPCODE_FB_WRITE = 128, |
SHADER_OPCODE_RCP, |
SHADER_OPCODE_RSQ, |
SHADER_OPCODE_SQRT, |
SHADER_OPCODE_EXP2, |
SHADER_OPCODE_LOG2, |
SHADER_OPCODE_POW, |
SHADER_OPCODE_SIN, |
SHADER_OPCODE_COS, |
FS_OPCODE_DDX, |
FS_OPCODE_DDY, |
FS_OPCODE_PIXEL_X, |
FS_OPCODE_PIXEL_Y, |
FS_OPCODE_CINTERP, |
FS_OPCODE_LINTERP, |
FS_OPCODE_TEX, |
FS_OPCODE_TXB, |
FS_OPCODE_TXD, |
FS_OPCODE_TXF, |
FS_OPCODE_TXL, |
FS_OPCODE_TXS, |
FS_OPCODE_DISCARD, |
FS_OPCODE_SPILL, |
FS_OPCODE_UNSPILL, |
FS_OPCODE_PULL_CONSTANT_LOAD, |
VS_OPCODE_URB_WRITE, |
VS_OPCODE_SCRATCH_READ, |
VS_OPCODE_SCRATCH_WRITE, |
VS_OPCODE_PULL_CONSTANT_LOAD, |
}; |
#define BRW_PREDICATE_NONE 0 |
#define BRW_PREDICATE_NORMAL 1 |
#define BRW_PREDICATE_ALIGN1_ANYV 2 |
#define BRW_PREDICATE_ALIGN1_ALLV 3 |
#define BRW_PREDICATE_ALIGN1_ANY2H 4 |
#define BRW_PREDICATE_ALIGN1_ALL2H 5 |
#define BRW_PREDICATE_ALIGN1_ANY4H 6 |
#define BRW_PREDICATE_ALIGN1_ALL4H 7 |
#define BRW_PREDICATE_ALIGN1_ANY8H 8 |
#define BRW_PREDICATE_ALIGN1_ALL8H 9 |
#define BRW_PREDICATE_ALIGN1_ANY16H 10 |
#define BRW_PREDICATE_ALIGN1_ALL16H 11 |
#define BRW_PREDICATE_ALIGN16_REPLICATE_X 2 |
#define BRW_PREDICATE_ALIGN16_REPLICATE_Y 3 |
#define BRW_PREDICATE_ALIGN16_REPLICATE_Z 4 |
#define BRW_PREDICATE_ALIGN16_REPLICATE_W 5 |
#define BRW_PREDICATE_ALIGN16_ANY4H 6 |
#define BRW_PREDICATE_ALIGN16_ALL4H 7 |
#define BRW_ARCHITECTURE_REGISTER_FILE 0 |
#define BRW_GENERAL_REGISTER_FILE 1 |
#define BRW_MESSAGE_REGISTER_FILE 2 |
#define BRW_IMMEDIATE_VALUE 3 |
#define BRW_REGISTER_TYPE_UD 0 |
#define BRW_REGISTER_TYPE_D 1 |
#define BRW_REGISTER_TYPE_UW 2 |
#define BRW_REGISTER_TYPE_W 3 |
#define BRW_REGISTER_TYPE_UB 4 |
#define BRW_REGISTER_TYPE_B 5 |
#define BRW_REGISTER_TYPE_VF 5 /* packed float vector, immediates only? */ |
#define BRW_REGISTER_TYPE_HF 6 |
#define BRW_REGISTER_TYPE_V 6 /* packed int vector, immediates only, uword dest only */ |
#define BRW_REGISTER_TYPE_F 7 |
#define BRW_ARF_NULL 0x00 |
#define BRW_ARF_ADDRESS 0x10 |
#define BRW_ARF_ACCUMULATOR 0x20 |
#define BRW_ARF_FLAG 0x30 |
#define BRW_ARF_MASK 0x40 |
#define BRW_ARF_MASK_STACK 0x50 |
#define BRW_ARF_MASK_STACK_DEPTH 0x60 |
#define BRW_ARF_STATE 0x70 |
#define BRW_ARF_CONTROL 0x80 |
#define BRW_ARF_NOTIFICATION_COUNT 0x90 |
#define BRW_ARF_IP 0xA0 |
#define BRW_MRF_COMPR4 (1 << 7) |
#define BRW_AMASK 0 |
#define BRW_IMASK 1 |
#define BRW_LMASK 2 |
#define BRW_CMASK 3 |
#define BRW_THREAD_NORMAL 0 |
#define BRW_THREAD_ATOMIC 1 |
#define BRW_THREAD_SWITCH 2 |
#define BRW_VERTICAL_STRIDE_0 0 |
#define BRW_VERTICAL_STRIDE_1 1 |
#define BRW_VERTICAL_STRIDE_2 2 |
#define BRW_VERTICAL_STRIDE_4 3 |
#define BRW_VERTICAL_STRIDE_8 4 |
#define BRW_VERTICAL_STRIDE_16 5 |
#define BRW_VERTICAL_STRIDE_32 6 |
#define BRW_VERTICAL_STRIDE_64 7 |
#define BRW_VERTICAL_STRIDE_128 8 |
#define BRW_VERTICAL_STRIDE_256 9 |
#define BRW_VERTICAL_STRIDE_ONE_DIMENSIONAL 0xF |
#define BRW_WIDTH_1 0 |
#define BRW_WIDTH_2 1 |
#define BRW_WIDTH_4 2 |
#define BRW_WIDTH_8 3 |
#define BRW_WIDTH_16 4 |
#define BRW_STATELESS_BUFFER_BOUNDARY_1K 0 |
#define BRW_STATELESS_BUFFER_BOUNDARY_2K 1 |
#define BRW_STATELESS_BUFFER_BOUNDARY_4K 2 |
#define BRW_STATELESS_BUFFER_BOUNDARY_8K 3 |
#define BRW_STATELESS_BUFFER_BOUNDARY_16K 4 |
#define BRW_STATELESS_BUFFER_BOUNDARY_32K 5 |
#define BRW_STATELESS_BUFFER_BOUNDARY_64K 6 |
#define BRW_STATELESS_BUFFER_BOUNDARY_128K 7 |
#define BRW_STATELESS_BUFFER_BOUNDARY_256K 8 |
#define BRW_STATELESS_BUFFER_BOUNDARY_512K 9 |
#define BRW_STATELESS_BUFFER_BOUNDARY_1M 10 |
#define BRW_STATELESS_BUFFER_BOUNDARY_2M 11 |
#define BRW_POLYGON_FACING_FRONT 0 |
#define BRW_POLYGON_FACING_BACK 1 |
#define BRW_MESSAGE_TARGET_NULL 0 |
#define BRW_MESSAGE_TARGET_MATH 1 /* reserved on GEN6 */ |
#define BRW_MESSAGE_TARGET_SAMPLER 2 |
#define BRW_MESSAGE_TARGET_GATEWAY 3 |
#define BRW_MESSAGE_TARGET_DATAPORT_READ 4 |
#define BRW_MESSAGE_TARGET_DATAPORT_WRITE 5 |
#define BRW_MESSAGE_TARGET_URB 6 |
#define BRW_MESSAGE_TARGET_THREAD_SPAWNER 7 |
#define GEN6_MESSAGE_TARGET_DP_SAMPLER_CACHE 4 |
#define GEN6_MESSAGE_TARGET_DP_RENDER_CACHE 5 |
#define GEN6_MESSAGE_TARGET_DP_CONST_CACHE 9 |
#define BRW_SAMPLER_RETURN_FORMAT_FLOAT32 0 |
#define BRW_SAMPLER_RETURN_FORMAT_UINT32 2 |
#define BRW_SAMPLER_RETURN_FORMAT_SINT32 3 |
#define BRW_SAMPLER_MESSAGE_SAMPLE 0 |
#define BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE 0 |
#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE 0 |
#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS 0 |
#define BRW_SAMPLER_MESSAGE_SIMD8_KILLPIX 1 |
#define BRW_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_LOD 1 |
#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_LOD 1 |
#define BRW_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_GRADIENTS 2 |
#define BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_GRADIENTS 2 |
#define BRW_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_COMPARE 0 |
#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_COMPARE 2 |
#define BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_BIAS_COMPARE 0 |
#define BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_LOD_COMPARE 1 |
#define BRW_SAMPLER_MESSAGE_SIMD4X2_RESINFO 2 |
#define BRW_SAMPLER_MESSAGE_SIMD16_RESINFO 2 |
#define BRW_SAMPLER_MESSAGE_SIMD4X2_LD 3 |
#define BRW_SAMPLER_MESSAGE_SIMD8_LD 3 |
#define BRW_SAMPLER_MESSAGE_SIMD16_LD 3 |
#define GEN5_SAMPLER_MESSAGE_SAMPLE 0 |
#define GEN5_SAMPLER_MESSAGE_SAMPLE_BIAS 1 |
#define GEN5_SAMPLER_MESSAGE_SAMPLE_LOD 2 |
#define GEN5_SAMPLER_MESSAGE_SAMPLE_COMPARE 3 |
#define GEN5_SAMPLER_MESSAGE_SAMPLE_DERIVS 4 |
#define GEN5_SAMPLER_MESSAGE_SAMPLE_BIAS_COMPARE 5 |
#define GEN5_SAMPLER_MESSAGE_SAMPLE_LOD_COMPARE 6 |
#define GEN5_SAMPLER_MESSAGE_SAMPLE_LD 7 |
#define GEN5_SAMPLER_MESSAGE_SAMPLE_RESINFO 10 |
/* for GEN5 only */ |
#define BRW_SAMPLER_SIMD_MODE_SIMD4X2 0 |
#define BRW_SAMPLER_SIMD_MODE_SIMD8 1 |
#define BRW_SAMPLER_SIMD_MODE_SIMD16 2 |
#define BRW_SAMPLER_SIMD_MODE_SIMD32_64 3 |
#define BRW_DATAPORT_OWORD_BLOCK_1_OWORDLOW 0 |
#define BRW_DATAPORT_OWORD_BLOCK_1_OWORDHIGH 1 |
#define BRW_DATAPORT_OWORD_BLOCK_2_OWORDS 2 |
#define BRW_DATAPORT_OWORD_BLOCK_4_OWORDS 3 |
#define BRW_DATAPORT_OWORD_BLOCK_8_OWORDS 4 |
#define BRW_DATAPORT_OWORD_DUAL_BLOCK_1OWORD 0 |
#define BRW_DATAPORT_OWORD_DUAL_BLOCK_4OWORDS 2 |
#define BRW_DATAPORT_DWORD_SCATTERED_BLOCK_8DWORDS 2 |
#define BRW_DATAPORT_DWORD_SCATTERED_BLOCK_16DWORDS 3 |
/* This one stays the same across generations. */ |
#define BRW_DATAPORT_READ_MESSAGE_OWORD_BLOCK_READ 0 |
/* GEN4 */ |
#define BRW_DATAPORT_READ_MESSAGE_OWORD_DUAL_BLOCK_READ 1 |
#define BRW_DATAPORT_READ_MESSAGE_MEDIA_BLOCK_READ 2 |
#define BRW_DATAPORT_READ_MESSAGE_DWORD_SCATTERED_READ 3 |
/* G45, GEN5 */ |
#define G45_DATAPORT_READ_MESSAGE_RENDER_UNORM_READ 1 |
#define G45_DATAPORT_READ_MESSAGE_OWORD_DUAL_BLOCK_READ 2 |
#define G45_DATAPORT_READ_MESSAGE_AVC_LOOP_FILTER_READ 3 |
#define G45_DATAPORT_READ_MESSAGE_MEDIA_BLOCK_READ 4 |
#define G45_DATAPORT_READ_MESSAGE_DWORD_SCATTERED_READ 6 |
/* GEN6 */ |
#define GEN6_DATAPORT_READ_MESSAGE_RENDER_UNORM_READ 1 |
#define GEN6_DATAPORT_READ_MESSAGE_OWORD_DUAL_BLOCK_READ 2 |
#define GEN6_DATAPORT_READ_MESSAGE_MEDIA_BLOCK_READ 4 |
#define GEN6_DATAPORT_READ_MESSAGE_OWORD_UNALIGN_BLOCK_READ 5 |
#define GEN6_DATAPORT_READ_MESSAGE_DWORD_SCATTERED_READ 6 |
#define BRW_DATAPORT_READ_TARGET_DATA_CACHE 0 |
#define BRW_DATAPORT_READ_TARGET_RENDER_CACHE 1 |
#define BRW_DATAPORT_READ_TARGET_SAMPLER_CACHE 2 |
#define BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD16_SINGLE_SOURCE 0 |
#define BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD16_SINGLE_SOURCE_REPLICATED 1 |
#define BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD8_DUAL_SOURCE_SUBSPAN01 2 |
#define BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD8_DUAL_SOURCE_SUBSPAN23 3 |
#define BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD8_SINGLE_SOURCE_SUBSPAN01 4 |
/** |
* Message target: Shared Function ID for where to SEND a message. |
* |
* These are enumerated in the ISA reference under "send - Send Message". |
* In particular, see the following tables: |
* - G45 PRM, Volume 4, Table 14-15 "Message Descriptor Definition" |
* - Sandybridge PRM, Volume 4 Part 2, Table 8-16 "Extended Message Descriptor" |
* - BSpec, Volume 1a (GPU Overview) / Graphics Processing Engine (GPE) / |
* Overview / GPE Function IDs |
*/ |
enum brw_message_target { |
BRW_SFID_NULL = 0, |
BRW_SFID_MATH = 1, /* Only valid on Gen4-5 */ |
BRW_SFID_SAMPLER = 2, |
BRW_SFID_MESSAGE_GATEWAY = 3, |
BRW_SFID_DATAPORT_READ = 4, |
BRW_SFID_DATAPORT_WRITE = 5, |
BRW_SFID_URB = 6, |
BRW_SFID_THREAD_SPAWNER = 7, |
GEN6_SFID_DATAPORT_SAMPLER_CACHE = 4, |
GEN6_SFID_DATAPORT_RENDER_CACHE = 5, |
GEN6_SFID_DATAPORT_CONSTANT_CACHE = 9, |
GEN7_SFID_DATAPORT_DATA_CACHE = 10, |
}; |
#define GEN7_MESSAGE_TARGET_DP_DATA_CACHE 10 |
#define BRW_DATAPORT_WRITE_MESSAGE_OWORD_BLOCK_WRITE 0 |
#define BRW_DATAPORT_WRITE_MESSAGE_OWORD_DUAL_BLOCK_WRITE 1 |
#define BRW_DATAPORT_WRITE_MESSAGE_MEDIA_BLOCK_WRITE 2 |
#define BRW_DATAPORT_WRITE_MESSAGE_DWORD_SCATTERED_WRITE 3 |
#define BRW_DATAPORT_WRITE_MESSAGE_RENDER_TARGET_WRITE 4 |
#define BRW_DATAPORT_WRITE_MESSAGE_STREAMED_VERTEX_BUFFER_WRITE 5 |
#define BRW_DATAPORT_WRITE_MESSAGE_FLUSH_RENDER_CACHE 7 |
/* GEN6 */ |
#define GEN6_DATAPORT_WRITE_MESSAGE_DWORD_ATOMIC_WRITE 7 |
#define GEN6_DATAPORT_WRITE_MESSAGE_OWORD_BLOCK_WRITE 8 |
#define GEN6_DATAPORT_WRITE_MESSAGE_OWORD_DUAL_BLOCK_WRITE 9 |
#define GEN6_DATAPORT_WRITE_MESSAGE_MEDIA_BLOCK_WRITE 10 |
#define GEN6_DATAPORT_WRITE_MESSAGE_DWORD_SCATTERED_WRITE 11 |
#define GEN6_DATAPORT_WRITE_MESSAGE_RENDER_TARGET_WRITE 12 |
#define GEN6_DATAPORT_WRITE_MESSAGE_STREAMED_VB_WRITE 13 |
#define GEN6_DATAPORT_WRITE_MESSAGE_RENDER_TARGET_UNORM_WRITE 14 |
#define BRW_MATH_FUNCTION_INV 1 |
#define BRW_MATH_FUNCTION_LOG 2 |
#define BRW_MATH_FUNCTION_EXP 3 |
#define BRW_MATH_FUNCTION_SQRT 4 |
#define BRW_MATH_FUNCTION_RSQ 5 |
#define BRW_MATH_FUNCTION_SIN 6 /* was 7 */ |
#define BRW_MATH_FUNCTION_COS 7 /* was 8 */ |
#define BRW_MATH_FUNCTION_SINCOS 8 /* was 6 */ |
#define BRW_MATH_FUNCTION_TAN 9 /* gen4 */ |
#define BRW_MATH_FUNCTION_FDIV 9 /* gen6+ */ |
#define BRW_MATH_FUNCTION_POW 10 |
#define BRW_MATH_FUNCTION_INT_DIV_QUOTIENT_AND_REMAINDER 11 |
#define BRW_MATH_FUNCTION_INT_DIV_QUOTIENT 12 |
#define BRW_MATH_FUNCTION_INT_DIV_REMAINDER 13 |
#define BRW_MATH_INTEGER_UNSIGNED 0 |
#define BRW_MATH_INTEGER_SIGNED 1 |
#define BRW_MATH_PRECISION_FULL 0 |
#define BRW_MATH_PRECISION_PARTIAL 1 |
#define BRW_MATH_SATURATE_NONE 0 |
#define BRW_MATH_SATURATE_SATURATE 1 |
#define BRW_MATH_DATA_VECTOR 0 |
#define BRW_MATH_DATA_SCALAR 1 |
#define BRW_URB_OPCODE_WRITE 0 |
#define BRW_URB_SWIZZLE_NONE 0 |
#define BRW_URB_SWIZZLE_INTERLEAVE 1 |
#define BRW_URB_SWIZZLE_TRANSPOSE 2 |
#define BRW_SCRATCH_SPACE_SIZE_1K 0 |
#define BRW_SCRATCH_SPACE_SIZE_2K 1 |
#define BRW_SCRATCH_SPACE_SIZE_4K 2 |
#define BRW_SCRATCH_SPACE_SIZE_8K 3 |
#define BRW_SCRATCH_SPACE_SIZE_16K 4 |
#define BRW_SCRATCH_SPACE_SIZE_32K 5 |
#define BRW_SCRATCH_SPACE_SIZE_64K 6 |
#define BRW_SCRATCH_SPACE_SIZE_128K 7 |
#define BRW_SCRATCH_SPACE_SIZE_256K 8 |
#define BRW_SCRATCH_SPACE_SIZE_512K 9 |
#define BRW_SCRATCH_SPACE_SIZE_1M 10 |
#define BRW_SCRATCH_SPACE_SIZE_2M 11 |
#define REG_SIZE (8*4) |
struct brw_instruction { |
struct { |
unsigned opcode:7; |
unsigned pad:1; |
unsigned access_mode:1; |
unsigned mask_control:1; |
unsigned dependency_control:2; |
unsigned compression_control:2; /* gen6: quater control */ |
unsigned thread_control:2; |
unsigned predicate_control:4; |
unsigned predicate_inverse:1; |
unsigned execution_size:3; |
/** |
* Conditional Modifier for most instructions. On Gen6+, this is also |
* used for the SEND instruction's Message Target/SFID. |
*/ |
unsigned destreg__conditionalmod:4; |
unsigned acc_wr_control:1; |
unsigned cmpt_control:1; |
unsigned debug_control:1; |
unsigned saturate:1; |
} header; |
union { |
struct { |
unsigned dest_reg_file:2; |
unsigned dest_reg_type:3; |
unsigned src0_reg_file:2; |
unsigned src0_reg_type:3; |
unsigned src1_reg_file:2; |
unsigned src1_reg_type:3; |
unsigned pad:1; |
unsigned dest_subreg_nr:5; |
unsigned dest_reg_nr:8; |
unsigned dest_horiz_stride:2; |
unsigned dest_address_mode:1; |
} da1; |
struct { |
unsigned dest_reg_file:2; |
unsigned dest_reg_type:3; |
unsigned src0_reg_file:2; |
unsigned src0_reg_type:3; |
unsigned src1_reg_file:2; /* 0x00000c00 */ |
unsigned src1_reg_type:3; /* 0x00007000 */ |
unsigned pad:1; |
int dest_indirect_offset:10; /* offset against the deref'd address reg */ |
unsigned dest_subreg_nr:3; /* subnr for the address reg a0.x */ |
unsigned dest_horiz_stride:2; |
unsigned dest_address_mode:1; |
} ia1; |
struct { |
unsigned dest_reg_file:2; |
unsigned dest_reg_type:3; |
unsigned src0_reg_file:2; |
unsigned src0_reg_type:3; |
unsigned src1_reg_file:2; |
unsigned src1_reg_type:3; |
unsigned pad:1; |
unsigned dest_writemask:4; |
unsigned dest_subreg_nr:1; |
unsigned dest_reg_nr:8; |
unsigned dest_horiz_stride:2; |
unsigned dest_address_mode:1; |
} da16; |
struct { |
unsigned dest_reg_file:2; |
unsigned dest_reg_type:3; |
unsigned src0_reg_file:2; |
unsigned src0_reg_type:3; |
unsigned pad0:6; |
unsigned dest_writemask:4; |
int dest_indirect_offset:6; |
unsigned dest_subreg_nr:3; |
unsigned dest_horiz_stride:2; |
unsigned dest_address_mode:1; |
} ia16; |
struct { |
unsigned dest_reg_file:2; |
unsigned dest_reg_type:3; |
unsigned src0_reg_file:2; |
unsigned src0_reg_type:3; |
unsigned src1_reg_file:2; |
unsigned src1_reg_type:3; |
unsigned pad:1; |
int jump_count:16; |
} branch_gen6; |
struct { |
unsigned dest_reg_file:1; |
unsigned flag_subreg_num:1; |
unsigned pad0:2; |
unsigned src0_abs:1; |
unsigned src0_negate:1; |
unsigned src1_abs:1; |
unsigned src1_negate:1; |
unsigned src2_abs:1; |
unsigned src2_negate:1; |
unsigned pad1:7; |
unsigned dest_writemask:4; |
unsigned dest_subreg_nr:3; |
unsigned dest_reg_nr:8; |
} da3src; |
} bits1; |
union { |
struct { |
unsigned src0_subreg_nr:5; |
unsigned src0_reg_nr:8; |
unsigned src0_abs:1; |
unsigned src0_negate:1; |
unsigned src0_address_mode:1; |
unsigned src0_horiz_stride:2; |
unsigned src0_width:3; |
unsigned src0_vert_stride:4; |
unsigned flag_subreg_nr:1; |
unsigned flag_reg_nr:1; |
unsigned pad:5; |
} da1; |
struct { |
int src0_indirect_offset:10; |
unsigned src0_subreg_nr:3; |
unsigned src0_abs:1; |
unsigned src0_negate:1; |
unsigned src0_address_mode:1; |
unsigned src0_horiz_stride:2; |
unsigned src0_width:3; |
unsigned src0_vert_stride:4; |
unsigned flag_subreg_nr:1; |
unsigned flag_reg_nr:1; |
unsigned pad:5; |
} ia1; |
struct { |
unsigned src0_swz_x:2; |
unsigned src0_swz_y:2; |
unsigned src0_subreg_nr:1; |
unsigned src0_reg_nr:8; |
unsigned src0_abs:1; |
unsigned src0_negate:1; |
unsigned src0_address_mode:1; |
unsigned src0_swz_z:2; |
unsigned src0_swz_w:2; |
unsigned pad0:1; |
unsigned src0_vert_stride:4; |
unsigned flag_subreg_nr:1; |
unsigned flag_reg_nr:1; |
unsigned pad1:5; |
} da16; |
struct { |
unsigned src0_swz_x:2; |
unsigned src0_swz_y:2; |
int src0_indirect_offset:6; |
unsigned src0_subreg_nr:3; |
unsigned src0_abs:1; |
unsigned src0_negate:1; |
unsigned src0_address_mode:1; |
unsigned src0_swz_z:2; |
unsigned src0_swz_w:2; |
unsigned pad0:1; |
unsigned src0_vert_stride:4; |
unsigned flag_subreg_nr:1; |
unsigned flag_reg_nr:1; |
unsigned pad1:5; |
} ia16; |
/* Extended Message Descriptor for Ironlake (Gen5) SEND instruction. |
* |
* Does not apply to Gen6+. The SFID/message target moved to bits |
* 27:24 of the header (destreg__conditionalmod); EOT is in bits3. |
*/ |
struct { |
unsigned pad:26; |
unsigned end_of_thread:1; |
unsigned pad1:1; |
unsigned sfid:4; |
} send_gen5; /* for Ironlake only */ |
struct { |
unsigned src0_rep_ctrl:1; |
unsigned src0_swizzle:8; |
unsigned src0_subreg_nr:3; |
unsigned src0_reg_nr:8; |
unsigned pad0:1; |
unsigned src1_rep_ctrl:1; |
unsigned src1_swizzle:8; |
unsigned src1_subreg_nr_low:2; |
} da3src; |
} bits2; |
union { |
struct { |
unsigned src1_subreg_nr:5; |
unsigned src1_reg_nr:8; |
unsigned src1_abs:1; |
unsigned src1_negate:1; |
unsigned src1_address_mode:1; |
unsigned src1_horiz_stride:2; |
unsigned src1_width:3; |
unsigned src1_vert_stride:4; |
unsigned pad0:7; |
} da1; |
struct { |
unsigned src1_swz_x:2; |
unsigned src1_swz_y:2; |
unsigned src1_subreg_nr:1; |
unsigned src1_reg_nr:8; |
unsigned src1_abs:1; |
unsigned src1_negate:1; |
unsigned src1_address_mode:1; |
unsigned src1_swz_z:2; |
unsigned src1_swz_w:2; |
unsigned pad1:1; |
unsigned src1_vert_stride:4; |
unsigned pad2:7; |
} da16; |
struct { |
int src1_indirect_offset:10; |
unsigned src1_subreg_nr:3; |
unsigned src1_abs:1; |
unsigned src1_negate:1; |
unsigned src1_address_mode:1; |
unsigned src1_horiz_stride:2; |
unsigned src1_width:3; |
unsigned src1_vert_stride:4; |
unsigned flag_subreg_nr:1; |
unsigned flag_reg_nr:1; |
unsigned pad1:5; |
} ia1; |
struct { |
unsigned src1_swz_x:2; |
unsigned src1_swz_y:2; |
int src1_indirect_offset:6; |
unsigned src1_subreg_nr:3; |
unsigned src1_abs:1; |
unsigned src1_negate:1; |
unsigned pad0:1; |
unsigned src1_swz_z:2; |
unsigned src1_swz_w:2; |
unsigned pad1:1; |
unsigned src1_vert_stride:4; |
unsigned flag_subreg_nr:1; |
unsigned flag_reg_nr:1; |
unsigned pad2:5; |
} ia16; |
struct { |
int jump_count:16; /* note: signed */ |
unsigned pop_count:4; |
unsigned pad0:12; |
} if_else; |
/* This is also used for gen7 IF/ELSE instructions */ |
struct { |
/* Signed jump distance to the ip to jump to if all channels |
* are disabled after the break or continue. It should point |
* to the end of the innermost control flow block, as that's |
* where some channel could get re-enabled. |
*/ |
int jip:16; |
/* Signed jump distance to the location to resume execution |
* of this channel if it's enabled for the break or continue. |
*/ |
int uip:16; |
} break_cont; |
/** |
* \defgroup SEND instructions / Message Descriptors |
* |
* @{ |
*/ |
/** |
* Generic Message Descriptor for Gen4 SEND instructions. The structs |
* below expand function_control to something specific for their |
* message. Due to struct packing issues, they duplicate these bits. |
* |
* See the G45 PRM, Volume 4, Table 14-15. |
*/ |
struct { |
unsigned function_control:16; |
unsigned response_length:4; |
unsigned msg_length:4; |
unsigned msg_target:4; |
unsigned pad1:3; |
unsigned end_of_thread:1; |
} generic; |
/** |
* Generic Message Descriptor for Gen5-7 SEND instructions. |
* |
* See the Sandybridge PRM, Volume 2 Part 2, Table 8-15. (Sadly, most |
* of the information on the SEND instruction is missing from the public |
* Ironlake PRM.) |
* |
* The table claims that bit 31 is reserved/MBZ on Gen6+, but it lies. |
* According to the SEND instruction description: |
* "The MSb of the message description, the EOT field, always comes from |
* bit 127 of the instruction word"...which is bit 31 of this field. |
*/ |
struct { |
unsigned function_control:19; |
unsigned header_present:1; |
unsigned response_length:5; |
unsigned msg_length:4; |
unsigned pad1:2; |
unsigned end_of_thread:1; |
} generic_gen5; |
/** G45 PRM, Volume 4, Section 6.1.1.1 */ |
struct { |
unsigned function:4; |
unsigned int_type:1; |
unsigned precision:1; |
unsigned saturate:1; |
unsigned data_type:1; |
unsigned pad0:8; |
unsigned response_length:4; |
unsigned msg_length:4; |
unsigned msg_target:4; |
unsigned pad1:3; |
unsigned end_of_thread:1; |
} math; |
/** Ironlake PRM, Volume 4 Part 1, Section 6.1.1.1 */ |
struct { |
unsigned function:4; |
unsigned int_type:1; |
unsigned precision:1; |
unsigned saturate:1; |
unsigned data_type:1; |
unsigned snapshot:1; |
unsigned pad0:10; |
unsigned header_present:1; |
unsigned response_length:5; |
unsigned msg_length:4; |
unsigned pad1:2; |
unsigned end_of_thread:1; |
} math_gen5; |
/** G45 PRM, Volume 4, Section 4.8.1.1.1 [DevBW] and [DevCL] */ |
struct { |
unsigned binding_table_index:8; |
unsigned sampler:4; |
unsigned return_format:2; |
unsigned msg_type:2; |
unsigned response_length:4; |
unsigned msg_length:4; |
unsigned msg_target:4; |
unsigned pad1:3; |
unsigned end_of_thread:1; |
} sampler; |
/** G45 PRM, Volume 4, Section 4.8.1.1.2 [DevCTG] */ |
struct { |
unsigned binding_table_index:8; |
unsigned sampler:4; |
unsigned msg_type:4; |
unsigned response_length:4; |
unsigned msg_length:4; |
unsigned msg_target:4; |
unsigned pad1:3; |
unsigned end_of_thread:1; |
} sampler_g4x; |
/** Ironlake PRM, Volume 4 Part 1, Section 4.11.1.1.3 */ |
struct { |
unsigned binding_table_index:8; |
unsigned sampler:4; |
unsigned msg_type:4; |
unsigned simd_mode:2; |
unsigned pad0:1; |
unsigned header_present:1; |
unsigned response_length:5; |
unsigned msg_length:4; |
unsigned pad1:2; |
unsigned end_of_thread:1; |
} sampler_gen5; |
struct { |
unsigned binding_table_index:8; |
unsigned sampler:4; |
unsigned msg_type:5; |
unsigned simd_mode:2; |
unsigned header_present:1; |
unsigned response_length:5; |
unsigned msg_length:4; |
unsigned pad1:2; |
unsigned end_of_thread:1; |
} sampler_gen7; |
struct brw_urb_immediate { |
unsigned opcode:4; |
unsigned offset:6; |
unsigned swizzle_control:2; |
unsigned pad:1; |
unsigned allocate:1; |
unsigned used:1; |
unsigned complete:1; |
unsigned response_length:4; |
unsigned msg_length:4; |
unsigned msg_target:4; |
unsigned pad1:3; |
unsigned end_of_thread:1; |
} urb; |
struct { |
unsigned opcode:4; |
unsigned offset:6; |
unsigned swizzle_control:2; |
unsigned pad:1; |
unsigned allocate:1; |
unsigned used:1; |
unsigned complete:1; |
unsigned pad0:3; |
unsigned header_present:1; |
unsigned response_length:5; |
unsigned msg_length:4; |
unsigned pad1:2; |
unsigned end_of_thread:1; |
} urb_gen5; |
struct { |
unsigned opcode:3; |
unsigned offset:11; |
unsigned swizzle_control:1; |
unsigned complete:1; |
unsigned per_slot_offset:1; |
unsigned pad0:2; |
unsigned header_present:1; |
unsigned response_length:5; |
unsigned msg_length:4; |
unsigned pad1:2; |
unsigned end_of_thread:1; |
} urb_gen7; |
/** 965 PRM, Volume 4, Section 5.10.1.1: Message Descriptor */ |
struct { |
unsigned binding_table_index:8; |
unsigned msg_control:4; |
unsigned msg_type:2; |
unsigned target_cache:2; |
unsigned response_length:4; |
unsigned msg_length:4; |
unsigned msg_target:4; |
unsigned pad1:3; |
unsigned end_of_thread:1; |
} dp_read; |
/** G45 PRM, Volume 4, Section 5.10.1.1.2 */ |
struct { |
unsigned binding_table_index:8; |
unsigned msg_control:3; |
unsigned msg_type:3; |
unsigned target_cache:2; |
unsigned response_length:4; |
unsigned msg_length:4; |
unsigned msg_target:4; |
unsigned pad1:3; |
unsigned end_of_thread:1; |
} dp_read_g4x; |
/** Ironlake PRM, Volume 4 Part 1, Section 5.10.2.1.2. */ |
struct { |
unsigned binding_table_index:8; |
unsigned msg_control:3; |
unsigned msg_type:3; |
unsigned target_cache:2; |
unsigned pad0:3; |
unsigned header_present:1; |
unsigned response_length:5; |
unsigned msg_length:4; |
unsigned pad1:2; |
unsigned end_of_thread:1; |
} dp_read_gen5; |
/** G45 PRM, Volume 4, Section 5.10.1.1.2. For both Gen4 and G45. */ |
struct { |
unsigned binding_table_index:8; |
unsigned msg_control:3; |
unsigned last_render_target:1; |
unsigned msg_type:3; |
unsigned send_commit_msg:1; |
unsigned response_length:4; |
unsigned msg_length:4; |
unsigned msg_target:4; |
unsigned pad1:3; |
unsigned end_of_thread:1; |
} dp_write; |
/** Ironlake PRM, Volume 4 Part 1, Section 5.10.2.1.2. */ |
struct { |
unsigned binding_table_index:8; |
unsigned msg_control:3; |
unsigned last_render_target:1; |
unsigned msg_type:3; |
unsigned send_commit_msg:1; |
unsigned pad0:3; |
unsigned header_present:1; |
unsigned response_length:5; |
unsigned msg_length:4; |
unsigned pad1:2; |
unsigned end_of_thread:1; |
} dp_write_gen5; |
/** |
* Message for the Sandybridge Sampler Cache or Constant Cache Data Port. |
* |
* See the Sandybridge PRM, Volume 4 Part 1, Section 3.9.2.1.1. |
**/ |
struct { |
unsigned binding_table_index:8; |
unsigned msg_control:5; |
unsigned msg_type:3; |
unsigned pad0:3; |
unsigned header_present:1; |
unsigned response_length:5; |
unsigned msg_length:4; |
unsigned pad1:2; |
unsigned end_of_thread:1; |
} gen6_dp_sampler_const_cache; |
/** |
* Message for the Sandybridge Render Cache Data Port. |
* |
* Most fields are defined in the Sandybridge PRM, Volume 4 Part 1, |
* Section 3.9.2.1.1: Message Descriptor. |
* |
* "Slot Group Select" and "Last Render Target" are part of the |
* 5-bit message control for Render Target Write messages. See |
* Section 3.9.9.2.1 of the same volume. |
*/ |
struct { |
unsigned binding_table_index:8; |
unsigned msg_control:3; |
unsigned slot_group_select:1; |
unsigned last_render_target:1; |
unsigned msg_type:4; |
unsigned send_commit_msg:1; |
unsigned pad0:1; |
unsigned header_present:1; |
unsigned response_length:5; |
unsigned msg_length:4; |
unsigned pad1:2; |
unsigned end_of_thread:1; |
} gen6_dp; |
/** |
* Message for any of the Gen7 Data Port caches. |
* |
* Most fields are defined in BSpec volume 5c.2 Data Port / Messages / |
* Data Port Messages / Message Descriptor. Once again, "Slot Group |
* Select" and "Last Render Target" are part of the 6-bit message |
* control for Render Target Writes. |
*/ |
struct { |
unsigned binding_table_index:8; |
unsigned msg_control:3; |
unsigned slot_group_select:1; |
unsigned last_render_target:1; |
unsigned msg_control_pad:1; |
unsigned msg_type:4; |
unsigned pad1:1; |
unsigned header_present:1; |
unsigned response_length:5; |
unsigned msg_length:4; |
unsigned pad2:2; |
unsigned end_of_thread:1; |
} gen7_dp; |
/** @} */ |
struct { |
unsigned src1_subreg_nr_high:1; |
unsigned src1_reg_nr:8; |
unsigned pad0:1; |
unsigned src2_rep_ctrl:1; |
unsigned src2_swizzle:8; |
unsigned src2_subreg_nr:3; |
unsigned src2_reg_nr:8; |
unsigned pad1:2; |
} da3src; |
int d; |
unsigned ud; |
float f; |
} bits3; |
}; |
/* These aren't hardware structs, just something useful for us to pass around: |
* |
* Align1 operation has a lot of control over input ranges. Used in |
* WM programs to implement shaders decomposed into "channel serial" |
* or "structure of array" form: |
*/ |
struct brw_reg { |
unsigned type:4; |
unsigned file:2; |
unsigned nr:8; |
unsigned subnr:5; /* :1 in align16 */ |
unsigned negate:1; /* source only */ |
unsigned abs:1; /* source only */ |
unsigned vstride:4; /* source only */ |
unsigned width:3; /* src only, align1 only */ |
unsigned hstride:2; /* align1 only */ |
unsigned address_mode:1; /* relative addressing, hopefully! */ |
unsigned pad0:1; |
union { |
struct { |
unsigned swizzle:8; /* src only, align16 only */ |
unsigned writemask:4; /* dest only, align16 only */ |
int indirect_offset:10; /* relative addressing offset */ |
unsigned pad1:10; /* two dwords total */ |
} bits; |
float f; |
int d; |
unsigned ud; |
} dw1; |
}; |
struct brw_indirect { |
unsigned addr_subnr:4; |
int addr_offset:10; |
unsigned pad:18; |
}; |
#define BRW_EU_MAX_INSN_STACK 5 |
#define BRW_EU_MAX_INSN 10000 |
struct brw_compile { |
struct brw_instruction *store; |
unsigned nr_insn; |
int gen; |
/* Allow clients to push/pop instruction state: |
*/ |
struct brw_instruction stack[BRW_EU_MAX_INSN_STACK]; |
bool compressed_stack[BRW_EU_MAX_INSN_STACK]; |
struct brw_instruction *current; |
unsigned flag_value; |
bool single_program_flow; |
bool compressed; |
/* Control flow stacks: |
* - if_stack contains IF and ELSE instructions which must be patched |
* (and popped) once the matching ENDIF instruction is encountered. |
*/ |
struct brw_instruction **if_stack; |
int if_stack_depth; |
int if_stack_array_size; |
}; |
static inline int type_sz(unsigned type) |
{ |
switch (type) { |
case BRW_REGISTER_TYPE_UD: |
case BRW_REGISTER_TYPE_D: |
case BRW_REGISTER_TYPE_F: |
return 4; |
case BRW_REGISTER_TYPE_HF: |
case BRW_REGISTER_TYPE_UW: |
case BRW_REGISTER_TYPE_W: |
return 2; |
case BRW_REGISTER_TYPE_UB: |
case BRW_REGISTER_TYPE_B: |
return 1; |
default: |
return 0; |
} |
} |
/** |
* Construct a brw_reg. |
* \param file one of the BRW_x_REGISTER_FILE values |
* \param nr register number/index |
* \param subnr register sub number |
* \param type one of BRW_REGISTER_TYPE_x |
* \param vstride one of BRW_VERTICAL_STRIDE_x |
* \param width one of BRW_WIDTH_x |
* \param hstride one of BRW_HORIZONTAL_STRIDE_x |
* \param swizzle one of BRW_SWIZZLE_x |
* \param writemask WRITEMASK_X/Y/Z/W bitfield |
*/ |
static inline struct brw_reg brw_reg(unsigned file, |
unsigned nr, |
unsigned subnr, |
unsigned type, |
unsigned vstride, |
unsigned width, |
unsigned hstride, |
unsigned swizzle, |
unsigned writemask) |
{ |
struct brw_reg reg; |
if (file == BRW_GENERAL_REGISTER_FILE) |
assert(nr < BRW_MAX_GRF); |
else if (file == BRW_MESSAGE_REGISTER_FILE) |
assert((nr & ~(1 << 7)) < BRW_MAX_MRF); |
else if (file == BRW_ARCHITECTURE_REGISTER_FILE) |
assert(nr <= BRW_ARF_IP); |
reg.type = type; |
reg.file = file; |
reg.nr = nr; |
reg.subnr = subnr * type_sz(type); |
reg.negate = 0; |
reg.abs = 0; |
reg.vstride = vstride; |
reg.width = width; |
reg.hstride = hstride; |
reg.address_mode = BRW_ADDRESS_DIRECT; |
reg.pad0 = 0; |
/* Could do better: If the reg is r5.3<0;1,0>, we probably want to |
* set swizzle and writemask to W, as the lower bits of subnr will |
* be lost when converted to align16. This is probably too much to |
* keep track of as you'd want it adjusted by suboffset(), etc. |
* Perhaps fix up when converting to align16? |
*/ |
reg.dw1.bits.swizzle = swizzle; |
reg.dw1.bits.writemask = writemask; |
reg.dw1.bits.indirect_offset = 0; |
reg.dw1.bits.pad1 = 0; |
return reg; |
} |
/** Construct float[16] register */ |
static inline struct brw_reg brw_vec16_reg(unsigned file, |
unsigned nr, |
unsigned subnr) |
{ |
return brw_reg(file, |
nr, |
subnr, |
BRW_REGISTER_TYPE_F, |
BRW_VERTICAL_STRIDE_16, |
BRW_WIDTH_16, |
BRW_HORIZONTAL_STRIDE_1, |
BRW_SWIZZLE_XYZW, |
WRITEMASK_XYZW); |
} |
/** Construct float[8] register */ |
static inline struct brw_reg brw_vec8_reg(unsigned file, |
unsigned nr, |
unsigned subnr) |
{ |
return brw_reg(file, |
nr, |
subnr, |
BRW_REGISTER_TYPE_F, |
BRW_VERTICAL_STRIDE_8, |
BRW_WIDTH_8, |
BRW_HORIZONTAL_STRIDE_1, |
BRW_SWIZZLE_XYZW, |
WRITEMASK_XYZW); |
} |
/** Construct float[4] register */ |
static inline struct brw_reg brw_vec4_reg(unsigned file, |
unsigned nr, |
unsigned subnr) |
{ |
return brw_reg(file, |
nr, |
subnr, |
BRW_REGISTER_TYPE_F, |
BRW_VERTICAL_STRIDE_4, |
BRW_WIDTH_4, |
BRW_HORIZONTAL_STRIDE_1, |
BRW_SWIZZLE_XYZW, |
WRITEMASK_XYZW); |
} |
/** Construct float[2] register */ |
static inline struct brw_reg brw_vec2_reg(unsigned file, |
unsigned nr, |
unsigned subnr) |
{ |
return brw_reg(file, |
nr, |
subnr, |
BRW_REGISTER_TYPE_F, |
BRW_VERTICAL_STRIDE_2, |
BRW_WIDTH_2, |
BRW_HORIZONTAL_STRIDE_1, |
BRW_SWIZZLE_XYXY, |
WRITEMASK_XY); |
} |
/** Construct float[1] register */ |
static inline struct brw_reg brw_vec1_reg(unsigned file, |
unsigned nr, |
unsigned subnr) |
{ |
return brw_reg(file, |
nr, |
subnr, |
BRW_REGISTER_TYPE_F, |
BRW_VERTICAL_STRIDE_0, |
BRW_WIDTH_1, |
BRW_HORIZONTAL_STRIDE_0, |
BRW_SWIZZLE_XXXX, |
WRITEMASK_X); |
} |
static inline struct brw_reg __retype(struct brw_reg reg, |
unsigned type) |
{ |
reg.type = type; |
return reg; |
} |
static inline struct brw_reg __retype_d(struct brw_reg reg) |
{ |
return __retype(reg, BRW_REGISTER_TYPE_D); |
} |
static inline struct brw_reg __retype_ud(struct brw_reg reg) |
{ |
return __retype(reg, BRW_REGISTER_TYPE_UD); |
} |
static inline struct brw_reg __retype_uw(struct brw_reg reg) |
{ |
return __retype(reg, BRW_REGISTER_TYPE_UW); |
} |
static inline struct brw_reg __sechalf(struct brw_reg reg) |
{ |
if (reg.vstride) |
reg.nr++; |
return reg; |
} |
static inline struct brw_reg __suboffset(struct brw_reg reg, |
unsigned delta) |
{ |
reg.subnr += delta * type_sz(reg.type); |
return reg; |
} |
static inline struct brw_reg __offset(struct brw_reg reg, |
unsigned delta) |
{ |
reg.nr += delta; |
return reg; |
} |
static inline struct brw_reg byte_offset(struct brw_reg reg, |
unsigned bytes) |
{ |
unsigned newoffset = reg.nr * REG_SIZE + reg.subnr + bytes; |
reg.nr = newoffset / REG_SIZE; |
reg.subnr = newoffset % REG_SIZE; |
return reg; |
} |
/** Construct unsigned word[16] register */ |
static inline struct brw_reg brw_uw16_reg(unsigned file, |
unsigned nr, |
unsigned subnr) |
{ |
return __suboffset(__retype(brw_vec16_reg(file, nr, 0), BRW_REGISTER_TYPE_UW), subnr); |
} |
/** Construct unsigned word[8] register */ |
static inline struct brw_reg brw_uw8_reg(unsigned file, |
unsigned nr, |
unsigned subnr) |
{ |
return __suboffset(__retype(brw_vec8_reg(file, nr, 0), BRW_REGISTER_TYPE_UW), subnr); |
} |
/** Construct unsigned word[1] register */ |
static inline struct brw_reg brw_uw1_reg(unsigned file, |
unsigned nr, |
unsigned subnr) |
{ |
return __suboffset(__retype(brw_vec1_reg(file, nr, 0), BRW_REGISTER_TYPE_UW), subnr); |
} |
static inline struct brw_reg brw_imm_reg(unsigned type) |
{ |
return brw_reg( BRW_IMMEDIATE_VALUE, |
0, |
0, |
type, |
BRW_VERTICAL_STRIDE_0, |
BRW_WIDTH_1, |
BRW_HORIZONTAL_STRIDE_0, |
0, |
0); |
} |
/** Construct float immediate register */ |
static inline struct brw_reg brw_imm_f(float f) |
{ |
struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_F); |
imm.dw1.f = f; |
return imm; |
} |
/** Construct integer immediate register */ |
static inline struct brw_reg brw_imm_d(int d) |
{ |
struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_D); |
imm.dw1.d = d; |
return imm; |
} |
/** Construct uint immediate register */ |
static inline struct brw_reg brw_imm_ud(unsigned ud) |
{ |
struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_UD); |
imm.dw1.ud = ud; |
return imm; |
} |
/** Construct ushort immediate register */ |
static inline struct brw_reg brw_imm_uw(uint16_t uw) |
{ |
struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_UW); |
imm.dw1.ud = uw | (uw << 16); |
return imm; |
} |
/** Construct short immediate register */ |
static inline struct brw_reg brw_imm_w(int16_t w) |
{ |
struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_W); |
imm.dw1.d = w | (w << 16); |
return imm; |
} |
/* brw_imm_b and brw_imm_ub aren't supported by hardware - the type |
* numbers alias with _V and _VF below: |
*/ |
/** Construct vector of eight signed half-byte values */ |
static inline struct brw_reg brw_imm_v(unsigned v) |
{ |
struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_V); |
imm.vstride = BRW_VERTICAL_STRIDE_0; |
imm.width = BRW_WIDTH_8; |
imm.hstride = BRW_HORIZONTAL_STRIDE_1; |
imm.dw1.ud = v; |
return imm; |
} |
/** Construct vector of four 8-bit float values */ |
static inline struct brw_reg brw_imm_vf(unsigned v) |
{ |
struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_VF); |
imm.vstride = BRW_VERTICAL_STRIDE_0; |
imm.width = BRW_WIDTH_4; |
imm.hstride = BRW_HORIZONTAL_STRIDE_1; |
imm.dw1.ud = v; |
return imm; |
} |
#define VF_ZERO 0x0 |
#define VF_ONE 0x30 |
#define VF_NEG (1<<7) |
static inline struct brw_reg brw_imm_vf4(unsigned v0, |
unsigned v1, |
unsigned v2, |
unsigned v3) |
{ |
struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_VF); |
imm.vstride = BRW_VERTICAL_STRIDE_0; |
imm.width = BRW_WIDTH_4; |
imm.hstride = BRW_HORIZONTAL_STRIDE_1; |
imm.dw1.ud = ((v0 << 0) | |
(v1 << 8) | |
(v2 << 16) | |
(v3 << 24)); |
return imm; |
} |
static inline struct brw_reg brw_address(struct brw_reg reg) |
{ |
return brw_imm_uw(reg.nr * REG_SIZE + reg.subnr); |
} |
/** Construct float[1] general-purpose register */ |
static inline struct brw_reg brw_vec1_grf(unsigned nr, unsigned subnr) |
{ |
return brw_vec1_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr); |
} |
/** Construct float[2] general-purpose register */ |
static inline struct brw_reg brw_vec2_grf(unsigned nr, unsigned subnr) |
{ |
return brw_vec2_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr); |
} |
/** Construct float[4] general-purpose register */ |
static inline struct brw_reg brw_vec4_grf(unsigned nr, unsigned subnr) |
{ |
return brw_vec4_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr); |
} |
/** Construct float[8] general-purpose register */ |
static inline struct brw_reg brw_vec8_grf(unsigned nr, unsigned subnr) |
{ |
return brw_vec8_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr); |
} |
static inline struct brw_reg brw_uw8_grf(unsigned nr, unsigned subnr) |
{ |
return brw_uw8_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr); |
} |
static inline struct brw_reg brw_uw16_grf(unsigned nr, unsigned subnr) |
{ |
return brw_uw16_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr); |
} |
/** Construct null register (usually used for setting condition codes) */ |
static inline struct brw_reg brw_null_reg(void) |
{ |
return brw_vec8_reg(BRW_ARCHITECTURE_REGISTER_FILE, |
BRW_ARF_NULL, |
0); |
} |
static inline struct brw_reg brw_address_reg(unsigned subnr) |
{ |
return brw_uw1_reg(BRW_ARCHITECTURE_REGISTER_FILE, |
BRW_ARF_ADDRESS, |
subnr); |
} |
/* If/else instructions break in align16 mode if writemask & swizzle |
* aren't xyzw. This goes against the convention for other scalar |
* regs: |
*/ |
static inline struct brw_reg brw_ip_reg(void) |
{ |
return brw_reg(BRW_ARCHITECTURE_REGISTER_FILE, |
BRW_ARF_IP, |
0, |
BRW_REGISTER_TYPE_UD, |
BRW_VERTICAL_STRIDE_4, /* ? */ |
BRW_WIDTH_1, |
BRW_HORIZONTAL_STRIDE_0, |
BRW_SWIZZLE_XYZW, /* NOTE! */ |
WRITEMASK_XYZW); /* NOTE! */ |
} |
static inline struct brw_reg brw_acc_reg(void) |
{ |
return brw_vec8_reg(BRW_ARCHITECTURE_REGISTER_FILE, |
BRW_ARF_ACCUMULATOR, |
0); |
} |
static inline struct brw_reg brw_notification_1_reg(void) |
{ |
return brw_reg(BRW_ARCHITECTURE_REGISTER_FILE, |
BRW_ARF_NOTIFICATION_COUNT, |
1, |
BRW_REGISTER_TYPE_UD, |
BRW_VERTICAL_STRIDE_0, |
BRW_WIDTH_1, |
BRW_HORIZONTAL_STRIDE_0, |
BRW_SWIZZLE_XXXX, |
WRITEMASK_X); |
} |
static inline struct brw_reg brw_flag_reg(void) |
{ |
return brw_uw1_reg(BRW_ARCHITECTURE_REGISTER_FILE, |
BRW_ARF_FLAG, |
0); |
} |
static inline struct brw_reg brw_mask_reg(unsigned subnr) |
{ |
return brw_uw1_reg(BRW_ARCHITECTURE_REGISTER_FILE, |
BRW_ARF_MASK, |
subnr); |
} |
static inline struct brw_reg brw_message_reg(unsigned nr) |
{ |
assert((nr & ~(1 << 7)) < BRW_MAX_MRF); |
return brw_vec8_reg(BRW_MESSAGE_REGISTER_FILE, nr, 0); |
} |
static inline struct brw_reg brw_message4_reg(unsigned nr, |
int subnr) |
{ |
assert((nr & ~(1 << 7)) < BRW_MAX_MRF); |
return brw_vec4_reg(BRW_MESSAGE_REGISTER_FILE, nr, subnr); |
} |
/* This is almost always called with a numeric constant argument, so |
* make things easy to evaluate at compile time: |
*/ |
static inline unsigned cvt(unsigned val) |
{ |
switch (val) { |
case 0: return 0; |
case 1: return 1; |
case 2: return 2; |
case 4: return 3; |
case 8: return 4; |
case 16: return 5; |
case 32: return 6; |
} |
return 0; |
} |
static inline struct brw_reg __stride(struct brw_reg reg, |
unsigned vstride, |
unsigned width, |
unsigned hstride) |
{ |
reg.vstride = cvt(vstride); |
reg.width = cvt(width) - 1; |
reg.hstride = cvt(hstride); |
return reg; |
} |
static inline struct brw_reg vec16(struct brw_reg reg) |
{ |
return __stride(reg, 16,16,1); |
} |
static inline struct brw_reg vec8(struct brw_reg reg) |
{ |
return __stride(reg, 8,8,1); |
} |
static inline struct brw_reg vec4(struct brw_reg reg) |
{ |
return __stride(reg, 4,4,1); |
} |
static inline struct brw_reg vec2(struct brw_reg reg) |
{ |
return __stride(reg, 2,2,1); |
} |
static inline struct brw_reg vec1(struct brw_reg reg) |
{ |
return __stride(reg, 0,1,0); |
} |
static inline struct brw_reg get_element(struct brw_reg reg, unsigned elt) |
{ |
return vec1(__suboffset(reg, elt)); |
} |
static inline struct brw_reg get_element_ud(struct brw_reg reg, unsigned elt) |
{ |
return vec1(__suboffset(__retype(reg, BRW_REGISTER_TYPE_UD), elt)); |
} |
static inline struct brw_reg brw_swizzle(struct brw_reg reg, |
unsigned x, |
unsigned y, |
unsigned z, |
unsigned w) |
{ |
assert(reg.file != BRW_IMMEDIATE_VALUE); |
reg.dw1.bits.swizzle = BRW_SWIZZLE4(BRW_GET_SWZ(reg.dw1.bits.swizzle, x), |
BRW_GET_SWZ(reg.dw1.bits.swizzle, y), |
BRW_GET_SWZ(reg.dw1.bits.swizzle, z), |
BRW_GET_SWZ(reg.dw1.bits.swizzle, w)); |
return reg; |
} |
static inline struct brw_reg brw_swizzle1(struct brw_reg reg, |
unsigned x) |
{ |
return brw_swizzle(reg, x, x, x, x); |
} |
static inline struct brw_reg brw_writemask(struct brw_reg reg, |
unsigned mask) |
{ |
assert(reg.file != BRW_IMMEDIATE_VALUE); |
reg.dw1.bits.writemask &= mask; |
return reg; |
} |
static inline struct brw_reg brw_set_writemask(struct brw_reg reg, |
unsigned mask) |
{ |
assert(reg.file != BRW_IMMEDIATE_VALUE); |
reg.dw1.bits.writemask = mask; |
return reg; |
} |
static inline struct brw_reg brw_negate(struct brw_reg reg) |
{ |
reg.negate ^= 1; |
return reg; |
} |
static inline struct brw_reg brw_abs(struct brw_reg reg) |
{ |
reg.abs = 1; |
return reg; |
} |
/*********************************************************************** |
*/ |
static inline struct brw_reg brw_vec4_indirect(unsigned subnr, |
int offset) |
{ |
struct brw_reg reg = brw_vec4_grf(0, 0); |
reg.subnr = subnr; |
reg.address_mode = BRW_ADDRESS_REGISTER_INDIRECT_REGISTER; |
reg.dw1.bits.indirect_offset = offset; |
return reg; |
} |
static inline struct brw_reg brw_vec1_indirect(unsigned subnr, |
int offset) |
{ |
struct brw_reg reg = brw_vec1_grf(0, 0); |
reg.subnr = subnr; |
reg.address_mode = BRW_ADDRESS_REGISTER_INDIRECT_REGISTER; |
reg.dw1.bits.indirect_offset = offset; |
return reg; |
} |
static inline struct brw_reg deref_4f(struct brw_indirect ptr, int offset) |
{ |
return brw_vec4_indirect(ptr.addr_subnr, ptr.addr_offset + offset); |
} |
static inline struct brw_reg deref_1f(struct brw_indirect ptr, int offset) |
{ |
return brw_vec1_indirect(ptr.addr_subnr, ptr.addr_offset + offset); |
} |
static inline struct brw_reg deref_4b(struct brw_indirect ptr, int offset) |
{ |
return __retype(deref_4f(ptr, offset), BRW_REGISTER_TYPE_B); |
} |
static inline struct brw_reg deref_1uw(struct brw_indirect ptr, int offset) |
{ |
return __retype(deref_1f(ptr, offset), BRW_REGISTER_TYPE_UW); |
} |
static inline struct brw_reg deref_1d(struct brw_indirect ptr, int offset) |
{ |
return __retype(deref_1f(ptr, offset), BRW_REGISTER_TYPE_D); |
} |
static inline struct brw_reg deref_1ud(struct brw_indirect ptr, int offset) |
{ |
return __retype(deref_1f(ptr, offset), BRW_REGISTER_TYPE_UD); |
} |
static inline struct brw_reg get_addr_reg(struct brw_indirect ptr) |
{ |
return brw_address_reg(ptr.addr_subnr); |
} |
static inline struct brw_indirect brw_indirect_offset(struct brw_indirect ptr, int offset) |
{ |
ptr.addr_offset += offset; |
return ptr; |
} |
static inline struct brw_indirect brw_indirect(unsigned addr_subnr, int offset) |
{ |
struct brw_indirect ptr; |
ptr.addr_subnr = addr_subnr; |
ptr.addr_offset = offset; |
ptr.pad = 0; |
return ptr; |
} |
/** Do two brw_regs refer to the same register? */ |
static inline bool brw_same_reg(struct brw_reg r1, struct brw_reg r2) |
{ |
return r1.file == r2.file && r1.nr == r2.nr; |
} |
static inline struct brw_instruction *current_insn( struct brw_compile *p) |
{ |
return &p->store[p->nr_insn]; |
} |
static inline void brw_set_predicate_control( struct brw_compile *p, unsigned pc ) |
{ |
p->current->header.predicate_control = pc; |
} |
static inline void brw_set_predicate_inverse(struct brw_compile *p, bool predicate_inverse) |
{ |
p->current->header.predicate_inverse = predicate_inverse; |
} |
static inline void brw_set_conditionalmod( struct brw_compile *p, unsigned conditional ) |
{ |
p->current->header.destreg__conditionalmod = conditional; |
} |
static inline void brw_set_access_mode(struct brw_compile *p, unsigned access_mode) |
{ |
p->current->header.access_mode = access_mode; |
} |
static inline void brw_set_mask_control(struct brw_compile *p, unsigned value) |
{ |
p->current->header.mask_control = value; |
} |
static inline void brw_set_saturate(struct brw_compile *p, unsigned value) |
{ |
p->current->header.saturate = value; |
} |
static inline void brw_set_acc_write_control(struct brw_compile *p, unsigned value) |
{ |
if (p->gen >= 060) |
p->current->header.acc_wr_control = value; |
} |
void brw_pop_insn_state(struct brw_compile *p); |
void brw_push_insn_state(struct brw_compile *p); |
void brw_set_compression_control(struct brw_compile *p, enum brw_compression control); |
void brw_set_predicate_control_flag_value( struct brw_compile *p, unsigned value ); |
void brw_compile_init(struct brw_compile *p, int gen, void *store); |
void brw_set_dest(struct brw_compile *p, struct brw_instruction *insn, |
struct brw_reg dest); |
void brw_set_src0(struct brw_compile *p, struct brw_instruction *insn, |
struct brw_reg reg); |
void brw_set_src1(struct brw_compile *p, |
struct brw_instruction *insn, |
struct brw_reg reg); |
void gen6_resolve_implied_move(struct brw_compile *p, |
struct brw_reg *src, |
unsigned msg_reg_nr); |
static inline struct brw_instruction * |
brw_next_insn(struct brw_compile *p, unsigned opcode) |
{ |
struct brw_instruction *insn; |
assert(p->nr_insn + 1 < BRW_EU_MAX_INSN); |
insn = &p->store[p->nr_insn++]; |
*insn = *p->current; |
if (p->current->header.destreg__conditionalmod) { |
p->current->header.destreg__conditionalmod = 0; |
p->current->header.predicate_control = BRW_PREDICATE_NORMAL; |
} |
insn->header.opcode = opcode; |
return insn; |
} |
/* Helpers for regular instructions: */ |
#define ALU1(OP) \ |
static inline struct brw_instruction *brw_##OP(struct brw_compile *p, \ |
struct brw_reg dest, \ |
struct brw_reg src0) \ |
{ \ |
return brw_alu1(p, BRW_OPCODE_##OP, dest, src0); \ |
} |
#define ALU2(OP) \ |
static inline struct brw_instruction *brw_##OP(struct brw_compile *p, \ |
struct brw_reg dest, \ |
struct brw_reg src0, \ |
struct brw_reg src1) \ |
{ \ |
return brw_alu2(p, BRW_OPCODE_##OP, dest, src0, src1); \ |
} |
/* Rounding operations (other than RNDD) require two instructions - the first |
* stores a rounded value (possibly the wrong way) in the dest register, but |
* also sets a per-channel "increment bit" in the flag register. A predicated |
* add of 1.0 fixes dest to contain the desired result. |
* |
* Sandybridge and later appear to round correctly without an ADD. |
*/ |
#define ROUND(OP) \ |
static inline void brw_##OP(struct brw_compile *p, \ |
struct brw_reg dest, \ |
struct brw_reg src) \ |
{ \ |
struct brw_instruction *rnd, *add; \ |
rnd = brw_next_insn(p, BRW_OPCODE_##OP); \ |
brw_set_dest(p, rnd, dest); \ |
brw_set_src0(p, rnd, src); \ |
if (p->gen < 060) { \ |
/* turn on round-increments */ \ |
rnd->header.destreg__conditionalmod = BRW_CONDITIONAL_R; \ |
add = brw_ADD(p, dest, dest, brw_imm_f(1.0f)); \ |
add->header.predicate_control = BRW_PREDICATE_NORMAL; \ |
} \ |
} |
static inline struct brw_instruction *brw_alu1(struct brw_compile *p, |
unsigned opcode, |
struct brw_reg dest, |
struct brw_reg src) |
{ |
struct brw_instruction *insn = brw_next_insn(p, opcode); |
brw_set_dest(p, insn, dest); |
brw_set_src0(p, insn, src); |
return insn; |
} |
static inline struct brw_instruction *brw_alu2(struct brw_compile *p, |
unsigned opcode, |
struct brw_reg dest, |
struct brw_reg src0, |
struct brw_reg src1 ) |
{ |
struct brw_instruction *insn = brw_next_insn(p, opcode); |
brw_set_dest(p, insn, dest); |
brw_set_src0(p, insn, src0); |
brw_set_src1(p, insn, src1); |
return insn; |
} |
static inline struct brw_instruction *brw_ADD(struct brw_compile *p, |
struct brw_reg dest, |
struct brw_reg src0, |
struct brw_reg src1) |
{ |
/* 6.2.2: add */ |
if (src0.type == BRW_REGISTER_TYPE_F || |
(src0.file == BRW_IMMEDIATE_VALUE && |
src0.type == BRW_REGISTER_TYPE_VF)) { |
assert(src1.type != BRW_REGISTER_TYPE_UD); |
assert(src1.type != BRW_REGISTER_TYPE_D); |
} |
if (src1.type == BRW_REGISTER_TYPE_F || |
(src1.file == BRW_IMMEDIATE_VALUE && |
src1.type == BRW_REGISTER_TYPE_VF)) { |
assert(src0.type != BRW_REGISTER_TYPE_UD); |
assert(src0.type != BRW_REGISTER_TYPE_D); |
} |
return brw_alu2(p, BRW_OPCODE_ADD, dest, src0, src1); |
} |
static inline struct brw_instruction *brw_MUL(struct brw_compile *p, |
struct brw_reg dest, |
struct brw_reg src0, |
struct brw_reg src1) |
{ |
/* 6.32.38: mul */ |
if (src0.type == BRW_REGISTER_TYPE_D || |
src0.type == BRW_REGISTER_TYPE_UD || |
src1.type == BRW_REGISTER_TYPE_D || |
src1.type == BRW_REGISTER_TYPE_UD) { |
assert(dest.type != BRW_REGISTER_TYPE_F); |
} |
if (src0.type == BRW_REGISTER_TYPE_F || |
(src0.file == BRW_IMMEDIATE_VALUE && |
src0.type == BRW_REGISTER_TYPE_VF)) { |
assert(src1.type != BRW_REGISTER_TYPE_UD); |
assert(src1.type != BRW_REGISTER_TYPE_D); |
} |
if (src1.type == BRW_REGISTER_TYPE_F || |
(src1.file == BRW_IMMEDIATE_VALUE && |
src1.type == BRW_REGISTER_TYPE_VF)) { |
assert(src0.type != BRW_REGISTER_TYPE_UD); |
assert(src0.type != BRW_REGISTER_TYPE_D); |
} |
assert(src0.file != BRW_ARCHITECTURE_REGISTER_FILE || |
src0.nr != BRW_ARF_ACCUMULATOR); |
assert(src1.file != BRW_ARCHITECTURE_REGISTER_FILE || |
src1.nr != BRW_ARF_ACCUMULATOR); |
return brw_alu2(p, BRW_OPCODE_MUL, dest, src0, src1); |
} |
static inline struct brw_instruction *brw_JMPI(struct brw_compile *p, |
struct brw_reg dest, |
struct brw_reg src0, |
struct brw_reg src1) |
{ |
struct brw_instruction *insn = brw_alu2(p, BRW_OPCODE_JMPI, dest, src0, src1); |
insn->header.execution_size = 1; |
insn->header.compression_control = BRW_COMPRESSION_NONE; |
insn->header.mask_control = BRW_MASK_DISABLE; |
p->current->header.predicate_control = BRW_PREDICATE_NONE; |
return insn; |
} |
ALU1(MOV); |
ALU2(SEL); |
ALU1(NOT); |
ALU2(AND); |
ALU2(OR); |
ALU2(XOR); |
ALU2(SHR); |
ALU2(SHL); |
ALU2(RSR); |
ALU2(RSL); |
ALU2(ASR); |
ALU1(FRC); |
ALU1(RNDD); |
ALU2(MAC); |
ALU2(MACH); |
ALU1(LZD); |
ALU2(DP4); |
ALU2(DPH); |
ALU2(DP3); |
ALU2(DP2); |
ALU2(LINE); |
ALU2(PLN); |
ROUND(RNDZ); |
ROUND(RNDE); |
#undef ALU1 |
#undef ALU2 |
#undef ROUND |
/* Helpers for SEND instruction */ |
void brw_set_dp_read_message(struct brw_compile *p, |
struct brw_instruction *insn, |
unsigned binding_table_index, |
unsigned msg_control, |
unsigned msg_type, |
unsigned target_cache, |
unsigned msg_length, |
unsigned response_length); |
void brw_set_dp_write_message(struct brw_compile *p, |
struct brw_instruction *insn, |
unsigned binding_table_index, |
unsigned msg_control, |
unsigned msg_type, |
unsigned msg_length, |
bool header_present, |
bool last_render_target, |
unsigned response_length, |
bool end_of_thread, |
bool send_commit_msg); |
void brw_urb_WRITE(struct brw_compile *p, |
struct brw_reg dest, |
unsigned msg_reg_nr, |
struct brw_reg src0, |
bool allocate, |
bool used, |
unsigned msg_length, |
unsigned response_length, |
bool eot, |
bool writes_complete, |
unsigned offset, |
unsigned swizzle); |
void brw_ff_sync(struct brw_compile *p, |
struct brw_reg dest, |
unsigned msg_reg_nr, |
struct brw_reg src0, |
bool allocate, |
unsigned response_length, |
bool eot); |
void brw_fb_WRITE(struct brw_compile *p, |
int dispatch_width, |
unsigned msg_reg_nr, |
struct brw_reg src0, |
unsigned msg_control, |
unsigned binding_table_index, |
unsigned msg_length, |
unsigned response_length, |
bool eot, |
bool header_present); |
void brw_SAMPLE(struct brw_compile *p, |
struct brw_reg dest, |
unsigned msg_reg_nr, |
struct brw_reg src0, |
unsigned binding_table_index, |
unsigned sampler, |
unsigned writemask, |
unsigned msg_type, |
unsigned response_length, |
unsigned msg_length, |
bool header_present, |
unsigned simd_mode); |
void brw_math_16(struct brw_compile *p, |
struct brw_reg dest, |
unsigned function, |
unsigned saturate, |
unsigned msg_reg_nr, |
struct brw_reg src, |
unsigned precision); |
void brw_math(struct brw_compile *p, |
struct brw_reg dest, |
unsigned function, |
unsigned saturate, |
unsigned msg_reg_nr, |
struct brw_reg src, |
unsigned data_type, |
unsigned precision); |
void brw_math2(struct brw_compile *p, |
struct brw_reg dest, |
unsigned function, |
struct brw_reg src0, |
struct brw_reg src1); |
void brw_oword_block_read(struct brw_compile *p, |
struct brw_reg dest, |
struct brw_reg mrf, |
uint32_t offset, |
uint32_t bind_table_index); |
void brw_oword_block_read_scratch(struct brw_compile *p, |
struct brw_reg dest, |
struct brw_reg mrf, |
int num_regs, |
unsigned offset); |
void brw_oword_block_write_scratch(struct brw_compile *p, |
struct brw_reg mrf, |
int num_regs, |
unsigned offset); |
void brw_dword_scattered_read(struct brw_compile *p, |
struct brw_reg dest, |
struct brw_reg mrf, |
uint32_t bind_table_index); |
void brw_dp_READ_4_vs(struct brw_compile *p, |
struct brw_reg dest, |
unsigned location, |
unsigned bind_table_index); |
void brw_dp_READ_4_vs_relative(struct brw_compile *p, |
struct brw_reg dest, |
struct brw_reg addrReg, |
unsigned offset, |
unsigned bind_table_index); |
/* If/else/endif. Works by manipulating the execution flags on each |
* channel. |
*/ |
struct brw_instruction *brw_IF(struct brw_compile *p, |
unsigned execute_size); |
struct brw_instruction *gen6_IF(struct brw_compile *p, uint32_t conditional, |
struct brw_reg src0, struct brw_reg src1); |
void brw_ELSE(struct brw_compile *p); |
void brw_ENDIF(struct brw_compile *p); |
/* DO/WHILE loops: |
*/ |
struct brw_instruction *brw_DO(struct brw_compile *p, |
unsigned execute_size); |
struct brw_instruction *brw_WHILE(struct brw_compile *p, |
struct brw_instruction *patch_insn); |
struct brw_instruction *brw_BREAK(struct brw_compile *p, int pop_count); |
struct brw_instruction *brw_CONT(struct brw_compile *p, int pop_count); |
struct brw_instruction *gen6_CONT(struct brw_compile *p, |
struct brw_instruction *do_insn); |
/* Forward jumps: |
*/ |
void brw_land_fwd_jump(struct brw_compile *p, |
struct brw_instruction *jmp_insn); |
void brw_NOP(struct brw_compile *p); |
void brw_WAIT(struct brw_compile *p); |
/* Special case: there is never a destination, execution size will be |
* taken from src0: |
*/ |
void brw_CMP(struct brw_compile *p, |
struct brw_reg dest, |
unsigned conditional, |
struct brw_reg src0, |
struct brw_reg src1); |
static inline void brw_math_invert(struct brw_compile *p, |
struct brw_reg dst, |
struct brw_reg src) |
{ |
brw_math(p, |
dst, |
BRW_MATH_FUNCTION_INV, |
BRW_MATH_SATURATE_NONE, |
0, |
src, |
BRW_MATH_PRECISION_FULL, |
BRW_MATH_DATA_VECTOR); |
} |
void brw_set_uip_jip(struct brw_compile *p); |
uint32_t brw_swap_cmod(uint32_t cmod); |
void brw_disasm(FILE *file, |
const struct brw_instruction *inst, |
int gen); |
#endif |
/drivers/video/Intel-2D/sna/brw/brw_eu_emit.c |
---|
0,0 → 1,2002 |
/* |
Copyright (C) Intel Corp. 2006. All Rights Reserved. |
Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to |
develop this 3D driver. |
Permission is hereby granted, free of charge, to any person obtaining |
a copy of this software and associated documentation files (the |
"Software"), to deal in the Software without restriction, including |
without limitation the rights to use, copy, modify, merge, publish, |
distribute, sublicense, and/or sell copies of the Software, and to |
permit persons to whom the Software is furnished to do so, subject to |
the following conditions: |
The above copyright notice and this permission notice (including the |
next paragraph) shall be included in all copies or substantial |
portions of the Software. |
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, |
EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF |
MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. |
IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE |
LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION |
OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION |
WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. |
**********************************************************************/ |
/* |
* Authors: |
* Keith Whitwell <keith@tungstengraphics.com> |
*/ |
#include "brw_eu.h" |
#include <string.h> |
#include <stdlib.h> |
/*********************************************************************** |
* Internal helper for constructing instructions |
*/ |
static void guess_execution_size(struct brw_compile *p, |
struct brw_instruction *insn, |
struct brw_reg reg) |
{ |
if (reg.width == BRW_WIDTH_8 && p->compressed) |
insn->header.execution_size = BRW_EXECUTE_16; |
else |
insn->header.execution_size = reg.width; |
} |
/** |
* Prior to Sandybridge, the SEND instruction accepted non-MRF source |
* registers, implicitly moving the operand to a message register. |
* |
* On Sandybridge, this is no longer the case. This function performs the |
* explicit move; it should be called before emitting a SEND instruction. |
*/ |
void |
gen6_resolve_implied_move(struct brw_compile *p, |
struct brw_reg *src, |
unsigned msg_reg_nr) |
{ |
if (p->gen < 060) |
return; |
if (src->file == BRW_MESSAGE_REGISTER_FILE) |
return; |
if (src->file != BRW_ARCHITECTURE_REGISTER_FILE || src->nr != BRW_ARF_NULL) { |
brw_push_insn_state(p); |
brw_set_mask_control(p, BRW_MASK_DISABLE); |
brw_set_compression_control(p, BRW_COMPRESSION_NONE); |
brw_MOV(p, __retype_ud(brw_message_reg(msg_reg_nr)), __retype_ud(*src)); |
brw_pop_insn_state(p); |
} |
*src = brw_message_reg(msg_reg_nr); |
} |
static void |
gen7_convert_mrf_to_grf(struct brw_compile *p, struct brw_reg *reg) |
{ |
/* From the BSpec / ISA Reference / send - [DevIVB+]: |
* "The send with EOT should use register space R112-R127 for <src>. This is |
* to enable loading of a new thread into the same slot while the message |
* with EOT for current thread is pending dispatch." |
* |
* Since we're pretending to have 16 MRFs anyway, we may as well use the |
* registers required for messages with EOT. |
*/ |
if (p->gen >= 070 && reg->file == BRW_MESSAGE_REGISTER_FILE) { |
reg->file = BRW_GENERAL_REGISTER_FILE; |
reg->nr += 111; |
} |
} |
void |
brw_set_dest(struct brw_compile *p, struct brw_instruction *insn, |
struct brw_reg dest) |
{ |
if (dest.file != BRW_ARCHITECTURE_REGISTER_FILE && |
dest.file != BRW_MESSAGE_REGISTER_FILE) |
assert(dest.nr < 128); |
gen7_convert_mrf_to_grf(p, &dest); |
insn->bits1.da1.dest_reg_file = dest.file; |
insn->bits1.da1.dest_reg_type = dest.type; |
insn->bits1.da1.dest_address_mode = dest.address_mode; |
if (dest.address_mode == BRW_ADDRESS_DIRECT) { |
insn->bits1.da1.dest_reg_nr = dest.nr; |
if (insn->header.access_mode == BRW_ALIGN_1) { |
insn->bits1.da1.dest_subreg_nr = dest.subnr; |
if (dest.hstride == BRW_HORIZONTAL_STRIDE_0) |
dest.hstride = BRW_HORIZONTAL_STRIDE_1; |
insn->bits1.da1.dest_horiz_stride = dest.hstride; |
} else { |
insn->bits1.da16.dest_subreg_nr = dest.subnr / 16; |
insn->bits1.da16.dest_writemask = dest.dw1.bits.writemask; |
/* even ignored in da16, still need to set as '01' */ |
insn->bits1.da16.dest_horiz_stride = 1; |
} |
} else { |
insn->bits1.ia1.dest_subreg_nr = dest.subnr; |
/* These are different sizes in align1 vs align16: |
*/ |
if (insn->header.access_mode == BRW_ALIGN_1) { |
insn->bits1.ia1.dest_indirect_offset = dest.dw1.bits.indirect_offset; |
if (dest.hstride == BRW_HORIZONTAL_STRIDE_0) |
dest.hstride = BRW_HORIZONTAL_STRIDE_1; |
insn->bits1.ia1.dest_horiz_stride = dest.hstride; |
} |
else { |
insn->bits1.ia16.dest_indirect_offset = dest.dw1.bits.indirect_offset; |
/* even ignored in da16, still need to set as '01' */ |
insn->bits1.ia16.dest_horiz_stride = 1; |
} |
} |
guess_execution_size(p, insn, dest); |
} |
static const int reg_type_size[8] = { |
[0] = 4, |
[1] = 4, |
[2] = 2, |
[3] = 2, |
[4] = 1, |
[5] = 1, |
[7] = 4 |
}; |
static void |
validate_reg(struct brw_instruction *insn, struct brw_reg reg) |
{ |
int hstride_for_reg[] = {0, 1, 2, 4}; |
int vstride_for_reg[] = {0, 1, 2, 4, 8, 16, 32, 64, 128, 256}; |
int width_for_reg[] = {1, 2, 4, 8, 16}; |
int execsize_for_reg[] = {1, 2, 4, 8, 16}; |
int width, hstride, vstride, execsize; |
if (reg.file == BRW_IMMEDIATE_VALUE) { |
/* 3.3.6: Region Parameters. Restriction: Immediate vectors |
* mean the destination has to be 128-bit aligned and the |
* destination horiz stride has to be a word. |
*/ |
if (reg.type == BRW_REGISTER_TYPE_V) { |
assert(hstride_for_reg[insn->bits1.da1.dest_horiz_stride] * |
reg_type_size[insn->bits1.da1.dest_reg_type] == 2); |
} |
return; |
} |
if (reg.file == BRW_ARCHITECTURE_REGISTER_FILE && |
reg.file == BRW_ARF_NULL) |
return; |
hstride = hstride_for_reg[reg.hstride]; |
if (reg.vstride == 0xf) { |
vstride = -1; |
} else { |
vstride = vstride_for_reg[reg.vstride]; |
} |
width = width_for_reg[reg.width]; |
execsize = execsize_for_reg[insn->header.execution_size]; |
/* Restrictions from 3.3.10: Register Region Restrictions. */ |
/* 3. */ |
assert(execsize >= width); |
/* 4. */ |
if (execsize == width && hstride != 0) { |
assert(vstride == -1 || vstride == width * hstride); |
} |
/* 5. */ |
if (execsize == width && hstride == 0) { |
/* no restriction on vstride. */ |
} |
/* 6. */ |
if (width == 1) { |
assert(hstride == 0); |
} |
/* 7. */ |
if (execsize == 1 && width == 1) { |
assert(hstride == 0); |
assert(vstride == 0); |
} |
/* 8. */ |
if (vstride == 0 && hstride == 0) { |
assert(width == 1); |
} |
/* 10. Check destination issues. */ |
} |
void |
brw_set_src0(struct brw_compile *p, struct brw_instruction *insn, |
struct brw_reg reg) |
{ |
if (reg.type != BRW_ARCHITECTURE_REGISTER_FILE) |
assert(reg.nr < 128); |
gen7_convert_mrf_to_grf(p, ®); |
validate_reg(insn, reg); |
insn->bits1.da1.src0_reg_file = reg.file; |
insn->bits1.da1.src0_reg_type = reg.type; |
insn->bits2.da1.src0_abs = reg.abs; |
insn->bits2.da1.src0_negate = reg.negate; |
insn->bits2.da1.src0_address_mode = reg.address_mode; |
if (reg.file == BRW_IMMEDIATE_VALUE) { |
insn->bits3.ud = reg.dw1.ud; |
/* Required to set some fields in src1 as well: |
*/ |
insn->bits1.da1.src1_reg_file = 0; /* arf */ |
insn->bits1.da1.src1_reg_type = reg.type; |
} else { |
if (reg.address_mode == BRW_ADDRESS_DIRECT) { |
if (insn->header.access_mode == BRW_ALIGN_1) { |
insn->bits2.da1.src0_subreg_nr = reg.subnr; |
insn->bits2.da1.src0_reg_nr = reg.nr; |
} else { |
insn->bits2.da16.src0_subreg_nr = reg.subnr / 16; |
insn->bits2.da16.src0_reg_nr = reg.nr; |
} |
} else { |
insn->bits2.ia1.src0_subreg_nr = reg.subnr; |
if (insn->header.access_mode == BRW_ALIGN_1) { |
insn->bits2.ia1.src0_indirect_offset = reg.dw1.bits.indirect_offset; |
} else { |
insn->bits2.ia16.src0_subreg_nr = reg.dw1.bits.indirect_offset; |
} |
} |
if (insn->header.access_mode == BRW_ALIGN_1) { |
if (reg.width == BRW_WIDTH_1 && |
insn->header.execution_size == BRW_EXECUTE_1) { |
insn->bits2.da1.src0_horiz_stride = BRW_HORIZONTAL_STRIDE_0; |
insn->bits2.da1.src0_width = BRW_WIDTH_1; |
insn->bits2.da1.src0_vert_stride = BRW_VERTICAL_STRIDE_0; |
} else { |
insn->bits2.da1.src0_horiz_stride = reg.hstride; |
insn->bits2.da1.src0_width = reg.width; |
insn->bits2.da1.src0_vert_stride = reg.vstride; |
} |
} else { |
insn->bits2.da16.src0_swz_x = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_X); |
insn->bits2.da16.src0_swz_y = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_Y); |
insn->bits2.da16.src0_swz_z = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_Z); |
insn->bits2.da16.src0_swz_w = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_W); |
/* This is an oddity of the fact we're using the same |
* descriptions for registers in align_16 as align_1: |
*/ |
if (reg.vstride == BRW_VERTICAL_STRIDE_8) |
insn->bits2.da16.src0_vert_stride = BRW_VERTICAL_STRIDE_4; |
else |
insn->bits2.da16.src0_vert_stride = reg.vstride; |
} |
} |
} |
void brw_set_src1(struct brw_compile *p, |
struct brw_instruction *insn, |
struct brw_reg reg) |
{ |
assert(reg.file != BRW_MESSAGE_REGISTER_FILE); |
assert(reg.nr < 128); |
gen7_convert_mrf_to_grf(p, ®); |
validate_reg(insn, reg); |
insn->bits1.da1.src1_reg_file = reg.file; |
insn->bits1.da1.src1_reg_type = reg.type; |
insn->bits3.da1.src1_abs = reg.abs; |
insn->bits3.da1.src1_negate = reg.negate; |
/* Only src1 can be immediate in two-argument instructions. */ |
assert(insn->bits1.da1.src0_reg_file != BRW_IMMEDIATE_VALUE); |
if (reg.file == BRW_IMMEDIATE_VALUE) { |
insn->bits3.ud = reg.dw1.ud; |
} else { |
/* This is a hardware restriction, which may or may not be lifted |
* in the future: |
*/ |
assert (reg.address_mode == BRW_ADDRESS_DIRECT); |
/* assert (reg.file == BRW_GENERAL_REGISTER_FILE); */ |
if (insn->header.access_mode == BRW_ALIGN_1) { |
insn->bits3.da1.src1_subreg_nr = reg.subnr; |
insn->bits3.da1.src1_reg_nr = reg.nr; |
} else { |
insn->bits3.da16.src1_subreg_nr = reg.subnr / 16; |
insn->bits3.da16.src1_reg_nr = reg.nr; |
} |
if (insn->header.access_mode == BRW_ALIGN_1) { |
if (reg.width == BRW_WIDTH_1 && |
insn->header.execution_size == BRW_EXECUTE_1) { |
insn->bits3.da1.src1_horiz_stride = BRW_HORIZONTAL_STRIDE_0; |
insn->bits3.da1.src1_width = BRW_WIDTH_1; |
insn->bits3.da1.src1_vert_stride = BRW_VERTICAL_STRIDE_0; |
} else { |
insn->bits3.da1.src1_horiz_stride = reg.hstride; |
insn->bits3.da1.src1_width = reg.width; |
insn->bits3.da1.src1_vert_stride = reg.vstride; |
} |
} else { |
insn->bits3.da16.src1_swz_x = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_X); |
insn->bits3.da16.src1_swz_y = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_Y); |
insn->bits3.da16.src1_swz_z = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_Z); |
insn->bits3.da16.src1_swz_w = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_W); |
/* This is an oddity of the fact we're using the same |
* descriptions for registers in align_16 as align_1: |
*/ |
if (reg.vstride == BRW_VERTICAL_STRIDE_8) |
insn->bits3.da16.src1_vert_stride = BRW_VERTICAL_STRIDE_4; |
else |
insn->bits3.da16.src1_vert_stride = reg.vstride; |
} |
} |
} |
/** |
* Set the Message Descriptor and Extended Message Descriptor fields |
* for SEND messages. |
* |
* \note This zeroes out the Function Control bits, so it must be called |
* \b before filling out any message-specific data. Callers can |
* choose not to fill in irrelevant bits; they will be zero. |
*/ |
static void |
brw_set_message_descriptor(struct brw_compile *p, |
struct brw_instruction *inst, |
enum brw_message_target sfid, |
unsigned msg_length, |
unsigned response_length, |
bool header_present, |
bool end_of_thread) |
{ |
brw_set_src1(p, inst, brw_imm_d(0)); |
if (p->gen >= 050) { |
inst->bits3.generic_gen5.header_present = header_present; |
inst->bits3.generic_gen5.response_length = response_length; |
inst->bits3.generic_gen5.msg_length = msg_length; |
inst->bits3.generic_gen5.end_of_thread = end_of_thread; |
if (p->gen >= 060) { |
/* On Gen6+ Message target/SFID goes in bits 27:24 of the header */ |
inst->header.destreg__conditionalmod = sfid; |
} else { |
/* Set Extended Message Descriptor (ex_desc) */ |
inst->bits2.send_gen5.sfid = sfid; |
inst->bits2.send_gen5.end_of_thread = end_of_thread; |
} |
} else { |
inst->bits3.generic.response_length = response_length; |
inst->bits3.generic.msg_length = msg_length; |
inst->bits3.generic.msg_target = sfid; |
inst->bits3.generic.end_of_thread = end_of_thread; |
} |
} |
static void brw_set_math_message(struct brw_compile *p, |
struct brw_instruction *insn, |
unsigned function, |
unsigned integer_type, |
bool low_precision, |
bool saturate, |
unsigned dataType) |
{ |
unsigned msg_length; |
unsigned response_length; |
/* Infer message length from the function */ |
switch (function) { |
case BRW_MATH_FUNCTION_POW: |
case BRW_MATH_FUNCTION_INT_DIV_QUOTIENT: |
case BRW_MATH_FUNCTION_INT_DIV_REMAINDER: |
case BRW_MATH_FUNCTION_INT_DIV_QUOTIENT_AND_REMAINDER: |
msg_length = 2; |
break; |
default: |
msg_length = 1; |
break; |
} |
/* Infer response length from the function */ |
switch (function) { |
case BRW_MATH_FUNCTION_SINCOS: |
case BRW_MATH_FUNCTION_INT_DIV_QUOTIENT_AND_REMAINDER: |
response_length = 2; |
break; |
default: |
response_length = 1; |
break; |
} |
brw_set_message_descriptor(p, insn, BRW_SFID_MATH, |
msg_length, response_length, |
false, false); |
if (p->gen == 050) { |
insn->bits3.math_gen5.function = function; |
insn->bits3.math_gen5.int_type = integer_type; |
insn->bits3.math_gen5.precision = low_precision; |
insn->bits3.math_gen5.saturate = saturate; |
insn->bits3.math_gen5.data_type = dataType; |
insn->bits3.math_gen5.snapshot = 0; |
} else { |
insn->bits3.math.function = function; |
insn->bits3.math.int_type = integer_type; |
insn->bits3.math.precision = low_precision; |
insn->bits3.math.saturate = saturate; |
insn->bits3.math.data_type = dataType; |
} |
} |
static void brw_set_ff_sync_message(struct brw_compile *p, |
struct brw_instruction *insn, |
bool allocate, |
unsigned response_length, |
bool end_of_thread) |
{ |
brw_set_message_descriptor(p, insn, BRW_SFID_URB, |
1, response_length, |
true, end_of_thread); |
insn->bits3.urb_gen5.opcode = 1; /* FF_SYNC */ |
insn->bits3.urb_gen5.offset = 0; /* Not used by FF_SYNC */ |
insn->bits3.urb_gen5.swizzle_control = 0; /* Not used by FF_SYNC */ |
insn->bits3.urb_gen5.allocate = allocate; |
insn->bits3.urb_gen5.used = 0; /* Not used by FF_SYNC */ |
insn->bits3.urb_gen5.complete = 0; /* Not used by FF_SYNC */ |
} |
static void brw_set_urb_message(struct brw_compile *p, |
struct brw_instruction *insn, |
bool allocate, |
bool used, |
unsigned msg_length, |
unsigned response_length, |
bool end_of_thread, |
bool complete, |
unsigned offset, |
unsigned swizzle_control) |
{ |
brw_set_message_descriptor(p, insn, BRW_SFID_URB, |
msg_length, response_length, true, end_of_thread); |
if (p->gen >= 070) { |
insn->bits3.urb_gen7.opcode = 0; /* URB_WRITE_HWORD */ |
insn->bits3.urb_gen7.offset = offset; |
assert(swizzle_control != BRW_URB_SWIZZLE_TRANSPOSE); |
insn->bits3.urb_gen7.swizzle_control = swizzle_control; |
/* per_slot_offset = 0 makes it ignore offsets in message header */ |
insn->bits3.urb_gen7.per_slot_offset = 0; |
insn->bits3.urb_gen7.complete = complete; |
} else if (p->gen >= 050) { |
insn->bits3.urb_gen5.opcode = 0; /* URB_WRITE */ |
insn->bits3.urb_gen5.offset = offset; |
insn->bits3.urb_gen5.swizzle_control = swizzle_control; |
insn->bits3.urb_gen5.allocate = allocate; |
insn->bits3.urb_gen5.used = used; /* ? */ |
insn->bits3.urb_gen5.complete = complete; |
} else { |
insn->bits3.urb.opcode = 0; /* ? */ |
insn->bits3.urb.offset = offset; |
insn->bits3.urb.swizzle_control = swizzle_control; |
insn->bits3.urb.allocate = allocate; |
insn->bits3.urb.used = used; /* ? */ |
insn->bits3.urb.complete = complete; |
} |
} |
void |
brw_set_dp_write_message(struct brw_compile *p, |
struct brw_instruction *insn, |
unsigned binding_table_index, |
unsigned msg_control, |
unsigned msg_type, |
unsigned msg_length, |
bool header_present, |
bool last_render_target, |
unsigned response_length, |
bool end_of_thread, |
bool send_commit_msg) |
{ |
unsigned sfid; |
if (p->gen >= 070) { |
/* Use the Render Cache for RT writes; otherwise use the Data Cache */ |
if (msg_type == GEN6_DATAPORT_WRITE_MESSAGE_RENDER_TARGET_WRITE) |
sfid = GEN6_SFID_DATAPORT_RENDER_CACHE; |
else |
sfid = GEN7_SFID_DATAPORT_DATA_CACHE; |
} else if (p->gen >= 060) { |
/* Use the render cache for all write messages. */ |
sfid = GEN6_SFID_DATAPORT_RENDER_CACHE; |
} else { |
sfid = BRW_SFID_DATAPORT_WRITE; |
} |
brw_set_message_descriptor(p, insn, sfid, |
msg_length, response_length, |
header_present, end_of_thread); |
if (p->gen >= 070) { |
insn->bits3.gen7_dp.binding_table_index = binding_table_index; |
insn->bits3.gen7_dp.msg_control = msg_control; |
insn->bits3.gen7_dp.last_render_target = last_render_target; |
insn->bits3.gen7_dp.msg_type = msg_type; |
} else if (p->gen >= 060) { |
insn->bits3.gen6_dp.binding_table_index = binding_table_index; |
insn->bits3.gen6_dp.msg_control = msg_control; |
insn->bits3.gen6_dp.last_render_target = last_render_target; |
insn->bits3.gen6_dp.msg_type = msg_type; |
insn->bits3.gen6_dp.send_commit_msg = send_commit_msg; |
} else if (p->gen >= 050) { |
insn->bits3.dp_write_gen5.binding_table_index = binding_table_index; |
insn->bits3.dp_write_gen5.msg_control = msg_control; |
insn->bits3.dp_write_gen5.last_render_target = last_render_target; |
insn->bits3.dp_write_gen5.msg_type = msg_type; |
insn->bits3.dp_write_gen5.send_commit_msg = send_commit_msg; |
} else { |
insn->bits3.dp_write.binding_table_index = binding_table_index; |
insn->bits3.dp_write.msg_control = msg_control; |
insn->bits3.dp_write.last_render_target = last_render_target; |
insn->bits3.dp_write.msg_type = msg_type; |
insn->bits3.dp_write.send_commit_msg = send_commit_msg; |
} |
} |
void |
brw_set_dp_read_message(struct brw_compile *p, |
struct brw_instruction *insn, |
unsigned binding_table_index, |
unsigned msg_control, |
unsigned msg_type, |
unsigned target_cache, |
unsigned msg_length, |
unsigned response_length) |
{ |
unsigned sfid; |
if (p->gen >= 070) { |
sfid = GEN7_SFID_DATAPORT_DATA_CACHE; |
} else if (p->gen >= 060) { |
if (target_cache == BRW_DATAPORT_READ_TARGET_RENDER_CACHE) |
sfid = GEN6_SFID_DATAPORT_RENDER_CACHE; |
else |
sfid = GEN6_SFID_DATAPORT_SAMPLER_CACHE; |
} else { |
sfid = BRW_SFID_DATAPORT_READ; |
} |
brw_set_message_descriptor(p, insn, sfid, |
msg_length, response_length, |
true, false); |
if (p->gen >= 070) { |
insn->bits3.gen7_dp.binding_table_index = binding_table_index; |
insn->bits3.gen7_dp.msg_control = msg_control; |
insn->bits3.gen7_dp.last_render_target = 0; |
insn->bits3.gen7_dp.msg_type = msg_type; |
} else if (p->gen >= 060) { |
insn->bits3.gen6_dp.binding_table_index = binding_table_index; |
insn->bits3.gen6_dp.msg_control = msg_control; |
insn->bits3.gen6_dp.last_render_target = 0; |
insn->bits3.gen6_dp.msg_type = msg_type; |
insn->bits3.gen6_dp.send_commit_msg = 0; |
} else if (p->gen >= 050) { |
insn->bits3.dp_read_gen5.binding_table_index = binding_table_index; |
insn->bits3.dp_read_gen5.msg_control = msg_control; |
insn->bits3.dp_read_gen5.msg_type = msg_type; |
insn->bits3.dp_read_gen5.target_cache = target_cache; |
} else if (p->gen >= 045) { |
insn->bits3.dp_read_g4x.binding_table_index = binding_table_index; /*0:7*/ |
insn->bits3.dp_read_g4x.msg_control = msg_control; /*8:10*/ |
insn->bits3.dp_read_g4x.msg_type = msg_type; /*11:13*/ |
insn->bits3.dp_read_g4x.target_cache = target_cache; /*14:15*/ |
} else { |
insn->bits3.dp_read.binding_table_index = binding_table_index; /*0:7*/ |
insn->bits3.dp_read.msg_control = msg_control; /*8:11*/ |
insn->bits3.dp_read.msg_type = msg_type; /*12:13*/ |
insn->bits3.dp_read.target_cache = target_cache; /*14:15*/ |
} |
} |
static void brw_set_sampler_message(struct brw_compile *p, |
struct brw_instruction *insn, |
unsigned binding_table_index, |
unsigned sampler, |
unsigned msg_type, |
unsigned response_length, |
unsigned msg_length, |
bool header_present, |
unsigned simd_mode) |
{ |
brw_set_message_descriptor(p, insn, BRW_SFID_SAMPLER, |
msg_length, response_length, |
header_present, false); |
if (p->gen >= 070) { |
insn->bits3.sampler_gen7.binding_table_index = binding_table_index; |
insn->bits3.sampler_gen7.sampler = sampler; |
insn->bits3.sampler_gen7.msg_type = msg_type; |
insn->bits3.sampler_gen7.simd_mode = simd_mode; |
} else if (p->gen >= 050) { |
insn->bits3.sampler_gen5.binding_table_index = binding_table_index; |
insn->bits3.sampler_gen5.sampler = sampler; |
insn->bits3.sampler_gen5.msg_type = msg_type; |
insn->bits3.sampler_gen5.simd_mode = simd_mode; |
} else if (p->gen >= 045) { |
insn->bits3.sampler_g4x.binding_table_index = binding_table_index; |
insn->bits3.sampler_g4x.sampler = sampler; |
insn->bits3.sampler_g4x.msg_type = msg_type; |
} else { |
insn->bits3.sampler.binding_table_index = binding_table_index; |
insn->bits3.sampler.sampler = sampler; |
insn->bits3.sampler.msg_type = msg_type; |
insn->bits3.sampler.return_format = BRW_SAMPLER_RETURN_FORMAT_FLOAT32; |
} |
} |
void brw_NOP(struct brw_compile *p) |
{ |
struct brw_instruction *insn = brw_next_insn(p, BRW_OPCODE_NOP); |
brw_set_dest(p, insn, __retype_ud(brw_vec4_grf(0,0))); |
brw_set_src0(p, insn, __retype_ud(brw_vec4_grf(0,0))); |
brw_set_src1(p, insn, brw_imm_ud(0x0)); |
} |
/*********************************************************************** |
* Comparisons, if/else/endif |
*/ |
static void |
push_if_stack(struct brw_compile *p, struct brw_instruction *inst) |
{ |
p->if_stack[p->if_stack_depth] = inst; |
p->if_stack_depth++; |
if (p->if_stack_array_size <= p->if_stack_depth) { |
p->if_stack_array_size *= 2; |
p->if_stack = realloc(p->if_stack, sizeof(struct brw_instruction *)*p->if_stack_array_size); |
} |
} |
/* EU takes the value from the flag register and pushes it onto some |
* sort of a stack (presumably merging with any flag value already on |
* the stack). Within an if block, the flags at the top of the stack |
* control execution on each channel of the unit, eg. on each of the |
* 16 pixel values in our wm programs. |
* |
* When the matching 'else' instruction is reached (presumably by |
* countdown of the instruction count patched in by our ELSE/ENDIF |
* functions), the relevent flags are inverted. |
* |
* When the matching 'endif' instruction is reached, the flags are |
* popped off. If the stack is now empty, normal execution resumes. |
*/ |
struct brw_instruction * |
brw_IF(struct brw_compile *p, unsigned execute_size) |
{ |
struct brw_instruction *insn; |
insn = brw_next_insn(p, BRW_OPCODE_IF); |
/* Override the defaults for this instruction: */ |
if (p->gen < 060) { |
brw_set_dest(p, insn, brw_ip_reg()); |
brw_set_src0(p, insn, brw_ip_reg()); |
brw_set_src1(p, insn, brw_imm_d(0x0)); |
} else if (p->gen < 070) { |
brw_set_dest(p, insn, brw_imm_w(0)); |
insn->bits1.branch_gen6.jump_count = 0; |
brw_set_src0(p, insn, __retype_d(brw_null_reg())); |
brw_set_src1(p, insn, __retype_d(brw_null_reg())); |
} else { |
brw_set_dest(p, insn, __retype_d(brw_null_reg())); |
brw_set_src0(p, insn, __retype_d(brw_null_reg())); |
brw_set_src1(p, insn, brw_imm_ud(0)); |
insn->bits3.break_cont.jip = 0; |
insn->bits3.break_cont.uip = 0; |
} |
insn->header.execution_size = execute_size; |
insn->header.compression_control = BRW_COMPRESSION_NONE; |
insn->header.predicate_control = BRW_PREDICATE_NORMAL; |
insn->header.mask_control = BRW_MASK_ENABLE; |
if (!p->single_program_flow) |
insn->header.thread_control = BRW_THREAD_SWITCH; |
p->current->header.predicate_control = BRW_PREDICATE_NONE; |
push_if_stack(p, insn); |
return insn; |
} |
/* This function is only used for gen6-style IF instructions with an |
* embedded comparison (conditional modifier). It is not used on gen7. |
*/ |
struct brw_instruction * |
gen6_IF(struct brw_compile *p, uint32_t conditional, |
struct brw_reg src0, struct brw_reg src1) |
{ |
struct brw_instruction *insn; |
insn = brw_next_insn(p, BRW_OPCODE_IF); |
brw_set_dest(p, insn, brw_imm_w(0)); |
if (p->compressed) { |
insn->header.execution_size = BRW_EXECUTE_16; |
} else { |
insn->header.execution_size = BRW_EXECUTE_8; |
} |
insn->bits1.branch_gen6.jump_count = 0; |
brw_set_src0(p, insn, src0); |
brw_set_src1(p, insn, src1); |
assert(insn->header.compression_control == BRW_COMPRESSION_NONE); |
assert(insn->header.predicate_control == BRW_PREDICATE_NONE); |
insn->header.destreg__conditionalmod = conditional; |
if (!p->single_program_flow) |
insn->header.thread_control = BRW_THREAD_SWITCH; |
push_if_stack(p, insn); |
return insn; |
} |
/** |
* In single-program-flow (SPF) mode, convert IF and ELSE into ADDs. |
*/ |
static void |
convert_IF_ELSE_to_ADD(struct brw_compile *p, |
struct brw_instruction *if_inst, |
struct brw_instruction *else_inst) |
{ |
/* The next instruction (where the ENDIF would be, if it existed) */ |
struct brw_instruction *next_inst = &p->store[p->nr_insn]; |
assert(p->single_program_flow); |
assert(if_inst != NULL && if_inst->header.opcode == BRW_OPCODE_IF); |
assert(else_inst == NULL || else_inst->header.opcode == BRW_OPCODE_ELSE); |
assert(if_inst->header.execution_size == BRW_EXECUTE_1); |
/* Convert IF to an ADD instruction that moves the instruction pointer |
* to the first instruction of the ELSE block. If there is no ELSE |
* block, point to where ENDIF would be. Reverse the predicate. |
* |
* There's no need to execute an ENDIF since we don't need to do any |
* stack operations, and if we're currently executing, we just want to |
* continue normally. |
*/ |
if_inst->header.opcode = BRW_OPCODE_ADD; |
if_inst->header.predicate_inverse = 1; |
if (else_inst != NULL) { |
/* Convert ELSE to an ADD instruction that points where the ENDIF |
* would be. |
*/ |
else_inst->header.opcode = BRW_OPCODE_ADD; |
if_inst->bits3.ud = (else_inst - if_inst + 1) * 16; |
else_inst->bits3.ud = (next_inst - else_inst) * 16; |
} else { |
if_inst->bits3.ud = (next_inst - if_inst) * 16; |
} |
} |
/** |
* Patch IF and ELSE instructions with appropriate jump targets. |
*/ |
static void |
patch_IF_ELSE(struct brw_compile *p, |
struct brw_instruction *if_inst, |
struct brw_instruction *else_inst, |
struct brw_instruction *endif_inst) |
{ |
unsigned br = 1; |
assert(!p->single_program_flow); |
assert(if_inst != NULL && if_inst->header.opcode == BRW_OPCODE_IF); |
assert(endif_inst != NULL); |
assert(else_inst == NULL || else_inst->header.opcode == BRW_OPCODE_ELSE); |
/* Jump count is for 64bit data chunk each, so one 128bit instruction |
* requires 2 chunks. |
*/ |
if (p->gen >= 050) |
br = 2; |
assert(endif_inst->header.opcode == BRW_OPCODE_ENDIF); |
endif_inst->header.execution_size = if_inst->header.execution_size; |
if (else_inst == NULL) { |
/* Patch IF -> ENDIF */ |
if (p->gen < 060) { |
/* Turn it into an IFF, which means no mask stack operations for |
* all-false and jumping past the ENDIF. |
*/ |
if_inst->header.opcode = BRW_OPCODE_IFF; |
if_inst->bits3.if_else.jump_count = br * (endif_inst - if_inst + 1); |
if_inst->bits3.if_else.pop_count = 0; |
if_inst->bits3.if_else.pad0 = 0; |
} else if (p->gen < 070) { |
/* As of gen6, there is no IFF and IF must point to the ENDIF. */ |
if_inst->bits1.branch_gen6.jump_count = br * (endif_inst - if_inst); |
} else { |
if_inst->bits3.break_cont.uip = br * (endif_inst - if_inst); |
if_inst->bits3.break_cont.jip = br * (endif_inst - if_inst); |
} |
} else { |
else_inst->header.execution_size = if_inst->header.execution_size; |
/* Patch IF -> ELSE */ |
if (p->gen < 060) { |
if_inst->bits3.if_else.jump_count = br * (else_inst - if_inst); |
if_inst->bits3.if_else.pop_count = 0; |
if_inst->bits3.if_else.pad0 = 0; |
} else if (p->gen <= 070) { |
if_inst->bits1.branch_gen6.jump_count = br * (else_inst - if_inst + 1); |
} |
/* Patch ELSE -> ENDIF */ |
if (p->gen < 060) { |
/* BRW_OPCODE_ELSE pre-gen6 should point just past the |
* matching ENDIF. |
*/ |
else_inst->bits3.if_else.jump_count = br*(endif_inst - else_inst + 1); |
else_inst->bits3.if_else.pop_count = 1; |
else_inst->bits3.if_else.pad0 = 0; |
} else if (p->gen < 070) { |
/* BRW_OPCODE_ELSE on gen6 should point to the matching ENDIF. */ |
else_inst->bits1.branch_gen6.jump_count = br*(endif_inst - else_inst); |
} else { |
/* The IF instruction's JIP should point just past the ELSE */ |
if_inst->bits3.break_cont.jip = br * (else_inst - if_inst + 1); |
/* The IF instruction's UIP and ELSE's JIP should point to ENDIF */ |
if_inst->bits3.break_cont.uip = br * (endif_inst - if_inst); |
else_inst->bits3.break_cont.jip = br * (endif_inst - else_inst); |
} |
} |
} |
void |
brw_ELSE(struct brw_compile *p) |
{ |
struct brw_instruction *insn; |
insn = brw_next_insn(p, BRW_OPCODE_ELSE); |
if (p->gen < 060) { |
brw_set_dest(p, insn, brw_ip_reg()); |
brw_set_src0(p, insn, brw_ip_reg()); |
brw_set_src1(p, insn, brw_imm_d(0x0)); |
} else if (p->gen < 070) { |
brw_set_dest(p, insn, brw_imm_w(0)); |
insn->bits1.branch_gen6.jump_count = 0; |
brw_set_src0(p, insn, __retype_d(brw_null_reg())); |
brw_set_src1(p, insn, __retype_d(brw_null_reg())); |
} else { |
brw_set_dest(p, insn, __retype_d(brw_null_reg())); |
brw_set_src0(p, insn, __retype_d(brw_null_reg())); |
brw_set_src1(p, insn, brw_imm_ud(0)); |
insn->bits3.break_cont.jip = 0; |
insn->bits3.break_cont.uip = 0; |
} |
insn->header.compression_control = BRW_COMPRESSION_NONE; |
insn->header.mask_control = BRW_MASK_ENABLE; |
if (!p->single_program_flow) |
insn->header.thread_control = BRW_THREAD_SWITCH; |
push_if_stack(p, insn); |
} |
void |
brw_ENDIF(struct brw_compile *p) |
{ |
struct brw_instruction *insn; |
struct brw_instruction *else_inst = NULL; |
struct brw_instruction *if_inst = NULL; |
/* Pop the IF and (optional) ELSE instructions from the stack */ |
p->if_stack_depth--; |
if (p->if_stack[p->if_stack_depth]->header.opcode == BRW_OPCODE_ELSE) { |
else_inst = p->if_stack[p->if_stack_depth]; |
p->if_stack_depth--; |
} |
if_inst = p->if_stack[p->if_stack_depth]; |
if (p->single_program_flow) { |
/* ENDIF is useless; don't bother emitting it. */ |
convert_IF_ELSE_to_ADD(p, if_inst, else_inst); |
return; |
} |
insn = brw_next_insn(p, BRW_OPCODE_ENDIF); |
if (p->gen < 060) { |
brw_set_dest(p, insn, __retype_ud(brw_vec4_grf(0,0))); |
brw_set_src0(p, insn, __retype_ud(brw_vec4_grf(0,0))); |
brw_set_src1(p, insn, brw_imm_d(0x0)); |
} else if (p->gen < 070) { |
brw_set_dest(p, insn, brw_imm_w(0)); |
brw_set_src0(p, insn, __retype_d(brw_null_reg())); |
brw_set_src1(p, insn, __retype_d(brw_null_reg())); |
} else { |
brw_set_dest(p, insn, __retype_d(brw_null_reg())); |
brw_set_src0(p, insn, __retype_d(brw_null_reg())); |
brw_set_src1(p, insn, brw_imm_ud(0)); |
} |
insn->header.compression_control = BRW_COMPRESSION_NONE; |
insn->header.mask_control = BRW_MASK_ENABLE; |
insn->header.thread_control = BRW_THREAD_SWITCH; |
/* Also pop item off the stack in the endif instruction: */ |
if (p->gen < 060) { |
insn->bits3.if_else.jump_count = 0; |
insn->bits3.if_else.pop_count = 1; |
insn->bits3.if_else.pad0 = 0; |
} else if (p->gen < 070) { |
insn->bits1.branch_gen6.jump_count = 2; |
} else { |
insn->bits3.break_cont.jip = 2; |
} |
patch_IF_ELSE(p, if_inst, else_inst, insn); |
} |
struct brw_instruction *brw_BREAK(struct brw_compile *p, int pop_count) |
{ |
struct brw_instruction *insn; |
insn = brw_next_insn(p, BRW_OPCODE_BREAK); |
if (p->gen >= 060) { |
brw_set_dest(p, insn, __retype_d(brw_null_reg())); |
brw_set_src0(p, insn, __retype_d(brw_null_reg())); |
brw_set_src1(p, insn, brw_imm_d(0x0)); |
} else { |
brw_set_dest(p, insn, brw_ip_reg()); |
brw_set_src0(p, insn, brw_ip_reg()); |
brw_set_src1(p, insn, brw_imm_d(0x0)); |
insn->bits3.if_else.pad0 = 0; |
insn->bits3.if_else.pop_count = pop_count; |
} |
insn->header.compression_control = BRW_COMPRESSION_NONE; |
insn->header.execution_size = BRW_EXECUTE_8; |
return insn; |
} |
struct brw_instruction *gen6_CONT(struct brw_compile *p, |
struct brw_instruction *do_insn) |
{ |
struct brw_instruction *insn; |
insn = brw_next_insn(p, BRW_OPCODE_CONTINUE); |
brw_set_dest(p, insn, __retype_d(brw_null_reg())); |
brw_set_src0(p, insn, __retype_d(brw_null_reg())); |
brw_set_dest(p, insn, brw_ip_reg()); |
brw_set_src0(p, insn, brw_ip_reg()); |
brw_set_src1(p, insn, brw_imm_d(0x0)); |
insn->header.compression_control = BRW_COMPRESSION_NONE; |
insn->header.execution_size = BRW_EXECUTE_8; |
return insn; |
} |
struct brw_instruction *brw_CONT(struct brw_compile *p, int pop_count) |
{ |
struct brw_instruction *insn; |
insn = brw_next_insn(p, BRW_OPCODE_CONTINUE); |
brw_set_dest(p, insn, brw_ip_reg()); |
brw_set_src0(p, insn, brw_ip_reg()); |
brw_set_src1(p, insn, brw_imm_d(0x0)); |
insn->header.compression_control = BRW_COMPRESSION_NONE; |
insn->header.execution_size = BRW_EXECUTE_8; |
/* insn->header.mask_control = BRW_MASK_DISABLE; */ |
insn->bits3.if_else.pad0 = 0; |
insn->bits3.if_else.pop_count = pop_count; |
return insn; |
} |
/* DO/WHILE loop: |
* |
* The DO/WHILE is just an unterminated loop -- break or continue are |
* used for control within the loop. We have a few ways they can be |
* done. |
* |
* For uniform control flow, the WHILE is just a jump, so ADD ip, ip, |
* jip and no DO instruction. |
* |
* For non-uniform control flow pre-gen6, there's a DO instruction to |
* push the mask, and a WHILE to jump back, and BREAK to get out and |
* pop the mask. |
* |
* For gen6, there's no more mask stack, so no need for DO. WHILE |
* just points back to the first instruction of the loop. |
*/ |
struct brw_instruction *brw_DO(struct brw_compile *p, unsigned execute_size) |
{ |
if (p->gen >= 060 || p->single_program_flow) { |
return &p->store[p->nr_insn]; |
} else { |
struct brw_instruction *insn = brw_next_insn(p, BRW_OPCODE_DO); |
/* Override the defaults for this instruction: |
*/ |
brw_set_dest(p, insn, brw_null_reg()); |
brw_set_src0(p, insn, brw_null_reg()); |
brw_set_src1(p, insn, brw_null_reg()); |
insn->header.compression_control = BRW_COMPRESSION_NONE; |
insn->header.execution_size = execute_size; |
insn->header.predicate_control = BRW_PREDICATE_NONE; |
/* insn->header.mask_control = BRW_MASK_ENABLE; */ |
/* insn->header.mask_control = BRW_MASK_DISABLE; */ |
return insn; |
} |
} |
struct brw_instruction *brw_WHILE(struct brw_compile *p, |
struct brw_instruction *do_insn) |
{ |
struct brw_instruction *insn; |
unsigned br = 1; |
if (p->gen >= 050) |
br = 2; |
if (p->gen >= 070) { |
insn = brw_next_insn(p, BRW_OPCODE_WHILE); |
brw_set_dest(p, insn, __retype_d(brw_null_reg())); |
brw_set_src0(p, insn, __retype_d(brw_null_reg())); |
brw_set_src1(p, insn, brw_imm_ud(0)); |
insn->bits3.break_cont.jip = br * (do_insn - insn); |
insn->header.execution_size = BRW_EXECUTE_8; |
} else if (p->gen >= 060) { |
insn = brw_next_insn(p, BRW_OPCODE_WHILE); |
brw_set_dest(p, insn, brw_imm_w(0)); |
insn->bits1.branch_gen6.jump_count = br * (do_insn - insn); |
brw_set_src0(p, insn, __retype_d(brw_null_reg())); |
brw_set_src1(p, insn, __retype_d(brw_null_reg())); |
insn->header.execution_size = BRW_EXECUTE_8; |
} else { |
if (p->single_program_flow) { |
insn = brw_next_insn(p, BRW_OPCODE_ADD); |
brw_set_dest(p, insn, brw_ip_reg()); |
brw_set_src0(p, insn, brw_ip_reg()); |
brw_set_src1(p, insn, brw_imm_d((do_insn - insn) * 16)); |
insn->header.execution_size = BRW_EXECUTE_1; |
} else { |
insn = brw_next_insn(p, BRW_OPCODE_WHILE); |
assert(do_insn->header.opcode == BRW_OPCODE_DO); |
brw_set_dest(p, insn, brw_ip_reg()); |
brw_set_src0(p, insn, brw_ip_reg()); |
brw_set_src1(p, insn, brw_imm_d(0)); |
insn->header.execution_size = do_insn->header.execution_size; |
insn->bits3.if_else.jump_count = br * (do_insn - insn + 1); |
insn->bits3.if_else.pop_count = 0; |
insn->bits3.if_else.pad0 = 0; |
} |
} |
insn->header.compression_control = BRW_COMPRESSION_NONE; |
p->current->header.predicate_control = BRW_PREDICATE_NONE; |
return insn; |
} |
/* FORWARD JUMPS: |
*/ |
void brw_land_fwd_jump(struct brw_compile *p, |
struct brw_instruction *jmp_insn) |
{ |
struct brw_instruction *landing = &p->store[p->nr_insn]; |
unsigned jmpi = 1; |
if (p->gen >= 050) |
jmpi = 2; |
assert(jmp_insn->header.opcode == BRW_OPCODE_JMPI); |
assert(jmp_insn->bits1.da1.src1_reg_file == BRW_IMMEDIATE_VALUE); |
jmp_insn->bits3.ud = jmpi * ((landing - jmp_insn) - 1); |
} |
/* To integrate with the above, it makes sense that the comparison |
* instruction should populate the flag register. It might be simpler |
* just to use the flag reg for most WM tasks? |
*/ |
void brw_CMP(struct brw_compile *p, |
struct brw_reg dest, |
unsigned conditional, |
struct brw_reg src0, |
struct brw_reg src1) |
{ |
struct brw_instruction *insn = brw_next_insn(p, BRW_OPCODE_CMP); |
insn->header.destreg__conditionalmod = conditional; |
brw_set_dest(p, insn, dest); |
brw_set_src0(p, insn, src0); |
brw_set_src1(p, insn, src1); |
/* Make it so that future instructions will use the computed flag |
* value until brw_set_predicate_control_flag_value() is called |
* again. |
*/ |
if (dest.file == BRW_ARCHITECTURE_REGISTER_FILE && |
dest.nr == 0) { |
p->current->header.predicate_control = BRW_PREDICATE_NORMAL; |
p->flag_value = 0xff; |
} |
} |
/* Issue 'wait' instruction for n1, host could program MMIO |
to wake up thread. */ |
void brw_WAIT(struct brw_compile *p) |
{ |
struct brw_instruction *insn = brw_next_insn(p, BRW_OPCODE_WAIT); |
struct brw_reg src = brw_notification_1_reg(); |
brw_set_dest(p, insn, src); |
brw_set_src0(p, insn, src); |
brw_set_src1(p, insn, brw_null_reg()); |
insn->header.execution_size = 0; /* must */ |
insn->header.predicate_control = 0; |
insn->header.compression_control = 0; |
} |
/*********************************************************************** |
* Helpers for the various SEND message types: |
*/ |
/** Extended math function, float[8]. |
*/ |
void brw_math(struct brw_compile *p, |
struct brw_reg dest, |
unsigned function, |
unsigned saturate, |
unsigned msg_reg_nr, |
struct brw_reg src, |
unsigned data_type, |
unsigned precision) |
{ |
if (p->gen >= 060) { |
struct brw_instruction *insn = brw_next_insn(p, BRW_OPCODE_MATH); |
assert(dest.file == BRW_GENERAL_REGISTER_FILE); |
assert(src.file == BRW_GENERAL_REGISTER_FILE); |
assert(dest.hstride == BRW_HORIZONTAL_STRIDE_1); |
assert(src.hstride == BRW_HORIZONTAL_STRIDE_1); |
/* Source modifiers are ignored for extended math instructions. */ |
assert(!src.negate); |
assert(!src.abs); |
if (function != BRW_MATH_FUNCTION_INT_DIV_QUOTIENT && |
function != BRW_MATH_FUNCTION_INT_DIV_QUOTIENT_AND_REMAINDER) { |
assert(src.type == BRW_REGISTER_TYPE_F); |
} |
/* Math is the same ISA format as other opcodes, except that CondModifier |
* becomes FC[3:0] and ThreadCtrl becomes FC[5:4]. |
*/ |
insn->header.destreg__conditionalmod = function; |
insn->header.saturate = saturate; |
brw_set_dest(p, insn, dest); |
brw_set_src0(p, insn, src); |
brw_set_src1(p, insn, brw_null_reg()); |
} else { |
struct brw_instruction *insn = brw_next_insn(p, BRW_OPCODE_SEND); |
/* Example code doesn't set predicate_control for send |
* instructions. |
*/ |
insn->header.predicate_control = 0; |
insn->header.destreg__conditionalmod = msg_reg_nr; |
brw_set_dest(p, insn, dest); |
brw_set_src0(p, insn, src); |
brw_set_math_message(p, insn, function, |
src.type == BRW_REGISTER_TYPE_D, |
precision, |
saturate, |
data_type); |
} |
} |
/** Extended math function, float[8]. |
*/ |
void brw_math2(struct brw_compile *p, |
struct brw_reg dest, |
unsigned function, |
struct brw_reg src0, |
struct brw_reg src1) |
{ |
struct brw_instruction *insn = brw_next_insn(p, BRW_OPCODE_MATH); |
assert(dest.file == BRW_GENERAL_REGISTER_FILE); |
assert(src0.file == BRW_GENERAL_REGISTER_FILE); |
assert(src1.file == BRW_GENERAL_REGISTER_FILE); |
assert(dest.hstride == BRW_HORIZONTAL_STRIDE_1); |
assert(src0.hstride == BRW_HORIZONTAL_STRIDE_1); |
assert(src1.hstride == BRW_HORIZONTAL_STRIDE_1); |
if (function != BRW_MATH_FUNCTION_INT_DIV_QUOTIENT && |
function != BRW_MATH_FUNCTION_INT_DIV_QUOTIENT_AND_REMAINDER) { |
assert(src0.type == BRW_REGISTER_TYPE_F); |
assert(src1.type == BRW_REGISTER_TYPE_F); |
} |
/* Source modifiers are ignored for extended math instructions. */ |
assert(!src0.negate); |
assert(!src0.abs); |
assert(!src1.negate); |
assert(!src1.abs); |
/* Math is the same ISA format as other opcodes, except that CondModifier |
* becomes FC[3:0] and ThreadCtrl becomes FC[5:4]. |
*/ |
insn->header.destreg__conditionalmod = function; |
brw_set_dest(p, insn, dest); |
brw_set_src0(p, insn, src0); |
brw_set_src1(p, insn, src1); |
} |
/** |
* Extended math function, float[16]. |
* Use 2 send instructions. |
*/ |
void brw_math_16(struct brw_compile *p, |
struct brw_reg dest, |
unsigned function, |
unsigned saturate, |
unsigned msg_reg_nr, |
struct brw_reg src, |
unsigned precision) |
{ |
struct brw_instruction *insn; |
if (p->gen >= 060) { |
insn = brw_next_insn(p, BRW_OPCODE_MATH); |
/* Math is the same ISA format as other opcodes, except that CondModifier |
* becomes FC[3:0] and ThreadCtrl becomes FC[5:4]. |
*/ |
insn->header.destreg__conditionalmod = function; |
insn->header.saturate = saturate; |
/* Source modifiers are ignored for extended math instructions. */ |
assert(!src.negate); |
assert(!src.abs); |
brw_set_dest(p, insn, dest); |
brw_set_src0(p, insn, src); |
brw_set_src1(p, insn, brw_null_reg()); |
return; |
} |
/* First instruction: |
*/ |
brw_push_insn_state(p); |
brw_set_predicate_control_flag_value(p, 0xff); |
brw_set_compression_control(p, BRW_COMPRESSION_NONE); |
insn = brw_next_insn(p, BRW_OPCODE_SEND); |
insn->header.destreg__conditionalmod = msg_reg_nr; |
brw_set_dest(p, insn, dest); |
brw_set_src0(p, insn, src); |
brw_set_math_message(p, insn, function, |
BRW_MATH_INTEGER_UNSIGNED, |
precision, |
saturate, |
BRW_MATH_DATA_VECTOR); |
/* Second instruction: |
*/ |
insn = brw_next_insn(p, BRW_OPCODE_SEND); |
insn->header.compression_control = BRW_COMPRESSION_2NDHALF; |
insn->header.destreg__conditionalmod = msg_reg_nr+1; |
brw_set_dest(p, insn, __offset(dest,1)); |
brw_set_src0(p, insn, src); |
brw_set_math_message(p, insn, function, |
BRW_MATH_INTEGER_UNSIGNED, |
precision, |
saturate, |
BRW_MATH_DATA_VECTOR); |
brw_pop_insn_state(p); |
} |
/** |
* Write a block of OWORDs (half a GRF each) from the scratch buffer, |
* using a constant offset per channel. |
* |
* The offset must be aligned to oword size (16 bytes). Used for |
* register spilling. |
*/ |
void brw_oword_block_write_scratch(struct brw_compile *p, |
struct brw_reg mrf, |
int num_regs, |
unsigned offset) |
{ |
uint32_t msg_control, msg_type; |
int mlen; |
if (p->gen >= 060) |
offset /= 16; |
mrf = __retype_ud(mrf); |
if (num_regs == 1) { |
msg_control = BRW_DATAPORT_OWORD_BLOCK_2_OWORDS; |
mlen = 2; |
} else { |
msg_control = BRW_DATAPORT_OWORD_BLOCK_4_OWORDS; |
mlen = 3; |
} |
/* Set up the message header. This is g0, with g0.2 filled with |
* the offset. We don't want to leave our offset around in g0 or |
* it'll screw up texture samples, so set it up inside the message |
* reg. |
*/ |
{ |
brw_push_insn_state(p); |
brw_set_mask_control(p, BRW_MASK_DISABLE); |
brw_set_compression_control(p, BRW_COMPRESSION_NONE); |
brw_MOV(p, mrf, __retype_ud(brw_vec8_grf(0, 0))); |
/* set message header global offset field (reg 0, element 2) */ |
brw_MOV(p, |
__retype_ud(brw_vec1_reg(BRW_MESSAGE_REGISTER_FILE, mrf.nr, 2)), |
brw_imm_ud(offset)); |
brw_pop_insn_state(p); |
} |
{ |
struct brw_reg dest; |
struct brw_instruction *insn = brw_next_insn(p, BRW_OPCODE_SEND); |
int send_commit_msg; |
struct brw_reg src_header = __retype_uw(brw_vec8_grf(0, 0)); |
if (insn->header.compression_control != BRW_COMPRESSION_NONE) { |
insn->header.compression_control = BRW_COMPRESSION_NONE; |
src_header = vec16(src_header); |
} |
assert(insn->header.predicate_control == BRW_PREDICATE_NONE); |
insn->header.destreg__conditionalmod = mrf.nr; |
/* Until gen6, writes followed by reads from the same location |
* are not guaranteed to be ordered unless write_commit is set. |
* If set, then a no-op write is issued to the destination |
* register to set a dependency, and a read from the destination |
* can be used to ensure the ordering. |
* |
* For gen6, only writes between different threads need ordering |
* protection. Our use of DP writes is all about register |
* spilling within a thread. |
*/ |
if (p->gen >= 060) { |
dest = __retype_uw(vec16(brw_null_reg())); |
send_commit_msg = 0; |
} else { |
dest = src_header; |
send_commit_msg = 1; |
} |
brw_set_dest(p, insn, dest); |
if (p->gen >= 060) { |
brw_set_src0(p, insn, mrf); |
} else { |
brw_set_src0(p, insn, brw_null_reg()); |
} |
if (p->gen >= 060) |
msg_type = GEN6_DATAPORT_WRITE_MESSAGE_OWORD_BLOCK_WRITE; |
else |
msg_type = BRW_DATAPORT_WRITE_MESSAGE_OWORD_BLOCK_WRITE; |
brw_set_dp_write_message(p, |
insn, |
255, /* binding table index (255=stateless) */ |
msg_control, |
msg_type, |
mlen, |
true, /* header_present */ |
0, /* pixel scoreboard */ |
send_commit_msg, /* response_length */ |
0, /* eot */ |
send_commit_msg); |
} |
} |
/** |
* Read a block of owords (half a GRF each) from the scratch buffer |
* using a constant index per channel. |
* |
* Offset must be aligned to oword size (16 bytes). Used for register |
* spilling. |
*/ |
void |
brw_oword_block_read_scratch(struct brw_compile *p, |
struct brw_reg dest, |
struct brw_reg mrf, |
int num_regs, |
unsigned offset) |
{ |
uint32_t msg_control; |
int rlen; |
if (p->gen >= 060) |
offset /= 16; |
mrf = __retype_ud(mrf); |
dest = __retype_uw(dest); |
if (num_regs == 1) { |
msg_control = BRW_DATAPORT_OWORD_BLOCK_2_OWORDS; |
rlen = 1; |
} else { |
msg_control = BRW_DATAPORT_OWORD_BLOCK_4_OWORDS; |
rlen = 2; |
} |
{ |
brw_push_insn_state(p); |
brw_set_compression_control(p, BRW_COMPRESSION_NONE); |
brw_set_mask_control(p, BRW_MASK_DISABLE); |
brw_MOV(p, mrf, __retype_ud(brw_vec8_grf(0, 0))); |
/* set message header global offset field (reg 0, element 2) */ |
brw_MOV(p, |
__retype_ud(brw_vec1_reg(BRW_MESSAGE_REGISTER_FILE, mrf.nr, 2)), |
brw_imm_ud(offset)); |
brw_pop_insn_state(p); |
} |
{ |
struct brw_instruction *insn = brw_next_insn(p, BRW_OPCODE_SEND); |
assert(insn->header.predicate_control == 0); |
insn->header.compression_control = BRW_COMPRESSION_NONE; |
insn->header.destreg__conditionalmod = mrf.nr; |
brw_set_dest(p, insn, dest); /* UW? */ |
if (p->gen >= 060) { |
brw_set_src0(p, insn, mrf); |
} else { |
brw_set_src0(p, insn, brw_null_reg()); |
} |
brw_set_dp_read_message(p, |
insn, |
255, /* binding table index (255=stateless) */ |
msg_control, |
BRW_DATAPORT_READ_MESSAGE_OWORD_BLOCK_READ, /* msg_type */ |
BRW_DATAPORT_READ_TARGET_RENDER_CACHE, |
1, /* msg_length */ |
rlen); |
} |
} |
/** |
* Read a float[4] vector from the data port Data Cache (const buffer). |
* Location (in buffer) should be a multiple of 16. |
* Used for fetching shader constants. |
*/ |
void brw_oword_block_read(struct brw_compile *p, |
struct brw_reg dest, |
struct brw_reg mrf, |
uint32_t offset, |
uint32_t bind_table_index) |
{ |
struct brw_instruction *insn; |
/* On newer hardware, offset is in units of owords. */ |
if (p->gen >= 060) |
offset /= 16; |
mrf = __retype_ud(mrf); |
brw_push_insn_state(p); |
brw_set_predicate_control(p, BRW_PREDICATE_NONE); |
brw_set_compression_control(p, BRW_COMPRESSION_NONE); |
brw_set_mask_control(p, BRW_MASK_DISABLE); |
brw_MOV(p, mrf, __retype_ud(brw_vec8_grf(0, 0))); |
/* set message header global offset field (reg 0, element 2) */ |
brw_MOV(p, |
__retype_ud(brw_vec1_reg(BRW_MESSAGE_REGISTER_FILE, mrf.nr, 2)), |
brw_imm_ud(offset)); |
insn = brw_next_insn(p, BRW_OPCODE_SEND); |
insn->header.destreg__conditionalmod = mrf.nr; |
/* cast dest to a uword[8] vector */ |
dest = __retype_uw(vec8(dest)); |
brw_set_dest(p, insn, dest); |
if (p->gen >= 060) { |
brw_set_src0(p, insn, mrf); |
} else { |
brw_set_src0(p, insn, brw_null_reg()); |
} |
brw_set_dp_read_message(p, |
insn, |
bind_table_index, |
BRW_DATAPORT_OWORD_BLOCK_1_OWORDLOW, |
BRW_DATAPORT_READ_MESSAGE_OWORD_BLOCK_READ, |
BRW_DATAPORT_READ_TARGET_DATA_CACHE, |
1, /* msg_length */ |
1); /* response_length (1 reg, 2 owords!) */ |
brw_pop_insn_state(p); |
} |
/** |
* Read a set of dwords from the data port Data Cache (const buffer). |
* |
* Location (in buffer) appears as UD offsets in the register after |
* the provided mrf header reg. |
*/ |
void brw_dword_scattered_read(struct brw_compile *p, |
struct brw_reg dest, |
struct brw_reg mrf, |
uint32_t bind_table_index) |
{ |
struct brw_instruction *insn; |
mrf = __retype_ud(mrf); |
brw_push_insn_state(p); |
brw_set_predicate_control(p, BRW_PREDICATE_NONE); |
brw_set_compression_control(p, BRW_COMPRESSION_NONE); |
brw_set_mask_control(p, BRW_MASK_DISABLE); |
brw_MOV(p, mrf, __retype_ud(brw_vec8_grf(0, 0))); |
brw_pop_insn_state(p); |
insn = brw_next_insn(p, BRW_OPCODE_SEND); |
insn->header.destreg__conditionalmod = mrf.nr; |
/* cast dest to a uword[8] vector */ |
dest = __retype_uw(vec8(dest)); |
brw_set_dest(p, insn, dest); |
brw_set_src0(p, insn, brw_null_reg()); |
brw_set_dp_read_message(p, |
insn, |
bind_table_index, |
BRW_DATAPORT_DWORD_SCATTERED_BLOCK_8DWORDS, |
BRW_DATAPORT_READ_MESSAGE_DWORD_SCATTERED_READ, |
BRW_DATAPORT_READ_TARGET_DATA_CACHE, |
2, /* msg_length */ |
1); /* response_length */ |
} |
/** |
* Read float[4] constant(s) from VS constant buffer. |
* For relative addressing, two float[4] constants will be read into 'dest'. |
* Otherwise, one float[4] constant will be read into the lower half of 'dest'. |
*/ |
void brw_dp_READ_4_vs(struct brw_compile *p, |
struct brw_reg dest, |
unsigned location, |
unsigned bind_table_index) |
{ |
struct brw_instruction *insn; |
unsigned msg_reg_nr = 1; |
if (p->gen >= 060) |
location /= 16; |
/* Setup MRF[1] with location/offset into const buffer */ |
brw_push_insn_state(p); |
brw_set_access_mode(p, BRW_ALIGN_1); |
brw_set_compression_control(p, BRW_COMPRESSION_NONE); |
brw_set_mask_control(p, BRW_MASK_DISABLE); |
brw_set_predicate_control(p, BRW_PREDICATE_NONE); |
brw_MOV(p, __retype_ud(brw_vec1_reg(BRW_MESSAGE_REGISTER_FILE, msg_reg_nr, 2)), |
brw_imm_ud(location)); |
brw_pop_insn_state(p); |
insn = brw_next_insn(p, BRW_OPCODE_SEND); |
insn->header.predicate_control = BRW_PREDICATE_NONE; |
insn->header.compression_control = BRW_COMPRESSION_NONE; |
insn->header.destreg__conditionalmod = msg_reg_nr; |
insn->header.mask_control = BRW_MASK_DISABLE; |
brw_set_dest(p, insn, dest); |
if (p->gen >= 060) { |
brw_set_src0(p, insn, brw_message_reg(msg_reg_nr)); |
} else { |
brw_set_src0(p, insn, brw_null_reg()); |
} |
brw_set_dp_read_message(p, |
insn, |
bind_table_index, |
0, |
BRW_DATAPORT_READ_MESSAGE_OWORD_BLOCK_READ, /* msg_type */ |
BRW_DATAPORT_READ_TARGET_DATA_CACHE, |
1, /* msg_length */ |
1); /* response_length (1 Oword) */ |
} |
/** |
* Read a float[4] constant per vertex from VS constant buffer, with |
* relative addressing. |
*/ |
void brw_dp_READ_4_vs_relative(struct brw_compile *p, |
struct brw_reg dest, |
struct brw_reg addr_reg, |
unsigned offset, |
unsigned bind_table_index) |
{ |
struct brw_reg src = brw_vec8_grf(0, 0); |
struct brw_instruction *insn; |
int msg_type; |
/* Setup MRF[1] with offset into const buffer */ |
brw_push_insn_state(p); |
brw_set_access_mode(p, BRW_ALIGN_1); |
brw_set_compression_control(p, BRW_COMPRESSION_NONE); |
brw_set_mask_control(p, BRW_MASK_DISABLE); |
brw_set_predicate_control(p, BRW_PREDICATE_NONE); |
/* M1.0 is block offset 0, M1.4 is block offset 1, all other |
* fields ignored. |
*/ |
brw_ADD(p, __retype_d(brw_message_reg(1)), |
addr_reg, brw_imm_d(offset)); |
brw_pop_insn_state(p); |
gen6_resolve_implied_move(p, &src, 0); |
insn = brw_next_insn(p, BRW_OPCODE_SEND); |
insn->header.predicate_control = BRW_PREDICATE_NONE; |
insn->header.compression_control = BRW_COMPRESSION_NONE; |
insn->header.destreg__conditionalmod = 0; |
insn->header.mask_control = BRW_MASK_DISABLE; |
brw_set_dest(p, insn, dest); |
brw_set_src0(p, insn, src); |
if (p->gen >= 060) |
msg_type = GEN6_DATAPORT_READ_MESSAGE_OWORD_DUAL_BLOCK_READ; |
else if (p->gen >= 045) |
msg_type = G45_DATAPORT_READ_MESSAGE_OWORD_DUAL_BLOCK_READ; |
else |
msg_type = BRW_DATAPORT_READ_MESSAGE_OWORD_DUAL_BLOCK_READ; |
brw_set_dp_read_message(p, |
insn, |
bind_table_index, |
BRW_DATAPORT_OWORD_DUAL_BLOCK_1OWORD, |
msg_type, |
BRW_DATAPORT_READ_TARGET_DATA_CACHE, |
2, /* msg_length */ |
1); /* response_length */ |
} |
void brw_fb_WRITE(struct brw_compile *p, |
int dispatch_width, |
unsigned msg_reg_nr, |
struct brw_reg src0, |
unsigned msg_control, |
unsigned binding_table_index, |
unsigned msg_length, |
unsigned response_length, |
bool eot, |
bool header_present) |
{ |
struct brw_instruction *insn; |
unsigned msg_type; |
struct brw_reg dest; |
if (dispatch_width == 16) |
dest = __retype_uw(vec16(brw_null_reg())); |
else |
dest = __retype_uw(vec8(brw_null_reg())); |
if (p->gen >= 060 && binding_table_index == 0) { |
insn = brw_next_insn(p, BRW_OPCODE_SENDC); |
} else { |
insn = brw_next_insn(p, BRW_OPCODE_SEND); |
} |
/* The execution mask is ignored for render target writes. */ |
insn->header.predicate_control = 0; |
insn->header.compression_control = BRW_COMPRESSION_NONE; |
if (p->gen >= 060) { |
/* headerless version, just submit color payload */ |
src0 = brw_message_reg(msg_reg_nr); |
msg_type = GEN6_DATAPORT_WRITE_MESSAGE_RENDER_TARGET_WRITE; |
} else { |
insn->header.destreg__conditionalmod = msg_reg_nr; |
msg_type = BRW_DATAPORT_WRITE_MESSAGE_RENDER_TARGET_WRITE; |
} |
brw_set_dest(p, insn, dest); |
brw_set_src0(p, insn, src0); |
brw_set_dp_write_message(p, |
insn, |
binding_table_index, |
msg_control, |
msg_type, |
msg_length, |
header_present, |
eot, |
response_length, |
eot, |
0 /* send_commit_msg */); |
} |
/** |
* Texture sample instruction. |
* Note: the msg_type plus msg_length values determine exactly what kind |
* of sampling operation is performed. See volume 4, page 161 of docs. |
*/ |
void brw_SAMPLE(struct brw_compile *p, |
struct brw_reg dest, |
unsigned msg_reg_nr, |
struct brw_reg src0, |
unsigned binding_table_index, |
unsigned sampler, |
unsigned writemask, |
unsigned msg_type, |
unsigned response_length, |
unsigned msg_length, |
bool header_present, |
unsigned simd_mode) |
{ |
assert(writemask); |
if (p->gen < 050 || writemask != WRITEMASK_XYZW) { |
struct brw_reg m1 = brw_message_reg(msg_reg_nr); |
writemask = ~writemask & WRITEMASK_XYZW; |
brw_push_insn_state(p); |
brw_set_compression_control(p, BRW_COMPRESSION_NONE); |
brw_set_mask_control(p, BRW_MASK_DISABLE); |
brw_MOV(p, __retype_ud(m1), __retype_ud(brw_vec8_grf(0,0))); |
brw_MOV(p, get_element_ud(m1, 2), brw_imm_ud(writemask << 12)); |
brw_pop_insn_state(p); |
src0 = __retype_uw(brw_null_reg()); |
} |
{ |
struct brw_instruction *insn; |
gen6_resolve_implied_move(p, &src0, msg_reg_nr); |
insn = brw_next_insn(p, BRW_OPCODE_SEND); |
insn->header.predicate_control = 0; /* XXX */ |
insn->header.compression_control = BRW_COMPRESSION_NONE; |
if (p->gen < 060) |
insn->header.destreg__conditionalmod = msg_reg_nr; |
brw_set_dest(p, insn, dest); |
brw_set_src0(p, insn, src0); |
brw_set_sampler_message(p, insn, |
binding_table_index, |
sampler, |
msg_type, |
response_length, |
msg_length, |
header_present, |
simd_mode); |
} |
} |
/* All these variables are pretty confusing - we might be better off |
* using bitmasks and macros for this, in the old style. Or perhaps |
* just having the caller instantiate the fields in dword3 itself. |
*/ |
void brw_urb_WRITE(struct brw_compile *p, |
struct brw_reg dest, |
unsigned msg_reg_nr, |
struct brw_reg src0, |
bool allocate, |
bool used, |
unsigned msg_length, |
unsigned response_length, |
bool eot, |
bool writes_complete, |
unsigned offset, |
unsigned swizzle) |
{ |
struct brw_instruction *insn; |
gen6_resolve_implied_move(p, &src0, msg_reg_nr); |
if (p->gen >= 070) { |
/* Enable Channel Masks in the URB_WRITE_HWORD message header */ |
brw_push_insn_state(p); |
brw_set_access_mode(p, BRW_ALIGN_1); |
brw_OR(p, __retype_ud(brw_vec1_reg(BRW_MESSAGE_REGISTER_FILE, msg_reg_nr, 5)), |
__retype_ud(brw_vec1_grf(0, 5)), |
brw_imm_ud(0xff00)); |
brw_pop_insn_state(p); |
} |
insn = brw_next_insn(p, BRW_OPCODE_SEND); |
assert(msg_length < BRW_MAX_MRF); |
brw_set_dest(p, insn, dest); |
brw_set_src0(p, insn, src0); |
brw_set_src1(p, insn, brw_imm_d(0)); |
if (p->gen <= 060) |
insn->header.destreg__conditionalmod = msg_reg_nr; |
brw_set_urb_message(p, |
insn, |
allocate, |
used, |
msg_length, |
response_length, |
eot, |
writes_complete, |
offset, |
swizzle); |
} |
static int |
brw_find_next_block_end(struct brw_compile *p, int start) |
{ |
int ip; |
for (ip = start + 1; ip < p->nr_insn; ip++) { |
struct brw_instruction *insn = &p->store[ip]; |
switch (insn->header.opcode) { |
case BRW_OPCODE_ENDIF: |
case BRW_OPCODE_ELSE: |
case BRW_OPCODE_WHILE: |
return ip; |
} |
} |
assert(!"not reached"); |
return start + 1; |
} |
/* There is no DO instruction on gen6, so to find the end of the loop |
* we have to see if the loop is jumping back before our start |
* instruction. |
*/ |
static int |
brw_find_loop_end(struct brw_compile *p, int start) |
{ |
int ip; |
int br = 2; |
for (ip = start + 1; ip < p->nr_insn; ip++) { |
struct brw_instruction *insn = &p->store[ip]; |
if (insn->header.opcode == BRW_OPCODE_WHILE) { |
int jip = p->gen <= 070 ? insn->bits1.branch_gen6.jump_count |
: insn->bits3.break_cont.jip; |
if (ip + jip / br <= start) |
return ip; |
} |
} |
assert(!"not reached"); |
return start + 1; |
} |
/* After program generation, go back and update the UIP and JIP of |
* BREAK and CONT instructions to their correct locations. |
*/ |
void |
brw_set_uip_jip(struct brw_compile *p) |
{ |
int ip; |
int br = 2; |
if (p->gen <= 060) |
return; |
for (ip = 0; ip < p->nr_insn; ip++) { |
struct brw_instruction *insn = &p->store[ip]; |
switch (insn->header.opcode) { |
case BRW_OPCODE_BREAK: |
insn->bits3.break_cont.jip = br * (brw_find_next_block_end(p, ip) - ip); |
/* Gen7 UIP points to WHILE; Gen6 points just after it */ |
insn->bits3.break_cont.uip = |
br * (brw_find_loop_end(p, ip) - ip + (p->gen <= 070 ? 1 : 0)); |
break; |
case BRW_OPCODE_CONTINUE: |
insn->bits3.break_cont.jip = br * (brw_find_next_block_end(p, ip) - ip); |
insn->bits3.break_cont.uip = br * (brw_find_loop_end(p, ip) - ip); |
assert(insn->bits3.break_cont.uip != 0); |
assert(insn->bits3.break_cont.jip != 0); |
break; |
} |
} |
} |
void brw_ff_sync(struct brw_compile *p, |
struct brw_reg dest, |
unsigned msg_reg_nr, |
struct brw_reg src0, |
bool allocate, |
unsigned response_length, |
bool eot) |
{ |
struct brw_instruction *insn; |
gen6_resolve_implied_move(p, &src0, msg_reg_nr); |
insn = brw_next_insn(p, BRW_OPCODE_SEND); |
brw_set_dest(p, insn, dest); |
brw_set_src0(p, insn, src0); |
brw_set_src1(p, insn, brw_imm_d(0)); |
if (p->gen < 060) |
insn->header.destreg__conditionalmod = msg_reg_nr; |
brw_set_ff_sync_message(p, |
insn, |
allocate, |
response_length, |
eot); |
} |
/drivers/video/Intel-2D/sna/brw/brw_sf.c |
---|
0,0 → 1,54 |
#include "brw.h" |
bool brw_sf_kernel__nomask(struct brw_compile *p) |
{ |
struct brw_reg inv, v0, v1, v2, delta; |
v0 = brw_vec4_grf(3, 0); |
v1 = brw_vec4_grf(4, 0); |
v2 = brw_vec4_grf(5, 0); |
delta = brw_vec8_grf(7, 0); |
inv = brw_vec4_grf(6, 0); |
brw_math_invert(p, inv, brw_vec4_grf(1, 11)); |
brw_MOV(p, brw_message_reg(3), v0); |
brw_ADD(p, delta, v1, brw_negate(v2)); |
brw_MUL(p, brw_message_reg(1), delta, brw_vec1_grf(6,0)); |
brw_ADD(p, delta, v2, brw_negate(v0)); |
brw_MUL(p, brw_message_reg(2), delta, brw_vec1_grf(6,2)); |
brw_urb_WRITE(p, brw_null_reg(), 0, brw_vec8_grf(0 ,0), |
false, true, 4, 0, true, true, 0, |
BRW_URB_SWIZZLE_TRANSPOSE); |
return true; |
} |
bool brw_sf_kernel__mask(struct brw_compile *p) |
{ |
struct brw_reg inv, v0, v1, v2; |
v0 = brw_vec8_grf(3, 0); |
v1 = brw_vec8_grf(4, 0); |
v2 = brw_vec8_grf(5, 0); |
inv = brw_vec4_grf(6, 0); |
brw_math_invert(p, inv, brw_vec4_grf(1, 11)); |
brw_MOV(p, brw_message_reg(3), v0); |
brw_ADD(p, brw_vec8_grf(7, 0), v1, brw_negate(v2)); |
brw_MUL(p, brw_message_reg(1), brw_vec8_grf(7, 0), brw_vec1_grf(6,0)); |
brw_ADD(p, brw_vec8_grf(7, 0), v2, brw_negate(v0)); |
brw_MUL(p, brw_message_reg(2), brw_vec8_grf(7, 0), brw_vec1_grf(6,2)); |
brw_urb_WRITE(p, brw_null_reg(), 0, brw_vec8_grf(0 ,0), |
false, true, 4, 0, true, true, 0, |
BRW_URB_SWIZZLE_TRANSPOSE); |
return true; |
} |
/drivers/video/Intel-2D/sna/brw/brw_wm.c |
---|
0,0 → 1,681 |
#include "brw.h" |
#define X16 8 |
#define Y16 10 |
static void brw_wm_xy(struct brw_compile *p, int dw) |
{ |
struct brw_reg r1 = brw_vec1_grf(1, 0); |
struct brw_reg r1_uw = __retype_uw(r1); |
struct brw_reg x_uw, y_uw; |
brw_set_compression_control(p, BRW_COMPRESSION_NONE); |
if (dw == 16) { |
x_uw = brw_uw16_grf(30, 0); |
y_uw = brw_uw16_grf(28, 0); |
} else { |
x_uw = brw_uw8_grf(30, 0); |
y_uw = brw_uw8_grf(28, 0); |
} |
brw_ADD(p, |
x_uw, |
__stride(__suboffset(r1_uw, 4), 2, 4, 0), |
brw_imm_v(0x10101010)); |
brw_ADD(p, |
y_uw, |
__stride(__suboffset(r1_uw, 5), 2, 4, 0), |
brw_imm_v(0x11001100)); |
brw_set_compression_control(p, BRW_COMPRESSION_COMPRESSED); |
brw_ADD(p, brw_vec8_grf(X16, 0), vec8(x_uw), brw_negate(r1)); |
brw_ADD(p, brw_vec8_grf(Y16, 0), vec8(y_uw), brw_negate(__suboffset(r1, 1))); |
} |
static void brw_wm_affine_st(struct brw_compile *p, int dw, |
int channel, int msg) |
{ |
int uv; |
if (dw == 16) { |
brw_set_compression_control(p, BRW_COMPRESSION_COMPRESSED); |
uv = p->gen >= 060 ? 6 : 3; |
} else { |
brw_set_compression_control(p, BRW_COMPRESSION_NONE); |
uv = p->gen >= 060 ? 4 : 3; |
} |
uv += 2*channel; |
msg++; |
if (p->gen >= 060) { |
brw_PLN(p, |
brw_message_reg(msg), |
brw_vec1_grf(uv, 0), |
brw_vec8_grf(2, 0)); |
msg += dw/8; |
brw_PLN(p, |
brw_message_reg(msg), |
brw_vec1_grf(uv, 4), |
brw_vec8_grf(2, 0)); |
} else { |
struct brw_reg r = brw_vec1_grf(uv, 0); |
brw_LINE(p, brw_null_reg(), __suboffset(r, 0), brw_vec8_grf(X16, 0)); |
brw_MAC(p, brw_message_reg(msg), __suboffset(r, 1), brw_vec8_grf(Y16, 0)); |
msg += dw/8; |
brw_LINE(p, brw_null_reg(), __suboffset(r, 4), brw_vec8_grf(X16, 0)); |
brw_MAC(p, brw_message_reg(msg), __suboffset(r, 5), brw_vec8_grf(Y16, 0)); |
} |
} |
static inline unsigned simd(int dw) |
{ |
return dw == 16 ? BRW_SAMPLER_SIMD_MODE_SIMD16 : BRW_SAMPLER_SIMD_MODE_SIMD8; |
} |
static inline struct brw_reg sample_result(int dw, int result) |
{ |
return brw_reg(BRW_GENERAL_REGISTER_FILE, result, 0, |
BRW_REGISTER_TYPE_UW, |
dw == 16 ? BRW_VERTICAL_STRIDE_16 : BRW_VERTICAL_STRIDE_8, |
dw == 16 ? BRW_WIDTH_16 : BRW_WIDTH_8, |
BRW_HORIZONTAL_STRIDE_1, |
BRW_SWIZZLE_XYZW, |
WRITEMASK_XYZW); |
} |
static int brw_wm_sample(struct brw_compile *p, int dw, |
int channel, int msg, int result) |
{ |
struct brw_reg src0; |
bool header; |
int len; |
len = dw == 16 ? 4 : 2; |
if (p->gen >= 060) { |
header = false; |
src0 = brw_message_reg(++msg); |
} else { |
header = true; |
src0 = brw_vec8_grf(0, 0); |
} |
brw_SAMPLE(p, sample_result(dw, result), msg, src0, |
channel+1, channel, WRITEMASK_XYZW, 0, |
2*len, len+header, header, simd(dw)); |
return result; |
} |
static int brw_wm_sample__alpha(struct brw_compile *p, int dw, |
int channel, int msg, int result) |
{ |
struct brw_reg src0; |
int mlen, rlen; |
if (dw == 8) { |
/* SIMD8 sample return is not masked */ |
mlen = 3; |
rlen = 4; |
} else { |
mlen = 5; |
rlen = 2; |
} |
if (p->gen >= 060) |
src0 = brw_message_reg(msg); |
else |
src0 = brw_vec8_grf(0, 0); |
brw_SAMPLE(p, sample_result(dw, result), msg, src0, |
channel+1, channel, WRITEMASK_W, 0, |
rlen, mlen, true, simd(dw)); |
if (dw == 8) |
result += 3; |
return result; |
} |
static int brw_wm_affine(struct brw_compile *p, int dw, |
int channel, int msg, int result) |
{ |
brw_wm_affine_st(p, dw, channel, msg); |
return brw_wm_sample(p, dw, channel, msg, result); |
} |
static int brw_wm_affine__alpha(struct brw_compile *p, int dw, |
int channel, int msg, int result) |
{ |
brw_wm_affine_st(p, dw, channel, msg); |
return brw_wm_sample__alpha(p, dw, channel, msg, result); |
} |
static inline struct brw_reg null_result(int dw) |
{ |
return brw_reg(BRW_ARCHITECTURE_REGISTER_FILE, BRW_ARF_NULL, 0, |
BRW_REGISTER_TYPE_UW, |
dw == 16 ? BRW_VERTICAL_STRIDE_16 : BRW_VERTICAL_STRIDE_8, |
dw == 16 ? BRW_WIDTH_16 : BRW_WIDTH_8, |
BRW_HORIZONTAL_STRIDE_1, |
BRW_SWIZZLE_XYZW, |
WRITEMASK_XYZW); |
} |
static void brw_fb_write(struct brw_compile *p, int dw) |
{ |
struct brw_instruction *insn; |
unsigned msg_control, msg_type, msg_len; |
struct brw_reg src0; |
bool header; |
if (dw == 16) { |
brw_set_compression_control(p, BRW_COMPRESSION_COMPRESSED); |
msg_control = BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD16_SINGLE_SOURCE; |
msg_len = 8; |
} else { |
brw_set_compression_control(p, BRW_COMPRESSION_NONE); |
msg_control = BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD8_SINGLE_SOURCE_SUBSPAN01; |
msg_len = 4; |
} |
if (p->gen < 060) { |
brw_push_insn_state(p); |
brw_set_compression_control(p, BRW_COMPRESSION_NONE); |
brw_set_mask_control(p, BRW_MASK_DISABLE); |
brw_MOV(p, brw_message_reg(1), brw_vec8_grf(1, 0)); |
brw_pop_insn_state(p); |
msg_len += 2; |
} |
/* The execution mask is ignored for render target writes. */ |
insn = brw_next_insn(p, BRW_OPCODE_SEND); |
insn->header.predicate_control = 0; |
insn->header.compression_control = BRW_COMPRESSION_NONE; |
if (p->gen >= 060) { |
msg_type = GEN6_DATAPORT_WRITE_MESSAGE_RENDER_TARGET_WRITE; |
src0 = brw_message_reg(2); |
header = false; |
} else { |
insn->header.destreg__conditionalmod = 0; |
msg_type = BRW_DATAPORT_WRITE_MESSAGE_RENDER_TARGET_WRITE; |
src0 = __retype_uw(brw_vec8_grf(0, 0)); |
header = true; |
} |
brw_set_dest(p, insn, null_result(dw)); |
brw_set_src0(p, insn, src0); |
brw_set_dp_write_message(p, insn, 0, |
msg_control, msg_type, msg_len, |
header, true, 0, true, false); |
} |
static void brw_wm_write(struct brw_compile *p, int dw, int src) |
{ |
int n; |
if (dw == 8 && p->gen >= 060) { |
/* XXX pixel execution mask? */ |
brw_set_compression_control(p, BRW_COMPRESSION_NONE); |
brw_MOV(p, brw_message_reg(2), brw_vec8_grf(src+0, 0)); |
brw_MOV(p, brw_message_reg(3), brw_vec8_grf(src+1, 0)); |
brw_MOV(p, brw_message_reg(4), brw_vec8_grf(src+2, 0)); |
brw_MOV(p, brw_message_reg(5), brw_vec8_grf(src+3, 0)); |
goto done; |
} |
brw_set_compression_control(p, BRW_COMPRESSION_COMPRESSED); |
for (n = 0; n < 4; n++) { |
if (p->gen >= 060) { |
brw_MOV(p, |
brw_message_reg(2 + 2*n), |
brw_vec8_grf(src + 2*n, 0)); |
} else if (p->gen >= 045 && dw == 16) { |
brw_MOV(p, |
brw_message_reg(2 + n + BRW_MRF_COMPR4), |
brw_vec8_grf(src + 2*n, 0)); |
} else { |
brw_set_compression_control(p, BRW_COMPRESSION_NONE); |
brw_MOV(p, |
brw_message_reg(2 + n), |
brw_vec8_grf(src + 2*n, 0)); |
if (dw == 16) { |
brw_set_compression_control(p, BRW_COMPRESSION_2NDHALF); |
brw_MOV(p, |
brw_message_reg(2 + n + 4), |
brw_vec8_grf(src + 2*n+1, 0)); |
} |
} |
} |
done: |
brw_fb_write(p, dw); |
} |
static void brw_wm_write__mask(struct brw_compile *p, int dw, |
int src, int mask) |
{ |
int n; |
if (dw == 8 && p->gen >= 060) { |
brw_set_compression_control(p, BRW_COMPRESSION_NONE); |
brw_MUL(p, |
brw_message_reg(2), |
brw_vec8_grf(src+0, 0), |
brw_vec8_grf(mask, 0)); |
brw_MUL(p, |
brw_message_reg(3), |
brw_vec8_grf(src+1, 0), |
brw_vec8_grf(mask, 0)); |
brw_MUL(p, |
brw_message_reg(4), |
brw_vec8_grf(src+2, 0), |
brw_vec8_grf(mask, 0)); |
brw_MUL(p, |
brw_message_reg(5), |
brw_vec8_grf(src+3, 0), |
brw_vec8_grf(mask, 0)); |
goto done; |
} |
brw_set_compression_control(p, BRW_COMPRESSION_COMPRESSED); |
for (n = 0; n < 4; n++) { |
if (p->gen >= 060) { |
brw_MUL(p, |
brw_message_reg(2 + 2*n), |
brw_vec8_grf(src + 2*n, 0), |
brw_vec8_grf(mask, 0)); |
} else if (p->gen >= 045 && dw == 16) { |
brw_MUL(p, |
brw_message_reg(2 + n + BRW_MRF_COMPR4), |
brw_vec8_grf(src + 2*n, 0), |
brw_vec8_grf(mask, 0)); |
} else { |
brw_set_compression_control(p, BRW_COMPRESSION_NONE); |
brw_MUL(p, |
brw_message_reg(2 + n), |
brw_vec8_grf(src + 2*n, 0), |
brw_vec8_grf(mask, 0)); |
if (dw == 16) { |
brw_set_compression_control(p, BRW_COMPRESSION_2NDHALF); |
brw_MUL(p, |
brw_message_reg(2 + n + 4), |
brw_vec8_grf(src + 2*n+1, 0), |
brw_vec8_grf(mask+1, 0)); |
} |
} |
} |
done: |
brw_fb_write(p, dw); |
} |
static void brw_wm_write__opacity(struct brw_compile *p, int dw, |
int src, int mask) |
{ |
int n; |
if (dw == 8 && p->gen >= 060) { |
brw_set_compression_control(p, BRW_COMPRESSION_NONE); |
brw_MUL(p, |
brw_message_reg(2), |
brw_vec8_grf(src+0, 0), |
brw_vec1_grf(mask, 3)); |
brw_MUL(p, |
brw_message_reg(3), |
brw_vec8_grf(src+1, 0), |
brw_vec1_grf(mask, 3)); |
brw_MUL(p, |
brw_message_reg(4), |
brw_vec8_grf(src+2, 0), |
brw_vec1_grf(mask, 3)); |
brw_MUL(p, |
brw_message_reg(5), |
brw_vec8_grf(src+3, 0), |
brw_vec1_grf(mask, 3)); |
goto done; |
} |
brw_set_compression_control(p, BRW_COMPRESSION_COMPRESSED); |
for (n = 0; n < 4; n++) { |
if (p->gen >= 060) { |
brw_MUL(p, |
brw_message_reg(2 + 2*n), |
brw_vec8_grf(src + 2*n, 0), |
brw_vec1_grf(mask, 3)); |
} else if (p->gen >= 045 && dw == 16) { |
brw_MUL(p, |
brw_message_reg(2 + n + BRW_MRF_COMPR4), |
brw_vec8_grf(src + 2*n, 0), |
brw_vec1_grf(mask, 3)); |
} else { |
brw_set_compression_control(p, BRW_COMPRESSION_NONE); |
brw_MUL(p, |
brw_message_reg(2 + n), |
brw_vec8_grf(src + 2*n, 0), |
brw_vec1_grf(mask, 3)); |
if (dw == 16) { |
brw_set_compression_control(p, BRW_COMPRESSION_2NDHALF); |
brw_MUL(p, |
brw_message_reg(2 + n + 4), |
brw_vec8_grf(src + 2*n+1, 0), |
brw_vec1_grf(mask, 3)); |
} |
} |
} |
done: |
brw_fb_write(p, dw); |
} |
static void brw_wm_write__mask_ca(struct brw_compile *p, int dw, |
int src, int mask) |
{ |
int n; |
if (dw == 8 && p->gen >= 060) { |
brw_set_compression_control(p, BRW_COMPRESSION_NONE); |
brw_MUL(p, |
brw_message_reg(2), |
brw_vec8_grf(src + 0, 0), |
brw_vec8_grf(mask + 0, 0)); |
brw_MUL(p, |
brw_message_reg(3), |
brw_vec8_grf(src + 1, 0), |
brw_vec8_grf(mask + 1, 0)); |
brw_MUL(p, |
brw_message_reg(4), |
brw_vec8_grf(src + 2, 0), |
brw_vec8_grf(mask + 2, 0)); |
brw_MUL(p, |
brw_message_reg(5), |
brw_vec8_grf(src + 3, 0), |
brw_vec8_grf(mask + 3, 0)); |
goto done; |
} |
brw_set_compression_control(p, BRW_COMPRESSION_COMPRESSED); |
for (n = 0; n < 4; n++) { |
if (p->gen >= 060) { |
brw_MUL(p, |
brw_message_reg(2 + 2*n), |
brw_vec8_grf(src + 2*n, 0), |
brw_vec8_grf(mask + 2*n, 0)); |
} else if (p->gen >= 045 && dw == 16) { |
brw_MUL(p, |
brw_message_reg(2 + n + BRW_MRF_COMPR4), |
brw_vec8_grf(src + 2*n, 0), |
brw_vec8_grf(mask + 2*n, 0)); |
} else { |
brw_set_compression_control(p, BRW_COMPRESSION_NONE); |
brw_MUL(p, |
brw_message_reg(2 + n), |
brw_vec8_grf(src + 2*n, 0), |
brw_vec8_grf(mask + 2*n, 0)); |
if (dw == 16) { |
brw_set_compression_control(p, BRW_COMPRESSION_2NDHALF); |
brw_MUL(p, |
brw_message_reg(2 + n + 4), |
brw_vec8_grf(src + 2*n + 1, 0), |
brw_vec8_grf(mask + 2*n + 1, 0)); |
} |
} |
} |
done: |
brw_fb_write(p, dw); |
} |
bool |
brw_wm_kernel__affine(struct brw_compile *p, int dispatch) |
{ |
if (p->gen < 060) |
brw_wm_xy(p, dispatch); |
brw_wm_write(p, dispatch, brw_wm_affine(p, dispatch, 0, 1, 12)); |
return true; |
} |
bool |
brw_wm_kernel__affine_mask(struct brw_compile *p, int dispatch) |
{ |
int src, mask; |
if (p->gen < 060) |
brw_wm_xy(p, dispatch); |
src = brw_wm_affine(p, dispatch, 0, 1, 12); |
mask = brw_wm_affine__alpha(p, dispatch, 1, 6, 20); |
brw_wm_write__mask(p, dispatch, src, mask); |
return true; |
} |
bool |
brw_wm_kernel__affine_mask_ca(struct brw_compile *p, int dispatch) |
{ |
int src, mask; |
if (p->gen < 060) |
brw_wm_xy(p, dispatch); |
src = brw_wm_affine(p, dispatch, 0, 1, 12); |
mask = brw_wm_affine(p, dispatch, 1, 6, 20); |
brw_wm_write__mask_ca(p, dispatch, src, mask); |
return true; |
} |
bool |
brw_wm_kernel__affine_mask_sa(struct brw_compile *p, int dispatch) |
{ |
int src, mask; |
if (p->gen < 060) |
brw_wm_xy(p, dispatch); |
src = brw_wm_affine__alpha(p, dispatch, 0, 1, 12); |
mask = brw_wm_affine(p, dispatch, 1, 6, 16); |
brw_wm_write__mask(p, dispatch, mask, src); |
return true; |
} |
/* Projective variants */ |
static void brw_wm_projective_st(struct brw_compile *p, int dw, |
int channel, int msg) |
{ |
int uv; |
if (dw == 16) { |
brw_set_compression_control(p, BRW_COMPRESSION_COMPRESSED); |
uv = p->gen >= 060 ? 6 : 3; |
} else { |
brw_set_compression_control(p, BRW_COMPRESSION_NONE); |
uv = p->gen >= 060 ? 4 : 3; |
} |
uv += 2*channel; |
msg++; |
if (p->gen >= 060) { |
/* First compute 1/z */ |
brw_PLN(p, |
brw_message_reg(msg), |
brw_vec1_grf(uv+1, 0), |
brw_vec8_grf(2, 0)); |
if (dw == 16) { |
brw_set_compression_control(p, BRW_COMPRESSION_NONE); |
brw_math_invert(p, brw_vec8_grf(30, 0), brw_vec8_grf(30, 0)); |
brw_math_invert(p, brw_vec8_grf(31, 0), brw_vec8_grf(31, 0)); |
brw_set_compression_control(p, BRW_COMPRESSION_COMPRESSED); |
} else |
brw_math_invert(p, brw_vec8_grf(30, 0), brw_vec8_grf(30, 0)); |
brw_PLN(p, |
brw_vec8_grf(28, 0), |
brw_vec1_grf(uv, 0), |
brw_vec8_grf(2, 0)); |
brw_MUL(p, |
brw_message_reg(msg), |
brw_vec8_grf(28, 0), |
brw_vec8_grf(30, 0)); |
msg += dw/8; |
brw_PLN(p, |
brw_vec8_grf(28, 0), |
brw_vec1_grf(uv, 0), |
brw_vec8_grf(4, 0)); |
brw_MUL(p, |
brw_message_reg(msg), |
brw_vec8_grf(28, 0), |
brw_vec8_grf(30, 0)); |
} else { |
struct brw_reg r = brw_vec1_grf(uv, 0); |
/* First compute 1/z */ |
brw_LINE(p, brw_null_reg(), brw_vec1_grf(uv+1, 0), brw_vec8_grf(X16, 0)); |
brw_MAC(p, brw_vec8_grf(30, 0), brw_vec1_grf(uv+1, 1), brw_vec8_grf(Y16, 0)); |
if (dw == 16) { |
brw_set_compression_control(p, BRW_COMPRESSION_NONE); |
brw_math_invert(p, brw_vec8_grf(30, 0), brw_vec8_grf(30, 0)); |
brw_math_invert(p, brw_vec8_grf(31, 0), brw_vec8_grf(31, 0)); |
brw_set_compression_control(p, BRW_COMPRESSION_COMPRESSED); |
} else |
brw_math_invert(p, brw_vec8_grf(30, 0), brw_vec8_grf(30, 0)); |
/* Now compute the output s,t values */ |
brw_LINE(p, brw_null_reg(), __suboffset(r, 0), brw_vec8_grf(X16, 0)); |
brw_MAC(p, brw_vec8_grf(28, 0), __suboffset(r, 1), brw_vec8_grf(Y16, 0)); |
brw_MUL(p, brw_message_reg(msg), brw_vec8_grf(28, 0), brw_vec8_grf(30, 0)); |
msg += dw/8; |
brw_LINE(p, brw_null_reg(), __suboffset(r, 4), brw_vec8_grf(X16, 0)); |
brw_MAC(p, brw_vec8_grf(28, 0), __suboffset(r, 5), brw_vec8_grf(Y16, 0)); |
brw_MUL(p, brw_message_reg(msg), brw_vec8_grf(28, 0), brw_vec8_grf(30, 0)); |
} |
} |
static int brw_wm_projective(struct brw_compile *p, int dw, |
int channel, int msg, int result) |
{ |
brw_wm_projective_st(p, dw, channel, msg); |
return brw_wm_sample(p, dw, channel, msg, result); |
} |
static int brw_wm_projective__alpha(struct brw_compile *p, int dw, |
int channel, int msg, int result) |
{ |
brw_wm_projective_st(p, dw, channel, msg); |
return brw_wm_sample__alpha(p, dw, channel, msg, result); |
} |
bool |
brw_wm_kernel__projective(struct brw_compile *p, int dispatch) |
{ |
if (p->gen < 060) |
brw_wm_xy(p, dispatch); |
brw_wm_write(p, dispatch, brw_wm_projective(p, dispatch, 0, 1, 12)); |
return true; |
} |
bool |
brw_wm_kernel__projective_mask(struct brw_compile *p, int dispatch) |
{ |
int src, mask; |
if (p->gen < 060) |
brw_wm_xy(p, dispatch); |
src = brw_wm_projective(p, dispatch, 0, 1, 12); |
mask = brw_wm_projective__alpha(p, dispatch, 1, 6, 20); |
brw_wm_write__mask(p, dispatch, src, mask); |
return true; |
} |
bool |
brw_wm_kernel__projective_mask_ca(struct brw_compile *p, int dispatch) |
{ |
int src, mask; |
if (p->gen < 060) |
brw_wm_xy(p, dispatch); |
src = brw_wm_projective(p, dispatch, 0, 1, 12); |
mask = brw_wm_projective(p, dispatch, 1, 6, 20); |
brw_wm_write__mask_ca(p, dispatch, src, mask); |
return true; |
} |
bool |
brw_wm_kernel__projective_mask_sa(struct brw_compile *p, int dispatch) |
{ |
int src, mask; |
if (p->gen < 060) |
brw_wm_xy(p, dispatch); |
src = brw_wm_projective__alpha(p, dispatch, 0, 1, 12); |
mask = brw_wm_projective(p, dispatch, 1, 6, 16); |
brw_wm_write__mask(p, dispatch, mask, src); |
return true; |
} |
bool |
brw_wm_kernel__affine_opacity(struct brw_compile *p, int dispatch) |
{ |
int src, mask; |
if (p->gen < 060) { |
brw_wm_xy(p, dispatch); |
mask = 5; |
} else |
mask = dispatch == 16 ? 8 : 6; |
src = brw_wm_affine(p, dispatch, 0, 1, 12); |
brw_wm_write__opacity(p, dispatch, src, mask); |
return true; |
} |
bool |
brw_wm_kernel__projective_opacity(struct brw_compile *p, int dispatch) |
{ |
int src, mask; |
if (p->gen < 060) { |
brw_wm_xy(p, dispatch); |
mask = 5; |
} else |
mask = dispatch == 16 ? 8 : 6; |
src = brw_wm_projective(p, dispatch, 0, 1, 12); |
brw_wm_write__opacity(p, dispatch, src, mask); |
return true; |
} |
/drivers/video/Intel-2D/sna/compiler.h |
---|
0,0 → 1,92 |
/* |
* Copyright (c) 2011 Intel Corporation |
* |
* Permission is hereby granted, free of charge, to any person obtaining a |
* copy of this software and associated documentation files (the "Software"), |
* to deal in the Software without restriction, including without limitation |
* the rights to use, copy, modify, merge, publish, distribute, sublicense, |
* and/or sell copies of the Software, and to permit persons to whom the |
* Software is furnished to do so, subject to the following conditions: |
* |
* The above copyright notice and this permission notice (including the next |
* paragraph) shall be included in all copies or substantial portions of the |
* Software. |
* |
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL |
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, |
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE |
* SOFTWARE. |
* |
* Authors: |
* Chris Wilson <chris@chris-wilson.co.uk> |
* |
*/ |
#ifndef _SNA_COMPILER_H_ |
#define _SNA_COMPILER_H_ |
#if defined(__GNUC__) && (__GNUC__ > 2) && defined(__OPTIMIZE__) |
#define likely(expr) (__builtin_expect (!!(expr), 1)) |
#define unlikely(expr) (__builtin_expect (!!(expr), 0)) |
#define noinline __attribute__((noinline)) |
#define force_inline inline __attribute__((always_inline)) |
#define fastcall __attribute__((regparm(3))) |
#define must_check __attribute__((warn_unused_result)) |
#define constant __attribute__((const)) |
#define pure __attribute__((pure)) |
#define tightly_packed __attribute__((__packed__)) |
#define flatten __attribute__((flatten)) |
#define page_aligned __attribute__((aligned(4096))) |
#else |
#define likely(expr) (expr) |
#define unlikely(expr) (expr) |
#define noinline |
#define force_inline inline |
#define fastcall |
#define must_check |
#define constant |
#define pure |
#define tighly_packed |
#define flatten |
#define page_aligned |
#endif |
#define HAS_GCC(major, minor) defined(__GNUC__) && (__GNUC__ > (major) || __GNUC__ == (major) && __GNUC_MINOR__ >= (minor)) |
#if HAS_GCC(4, 5) |
#define sse2 __attribute__((target("sse2,fpmath=sse"))) |
#define sse4_2 __attribute__((target("sse4.2,sse2,fpmath=sse"))) |
#endif |
#if HAS_GCC(4, 7) |
#define avx2 __attribute__((target("avx2,sse4.2,sse2,fpmath=sse"))) |
#endif |
#if HAS_GCC(4, 6) && defined(__OPTIMIZE__) |
#define fast __attribute__((optimize("Ofast"))) |
#else |
#define fast |
#endif |
#if HAS_GCC(4, 6) && defined(__OPTIMIZE__) |
#define fast_memcpy __attribute__((optimize("Ofast"))) __attribute__((target("inline-all-stringops"))) |
#elif HAS_GCC(4, 5) && defined(__OPTIMIZE__) |
#define fast_memcpy __attribute__((target("inline-all-stringops"))) |
#else |
#define fast_memcpy |
#endif |
#ifdef HAVE_VALGRIND |
#define VG(x) x |
#else |
#define VG(x) |
#endif |
#define VG_CLEAR(s) VG(memset(&s, 0, sizeof(s))) |
#define COMPILE_TIME_ASSERT(E) ((void)sizeof(char[1 - 2*!(E)])) |
#endif /* _SNA_COMPILER_H_ */ |
/drivers/video/Intel-2D/sna/gen3_render.c |
---|
0,0 → 1,2814 |
/* |
* Copyright © 2010-2011 Intel Corporation |
* |
* Permission is hereby granted, free of charge, to any person obtaining a |
* copy of this software and associated documentation files (the "Software"), |
* to deal in the Software without restriction, including without limitation |
* the rights to use, copy, modify, merge, publish, distribute, sublicense, |
* and/or sell copies of the Software, and to permit persons to whom the |
* Software is furnished to do so, subject to the following conditions: |
* |
* The above copyright notice and this permission notice (including the next |
* paragraph) shall be included in all copies or substantial portions of the |
* Software. |
* |
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL |
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, |
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE |
* SOFTWARE. |
* |
* Authors: |
* Chris Wilson <chris@chris-wilson.co.uk> |
* |
*/ |
#ifdef HAVE_CONFIG_H |
#include "config.h" |
#endif |
#include "sna.h" |
#include "sna_render.h" |
#include "sna_render_inline.h" |
#include "sna_reg.h" |
//#include "sna_video.h" |
#include "gen3_render.h" |
#define NO_COMPOSITE 0 |
#define NO_COMPOSITE_SPANS 0 |
#define NO_COPY 0 |
#define NO_COPY_BOXES 0 |
#define NO_FILL 0 |
#define NO_FILL_ONE 0 |
#define NO_FILL_BOXES 0 |
#define PREFER_BLT_FILL 1 |
enum { |
SHADER_NONE = 0, |
SHADER_ZERO, |
SHADER_BLACK, |
SHADER_WHITE, |
SHADER_CONSTANT, |
SHADER_LINEAR, |
SHADER_RADIAL, |
SHADER_TEXTURE, |
SHADER_OPACITY, |
}; |
#define MAX_3D_SIZE 2048 |
#define MAX_3D_PITCH 8192 |
#define OUT_BATCH(v) batch_emit(sna, v) |
#define OUT_BATCH_F(v) batch_emit_float(sna, v) |
#define OUT_VERTEX(v) vertex_emit(sna, v) |
enum gen3_radial_mode { |
RADIAL_ONE, |
RADIAL_TWO |
}; |
static const struct blendinfo { |
bool dst_alpha; |
bool src_alpha; |
uint32_t src_blend; |
uint32_t dst_blend; |
} gen3_blend_op[] = { |
/* Clear */ {0, 0, BLENDFACT_ZERO, BLENDFACT_ZERO}, |
/* Src */ {0, 0, BLENDFACT_ONE, BLENDFACT_ZERO}, |
/* Dst */ {0, 0, BLENDFACT_ZERO, BLENDFACT_ONE}, |
/* Over */ {0, 1, BLENDFACT_ONE, BLENDFACT_INV_SRC_ALPHA}, |
/* OverReverse */ {1, 0, BLENDFACT_INV_DST_ALPHA, BLENDFACT_ONE}, |
/* In */ {1, 0, BLENDFACT_DST_ALPHA, BLENDFACT_ZERO}, |
/* InReverse */ {0, 1, BLENDFACT_ZERO, BLENDFACT_SRC_ALPHA}, |
/* Out */ {1, 0, BLENDFACT_INV_DST_ALPHA, BLENDFACT_ZERO}, |
/* OutReverse */ {0, 1, BLENDFACT_ZERO, BLENDFACT_INV_SRC_ALPHA}, |
/* Atop */ {1, 1, BLENDFACT_DST_ALPHA, BLENDFACT_INV_SRC_ALPHA}, |
/* AtopReverse */ {1, 1, BLENDFACT_INV_DST_ALPHA, BLENDFACT_SRC_ALPHA}, |
/* Xor */ {1, 1, BLENDFACT_INV_DST_ALPHA, BLENDFACT_INV_SRC_ALPHA}, |
/* Add */ {0, 0, BLENDFACT_ONE, BLENDFACT_ONE}, |
}; |
#define S6_COLOR_WRITE_ONLY \ |
(S6_COLOR_WRITE_ENABLE | \ |
BLENDFUNC_ADD << S6_CBUF_BLEND_FUNC_SHIFT | \ |
BLENDFACT_ONE << S6_CBUF_SRC_BLEND_FACT_SHIFT | \ |
BLENDFACT_ZERO << S6_CBUF_DST_BLEND_FACT_SHIFT) |
static const struct formatinfo { |
unsigned int fmt, xfmt; |
uint32_t card_fmt; |
bool rb_reversed; |
} gen3_tex_formats[] = { |
{PICT_a8, 0, MAPSURF_8BIT | MT_8BIT_A8, false}, |
{PICT_a8r8g8b8, 0, MAPSURF_32BIT | MT_32BIT_ARGB8888, false}, |
{PICT_x8r8g8b8, 0, MAPSURF_32BIT | MT_32BIT_XRGB8888, false}, |
{PICT_a8b8g8r8, 0, MAPSURF_32BIT | MT_32BIT_ABGR8888, false}, |
{PICT_x8b8g8r8, 0, MAPSURF_32BIT | MT_32BIT_XBGR8888, false}, |
{PICT_a2r10g10b10, PICT_x2r10g10b10, MAPSURF_32BIT | MT_32BIT_ARGB2101010, false}, |
{PICT_a2b10g10r10, PICT_x2b10g10r10, MAPSURF_32BIT | MT_32BIT_ABGR2101010, false}, |
{PICT_r5g6b5, 0, MAPSURF_16BIT | MT_16BIT_RGB565, false}, |
{PICT_b5g6r5, 0, MAPSURF_16BIT | MT_16BIT_RGB565, true}, |
{PICT_a1r5g5b5, PICT_x1r5g5b5, MAPSURF_16BIT | MT_16BIT_ARGB1555, false}, |
{PICT_a1b5g5r5, PICT_x1b5g5r5, MAPSURF_16BIT | MT_16BIT_ARGB1555, true}, |
{PICT_a4r4g4b4, PICT_x4r4g4b4, MAPSURF_16BIT | MT_16BIT_ARGB4444, false}, |
{PICT_a4b4g4r4, PICT_x4b4g4r4, MAPSURF_16BIT | MT_16BIT_ARGB4444, true}, |
}; |
#define xFixedToDouble(f) pixman_fixed_to_double(f) |
static inline bool too_large(int width, int height) |
{ |
return width > MAX_3D_SIZE || height > MAX_3D_SIZE; |
} |
static inline uint32_t gen3_buf_tiling(uint32_t tiling) |
{ |
uint32_t v = 0; |
switch (tiling) { |
case I915_TILING_Y: v |= BUF_3D_TILE_WALK_Y; |
case I915_TILING_X: v |= BUF_3D_TILED_SURFACE; |
case I915_TILING_NONE: break; |
} |
return v; |
} |
static uint32_t gen3_get_blend_cntl(int op, |
bool has_component_alpha, |
uint32_t dst_format) |
{ |
uint32_t sblend; |
uint32_t dblend; |
sblend = BLENDFACT_ONE; |
dblend = BLENDFACT_INV_SRC_ALPHA; |
#if 0 |
if (op <= PictOpSrc) /* for clear and src disable blending */ |
return S6_COLOR_WRITE_ONLY; |
/* If there's no dst alpha channel, adjust the blend op so that we'll |
* treat it as always 1. |
*/ |
if (gen3_blend_op[op].dst_alpha) { |
if (PICT_FORMAT_A(dst_format) == 0) { |
if (sblend == BLENDFACT_DST_ALPHA) |
sblend = BLENDFACT_ONE; |
else if (sblend == BLENDFACT_INV_DST_ALPHA) |
sblend = BLENDFACT_ZERO; |
} |
/* gen3 engine reads 8bit color buffer into green channel |
* in cases like color buffer blending etc., and also writes |
* back green channel. So with dst_alpha blend we should use |
* color factor. See spec on "8-bit rendering". |
*/ |
if (dst_format == PICT_a8) { |
if (sblend == BLENDFACT_DST_ALPHA) |
sblend = BLENDFACT_DST_COLR; |
else if (sblend == BLENDFACT_INV_DST_ALPHA) |
sblend = BLENDFACT_INV_DST_COLR; |
} |
} |
/* If the source alpha is being used, then we should only be in a case |
* where the source blend factor is 0, and the source blend value is the |
* mask channels multiplied by the source picture's alpha. |
*/ |
if (has_component_alpha && gen3_blend_op[op].src_alpha) { |
if (dblend == BLENDFACT_SRC_ALPHA) |
dblend = BLENDFACT_SRC_COLR; |
else if (dblend == BLENDFACT_INV_SRC_ALPHA) |
dblend = BLENDFACT_INV_SRC_COLR; |
} |
#endif |
return (S6_CBUF_BLEND_ENABLE | S6_COLOR_WRITE_ENABLE | |
BLENDFUNC_ADD << S6_CBUF_BLEND_FUNC_SHIFT | |
sblend << S6_CBUF_SRC_BLEND_FACT_SHIFT | |
dblend << S6_CBUF_DST_BLEND_FACT_SHIFT); |
} |
static bool gen3_dst_rb_reversed(uint32_t format) |
{ |
switch (format) { |
case PICT_a8r8g8b8: |
case PICT_x8r8g8b8: |
case PICT_r5g6b5: |
case PICT_a1r5g5b5: |
case PICT_x1r5g5b5: |
case PICT_a2r10g10b10: |
case PICT_x2r10g10b10: |
case PICT_a8: |
case PICT_a4r4g4b4: |
case PICT_x4r4g4b4: |
return false; |
default: |
return true; |
} |
} |
#define DSTORG_HORT_BIAS(x) ((x)<<20) |
#define DSTORG_VERT_BIAS(x) ((x)<<16) |
static uint32_t gen3_get_dst_format(uint32_t format) |
{ |
#define BIAS (DSTORG_HORT_BIAS(0x8) | DSTORG_VERT_BIAS(0x8)) |
switch (format) { |
default: |
case PICT_a8r8g8b8: |
case PICT_x8r8g8b8: |
case PICT_a8b8g8r8: |
case PICT_x8b8g8r8: |
return BIAS | COLR_BUF_ARGB8888; |
case PICT_r5g6b5: |
case PICT_b5g6r5: |
return BIAS | COLR_BUF_RGB565; |
case PICT_a1r5g5b5: |
case PICT_x1r5g5b5: |
case PICT_a1b5g5r5: |
case PICT_x1b5g5r5: |
return BIAS | COLR_BUF_ARGB1555; |
case PICT_a2r10g10b10: |
case PICT_x2r10g10b10: |
case PICT_a2b10g10r10: |
case PICT_x2b10g10r10: |
return BIAS | COLR_BUF_ARGB2AAA; |
case PICT_a8: |
return BIAS | COLR_BUF_8BIT; |
case PICT_a4r4g4b4: |
case PICT_x4r4g4b4: |
case PICT_a4b4g4r4: |
case PICT_x4b4g4r4: |
return BIAS | COLR_BUF_ARGB4444; |
} |
#undef BIAS |
} |
#if 0 |
static bool gen3_check_repeat(PicturePtr p) |
{ |
if (!p->repeat) |
return true; |
switch (p->repeatType) { |
case RepeatNone: |
case RepeatNormal: |
case RepeatPad: |
case RepeatReflect: |
return true; |
default: |
return false; |
} |
} |
static uint32_t gen3_filter(uint32_t filter) |
{ |
switch (filter) { |
default: |
assert(0); |
case PictFilterNearest: |
return (FILTER_NEAREST << SS2_MAG_FILTER_SHIFT | |
FILTER_NEAREST << SS2_MIN_FILTER_SHIFT | |
MIPFILTER_NONE << SS2_MIP_FILTER_SHIFT); |
case PictFilterBilinear: |
return (FILTER_LINEAR << SS2_MAG_FILTER_SHIFT | |
FILTER_LINEAR << SS2_MIN_FILTER_SHIFT | |
MIPFILTER_NONE << SS2_MIP_FILTER_SHIFT); |
} |
} |
static bool gen3_check_filter(PicturePtr p) |
{ |
switch (p->filter) { |
case PictFilterNearest: |
case PictFilterBilinear: |
return true; |
default: |
return false; |
} |
} |
fastcall static void |
gen3_emit_composite_primitive_identity_gradient(struct sna *sna, |
const struct sna_composite_op *op, |
const struct sna_composite_rectangles *r) |
{ |
int16_t dst_x, dst_y; |
int16_t src_x, src_y; |
dst_x = r->dst.x + op->dst.x; |
dst_y = r->dst.y + op->dst.y; |
src_x = r->src.x + op->src.offset[0]; |
src_y = r->src.y + op->src.offset[1]; |
gen3_emit_composite_dstcoord(sna, dst_x + r->width, dst_y + r->height); |
OUT_VERTEX(src_x + r->width); |
OUT_VERTEX(src_y + r->height); |
gen3_emit_composite_dstcoord(sna, dst_x, dst_y + r->height); |
OUT_VERTEX(src_x); |
OUT_VERTEX(src_y + r->height); |
gen3_emit_composite_dstcoord(sna, dst_x, dst_y); |
OUT_VERTEX(src_x); |
OUT_VERTEX(src_y); |
} |
fastcall static void |
gen3_emit_composite_boxes_identity_gradient(const struct sna_composite_op *op, |
const BoxRec *box, int nbox, |
float *v) |
{ |
do { |
v[0] = box->x2; |
v[1] = box->y2; |
v[2] = box->x2 + op->src.offset[0]; |
v[3] = box->y2 + op->src.offset[1]; |
v[4] = box->x1; |
v[5] = box->y2; |
v[6] = box->x1 + op->src.offset[0]; |
v[7] = box->y2 + op->src.offset[1]; |
v[8] = box->x1; |
v[9] = box->y1; |
v[10] = box->x1 + op->src.offset[0]; |
v[11] = box->y1 + op->src.offset[1]; |
v += 12; |
box++; |
} while (--nbox); |
} |
fastcall static void |
gen3_emit_composite_boxes_affine_gradient(const struct sna_composite_op *op, |
const BoxRec *box, int nbox, |
float *v) |
{ |
const PictTransform *transform = op->src.transform; |
do { |
v[0] = box->x2; |
v[1] = box->y2; |
_sna_get_transformed_scaled(box->x2 + op->src.offset[0], |
box->y2 + op->src.offset[1], |
transform, op->src.scale, |
&v[2], &v[3]); |
v[4] = box->x1; |
v[5] = box->y2; |
_sna_get_transformed_scaled(box->x1 + op->src.offset[0], |
box->y2 + op->src.offset[1], |
transform, op->src.scale, |
&v[6], &v[7]); |
v[8] = box->x1; |
v[9] = box->y1; |
_sna_get_transformed_scaled(box->x1 + op->src.offset[0], |
box->y1 + op->src.offset[1], |
transform, op->src.scale, |
&v[10], &v[11]); |
box++; |
v += 12; |
} while (--nbox); |
} |
fastcall static void |
gen3_emit_composite_primitive_identity_source(struct sna *sna, |
const struct sna_composite_op *op, |
const struct sna_composite_rectangles *r) |
{ |
float w = r->width; |
float h = r->height; |
float *v; |
v = sna->render.vertices + sna->render.vertex_used; |
sna->render.vertex_used += 12; |
v[8] = v[4] = r->dst.x + op->dst.x; |
v[0] = v[4] + w; |
v[9] = r->dst.y + op->dst.y; |
v[5] = v[1] = v[9] + h; |
v[10] = v[6] = (r->src.x + op->src.offset[0]) * op->src.scale[0]; |
v[2] = v[6] + w * op->src.scale[0]; |
v[11] = (r->src.y + op->src.offset[1]) * op->src.scale[1]; |
v[7] = v[3] = v[11] + h * op->src.scale[1]; |
} |
fastcall static void |
gen3_emit_composite_boxes_identity_source(const struct sna_composite_op *op, |
const BoxRec *box, int nbox, |
float *v) |
{ |
do { |
v[0] = box->x2 + op->dst.x; |
v[8] = v[4] = box->x1 + op->dst.x; |
v[5] = v[1] = box->y2 + op->dst.y; |
v[9] = box->y1 + op->dst.y; |
v[10] = v[6] = (box->x1 + op->src.offset[0]) * op->src.scale[0]; |
v[2] = (box->x2 + op->src.offset[0]) * op->src.scale[0]; |
v[11] = (box->y1 + op->src.offset[1]) * op->src.scale[1]; |
v[7] = v[3] = (box->y2 + op->src.offset[1]) * op->src.scale[1]; |
v += 12; |
box++; |
} while (--nbox); |
} |
fastcall static void |
gen3_emit_composite_primitive_identity_source_no_offset(struct sna *sna, |
const struct sna_composite_op *op, |
const struct sna_composite_rectangles *r) |
{ |
float w = r->width; |
float h = r->height; |
float *v; |
v = sna->render.vertices + sna->render.vertex_used; |
sna->render.vertex_used += 12; |
v[8] = v[4] = r->dst.x; |
v[9] = r->dst.y; |
v[0] = v[4] + w; |
v[5] = v[1] = v[9] + h; |
v[10] = v[6] = r->src.x * op->src.scale[0]; |
v[11] = r->src.y * op->src.scale[1]; |
v[2] = v[6] + w * op->src.scale[0]; |
v[7] = v[3] = v[11] + h * op->src.scale[1]; |
} |
fastcall static void |
gen3_emit_composite_primitive_constant_identity_mask(struct sna *sna, |
const struct sna_composite_op *op, |
const struct sna_composite_rectangles *r) |
{ |
float w = r->width; |
float h = r->height; |
float *v; |
v = sna->render.vertices + sna->render.vertex_used; |
sna->render.vertex_used += 12; |
v[8] = v[4] = r->dst.x + op->dst.x; |
v[0] = v[4] + w; |
v[9] = r->dst.y + op->dst.y; |
v[5] = v[1] = v[9] + h; |
v[10] = v[6] = (r->mask.x + op->mask.offset[0]) * op->mask.scale[0]; |
v[2] = v[6] + w * op->mask.scale[0]; |
v[11] = (r->mask.y + op->mask.offset[1]) * op->mask.scale[1]; |
v[7] = v[3] = v[11] + h * op->mask.scale[1]; |
} |
#endif |
fastcall static void |
gen3_emit_composite_primitive_identity_source_mask(struct sna *sna, |
const struct sna_composite_op *op, |
const struct sna_composite_rectangles *r) |
{ |
float dst_x, dst_y; |
float src_x, src_y; |
float msk_x, msk_y; |
float w, h; |
float *v; |
dst_x = r->dst.x + op->dst.x; |
dst_y = r->dst.y + op->dst.y; |
src_x = r->src.x + op->src.offset[0]; |
src_y = r->src.y + op->src.offset[1]; |
msk_x = r->mask.x + op->mask.offset[0]; |
msk_y = r->mask.y + op->mask.offset[1]; |
w = r->width; |
h = r->height; |
v = sna->render.vertices + sna->render.vertex_used; |
sna->render.vertex_used += 18; |
v[0] = dst_x + w; |
v[1] = dst_y + h; |
v[2] = (src_x + w) * op->src.scale[0]; |
v[3] = (src_y + h) * op->src.scale[1]; |
v[4] = (msk_x + w) * op->mask.scale[0]; |
v[5] = (msk_y + h) * op->mask.scale[1]; |
v[6] = dst_x; |
v[7] = v[1]; |
v[8] = src_x * op->src.scale[0]; |
v[9] = v[3]; |
v[10] = msk_x * op->mask.scale[0]; |
v[11] =v[5]; |
v[12] = v[6]; |
v[13] = dst_y; |
v[14] = v[8]; |
v[15] = src_y * op->src.scale[1]; |
v[16] = v[10]; |
v[17] = msk_y * op->mask.scale[1]; |
} |
static inline void |
gen3_2d_perspective(struct sna *sna, int in, int out) |
{ |
gen3_fs_rcp(out, 0, gen3_fs_operand(in, W, W, W, W)); |
gen3_fs_mul(out, |
gen3_fs_operand(in, X, Y, ZERO, ONE), |
gen3_fs_operand_reg(out)); |
} |
static inline void |
gen3_linear_coord(struct sna *sna, |
const struct sna_composite_channel *channel, |
int in, int out) |
{ |
int c = channel->u.gen3.constants; |
if (!channel->is_affine) { |
gen3_2d_perspective(sna, in, FS_U0); |
in = FS_U0; |
} |
gen3_fs_mov(out, gen3_fs_operand_zero()); |
gen3_fs_dp3(out, MASK_X, |
gen3_fs_operand(in, X, Y, ONE, ZERO), |
gen3_fs_operand_reg(c)); |
} |
static void |
gen3_radial_coord(struct sna *sna, |
const struct sna_composite_channel *channel, |
int in, int out) |
{ |
int c = channel->u.gen3.constants; |
if (!channel->is_affine) { |
gen3_2d_perspective(sna, in, FS_U0); |
in = FS_U0; |
} |
switch (channel->u.gen3.mode) { |
case RADIAL_ONE: |
/* |
pdx = (x - c1x) / dr, pdy = (y - c1y) / dr; |
r? = pdx*pdx + pdy*pdy |
t = r?/sqrt(r?) - r1/dr; |
*/ |
gen3_fs_mad(FS_U0, MASK_X | MASK_Y, |
gen3_fs_operand(in, X, Y, ZERO, ZERO), |
gen3_fs_operand(c, Z, Z, ZERO, ZERO), |
gen3_fs_operand(c, NEG_X, NEG_Y, ZERO, ZERO)); |
gen3_fs_dp2add(FS_U0, MASK_X, |
gen3_fs_operand(FS_U0, X, Y, ZERO, ZERO), |
gen3_fs_operand(FS_U0, X, Y, ZERO, ZERO), |
gen3_fs_operand_zero()); |
gen3_fs_rsq(out, MASK_X, gen3_fs_operand(FS_U0, X, X, X, X)); |
gen3_fs_mad(out, 0, |
gen3_fs_operand(FS_U0, X, ZERO, ZERO, ZERO), |
gen3_fs_operand(out, X, ZERO, ZERO, ZERO), |
gen3_fs_operand(c, W, ZERO, ZERO, ZERO)); |
break; |
case RADIAL_TWO: |
/* |
pdx = x - c1x, pdy = y - c1y; |
A = dx? + dy? - dr? |
B = -2*(pdx*dx + pdy*dy + r1*dr); |
C = pdx? + pdy? - r1?; |
det = B*B - 4*A*C; |
t = (-B + sqrt (det)) / (2 * A) |
*/ |
/* u0.x = pdx, u0.y = pdy, u[0].z = r1; */ |
gen3_fs_add(FS_U0, |
gen3_fs_operand(in, X, Y, ZERO, ZERO), |
gen3_fs_operand(c, X, Y, Z, ZERO)); |
/* u0.x = pdx, u0.y = pdy, u[0].z = r1, u[0].w = B; */ |
gen3_fs_dp3(FS_U0, MASK_W, |
gen3_fs_operand(FS_U0, X, Y, ONE, ZERO), |
gen3_fs_operand(c+1, X, Y, Z, ZERO)); |
/* u1.x = pdx? + pdy? - r1?; [C] */ |
gen3_fs_dp3(FS_U1, MASK_X, |
gen3_fs_operand(FS_U0, X, Y, Z, ZERO), |
gen3_fs_operand(FS_U0, X, Y, NEG_Z, ZERO)); |
/* u1.x = C, u1.y = B, u1.z=-4*A; */ |
gen3_fs_mov_masked(FS_U1, MASK_Y, gen3_fs_operand(FS_U0, W, W, W, W)); |
gen3_fs_mov_masked(FS_U1, MASK_Z, gen3_fs_operand(c, W, W, W, W)); |
/* u1.x = B? - 4*A*C */ |
gen3_fs_dp2add(FS_U1, MASK_X, |
gen3_fs_operand(FS_U1, X, Y, ZERO, ZERO), |
gen3_fs_operand(FS_U1, Z, Y, ZERO, ZERO), |
gen3_fs_operand_zero()); |
/* out.x = -B + sqrt (B? - 4*A*C), */ |
gen3_fs_rsq(out, MASK_X, gen3_fs_operand(FS_U1, X, X, X, X)); |
gen3_fs_mad(out, MASK_X, |
gen3_fs_operand(out, X, ZERO, ZERO, ZERO), |
gen3_fs_operand(FS_U1, X, ZERO, ZERO, ZERO), |
gen3_fs_operand(FS_U0, NEG_W, ZERO, ZERO, ZERO)); |
/* out.x = (-B + sqrt (B? - 4*A*C)) / (2 * A), */ |
gen3_fs_mul(out, |
gen3_fs_operand(out, X, ZERO, ZERO, ZERO), |
gen3_fs_operand(c+1, W, ZERO, ZERO, ZERO)); |
break; |
} |
} |
static void |
gen3_composite_emit_shader(struct sna *sna, |
const struct sna_composite_op *op, |
uint8_t blend) |
{ |
bool dst_is_alpha = PIXMAN_FORMAT_RGB(op->dst.format) == 0; |
const struct sna_composite_channel *src, *mask; |
struct gen3_render_state *state = &sna->render_state.gen3; |
uint32_t shader_offset, id; |
int src_reg, mask_reg; |
int t, length; |
src = &op->src; |
mask = &op->mask; |
if (mask->u.gen3.type == SHADER_NONE) |
mask = NULL; |
id = (src->u.gen3.type | |
src->is_affine << 4 | |
src->alpha_fixup << 5 | |
src->rb_reversed << 6); |
if (mask) { |
id |= (mask->u.gen3.type << 8 | |
mask->is_affine << 12 | |
gen3_blend_op[blend].src_alpha << 13 | |
op->has_component_alpha << 14 | |
mask->alpha_fixup << 15 | |
mask->rb_reversed << 16); |
} |
id |= dst_is_alpha << 24; |
id |= op->rb_reversed << 25; |
if (id == state->last_shader) |
return; |
state->last_shader = id; |
shader_offset = sna->kgem.nbatch++; |
t = 0; |
switch (src->u.gen3.type) { |
case SHADER_NONE: |
case SHADER_OPACITY: |
assert(0); |
case SHADER_ZERO: |
case SHADER_BLACK: |
case SHADER_WHITE: |
break; |
case SHADER_CONSTANT: |
gen3_fs_dcl(FS_T8); |
src_reg = FS_T8; |
break; |
case SHADER_TEXTURE: |
case SHADER_RADIAL: |
case SHADER_LINEAR: |
gen3_fs_dcl(FS_S0); |
gen3_fs_dcl(FS_T0); |
t++; |
break; |
} |
if (mask == NULL) { |
switch (src->u.gen3.type) { |
case SHADER_ZERO: |
gen3_fs_mov(FS_OC, gen3_fs_operand_zero()); |
goto done; |
case SHADER_BLACK: |
if (dst_is_alpha) |
gen3_fs_mov(FS_OC, gen3_fs_operand_one()); |
else |
gen3_fs_mov(FS_OC, gen3_fs_operand(FS_R0, ZERO, ZERO, ZERO, ONE)); |
goto done; |
case SHADER_WHITE: |
gen3_fs_mov(FS_OC, gen3_fs_operand_one()); |
goto done; |
} |
if (src->alpha_fixup && dst_is_alpha) { |
gen3_fs_mov(FS_OC, gen3_fs_operand_one()); |
goto done; |
} |
/* No mask, so load directly to output color */ |
if (src->u.gen3.type != SHADER_CONSTANT) { |
if (dst_is_alpha || src->rb_reversed ^ op->rb_reversed) |
src_reg = FS_R0; |
else |
src_reg = FS_OC; |
} |
switch (src->u.gen3.type) { |
case SHADER_LINEAR: |
gen3_linear_coord(sna, src, FS_T0, FS_R0); |
gen3_fs_texld(src_reg, FS_S0, FS_R0); |
break; |
case SHADER_RADIAL: |
gen3_radial_coord(sna, src, FS_T0, FS_R0); |
gen3_fs_texld(src_reg, FS_S0, FS_R0); |
break; |
case SHADER_TEXTURE: |
if (src->is_affine) |
gen3_fs_texld(src_reg, FS_S0, FS_T0); |
else |
gen3_fs_texldp(src_reg, FS_S0, FS_T0); |
break; |
case SHADER_NONE: |
case SHADER_WHITE: |
case SHADER_BLACK: |
case SHADER_ZERO: |
assert(0); |
case SHADER_CONSTANT: |
break; |
} |
if (src_reg != FS_OC) { |
if (src->alpha_fixup) |
gen3_fs_mov(FS_OC, |
src->rb_reversed ^ op->rb_reversed ? |
gen3_fs_operand(src_reg, Z, Y, X, ONE) : |
gen3_fs_operand(src_reg, X, Y, Z, ONE)); |
else if (dst_is_alpha) |
gen3_fs_mov(FS_OC, gen3_fs_operand(src_reg, W, W, W, W)); |
else if (src->rb_reversed ^ op->rb_reversed) |
gen3_fs_mov(FS_OC, gen3_fs_operand(src_reg, Z, Y, X, W)); |
else |
gen3_fs_mov(FS_OC, gen3_fs_operand_reg(src_reg)); |
} else if (src->alpha_fixup) |
gen3_fs_mov_masked(FS_OC, MASK_W, gen3_fs_operand_one()); |
} else { |
int out_reg = FS_OC; |
if (op->rb_reversed) |
out_reg = FS_U0; |
switch (mask->u.gen3.type) { |
case SHADER_CONSTANT: |
gen3_fs_dcl(FS_T9); |
mask_reg = FS_T9; |
break; |
case SHADER_TEXTURE: |
case SHADER_LINEAR: |
case SHADER_RADIAL: |
gen3_fs_dcl(FS_S0 + t); |
/* fall through */ |
case SHADER_OPACITY: |
gen3_fs_dcl(FS_T0 + t); |
break; |
case SHADER_ZERO: |
case SHADER_BLACK: |
assert(0); |
case SHADER_NONE: |
case SHADER_WHITE: |
break; |
} |
t = 0; |
switch (src->u.gen3.type) { |
case SHADER_LINEAR: |
gen3_linear_coord(sna, src, FS_T0, FS_R0); |
gen3_fs_texld(FS_R0, FS_S0, FS_R0); |
src_reg = FS_R0; |
t++; |
break; |
case SHADER_RADIAL: |
gen3_radial_coord(sna, src, FS_T0, FS_R0); |
gen3_fs_texld(FS_R0, FS_S0, FS_R0); |
src_reg = FS_R0; |
t++; |
break; |
case SHADER_TEXTURE: |
if (src->is_affine) |
gen3_fs_texld(FS_R0, FS_S0, FS_T0); |
else |
gen3_fs_texldp(FS_R0, FS_S0, FS_T0); |
src_reg = FS_R0; |
t++; |
break; |
case SHADER_CONSTANT: |
case SHADER_NONE: |
case SHADER_ZERO: |
case SHADER_BLACK: |
case SHADER_WHITE: |
break; |
} |
if (src->alpha_fixup) |
gen3_fs_mov_masked(src_reg, MASK_W, gen3_fs_operand_one()); |
if (src->rb_reversed) |
gen3_fs_mov(src_reg, gen3_fs_operand(src_reg, Z, Y, X, W)); |
switch (mask->u.gen3.type) { |
case SHADER_LINEAR: |
gen3_linear_coord(sna, mask, FS_T0 + t, FS_R1); |
gen3_fs_texld(FS_R1, FS_S0 + t, FS_R1); |
mask_reg = FS_R1; |
break; |
case SHADER_RADIAL: |
gen3_radial_coord(sna, mask, FS_T0 + t, FS_R1); |
gen3_fs_texld(FS_R1, FS_S0 + t, FS_R1); |
mask_reg = FS_R1; |
break; |
case SHADER_TEXTURE: |
if (mask->is_affine) |
gen3_fs_texld(FS_R1, FS_S0 + t, FS_T0 + t); |
else |
gen3_fs_texldp(FS_R1, FS_S0 + t, FS_T0 + t); |
mask_reg = FS_R1; |
break; |
case SHADER_OPACITY: |
switch (src->u.gen3.type) { |
case SHADER_BLACK: |
case SHADER_WHITE: |
if (dst_is_alpha || src->u.gen3.type == SHADER_WHITE) { |
gen3_fs_mov(out_reg, |
gen3_fs_operand(FS_T0 + t, X, X, X, X)); |
} else { |
gen3_fs_mov(out_reg, |
gen3_fs_operand(FS_T0 + t, ZERO, ZERO, ZERO, X)); |
} |
break; |
default: |
if (dst_is_alpha) { |
gen3_fs_mul(out_reg, |
gen3_fs_operand(src_reg, W, W, W, W), |
gen3_fs_operand(FS_T0 + t, X, X, X, X)); |
} else { |
gen3_fs_mul(out_reg, |
gen3_fs_operand(src_reg, X, Y, Z, W), |
gen3_fs_operand(FS_T0 + t, X, X, X, X)); |
} |
} |
goto mask_done; |
case SHADER_CONSTANT: |
case SHADER_ZERO: |
case SHADER_BLACK: |
case SHADER_WHITE: |
case SHADER_NONE: |
break; |
} |
if (mask->alpha_fixup) |
gen3_fs_mov_masked(mask_reg, MASK_W, gen3_fs_operand_one()); |
if (mask->rb_reversed) |
gen3_fs_mov(mask_reg, gen3_fs_operand(mask_reg, Z, Y, X, W)); |
if (dst_is_alpha) { |
switch (src->u.gen3.type) { |
case SHADER_BLACK: |
case SHADER_WHITE: |
gen3_fs_mov(out_reg, |
gen3_fs_operand(mask_reg, W, W, W, W)); |
break; |
default: |
gen3_fs_mul(out_reg, |
gen3_fs_operand(src_reg, W, W, W, W), |
gen3_fs_operand(mask_reg, W, W, W, W)); |
break; |
} |
} else { |
/* If component alpha is active in the mask and the blend |
* operation uses the source alpha, then we know we don't |
* need the source value (otherwise we would have hit a |
* fallback earlier), so we provide the source alpha (src.A * |
* mask.X) as output color. |
* Conversely, if CA is set and we don't need the source alpha, |
* then we produce the source value (src.X * mask.X) and the |
* source alpha is unused. Otherwise, we provide the non-CA |
* source value (src.X * mask.A). |
*/ |
if (op->has_component_alpha) { |
switch (src->u.gen3.type) { |
case SHADER_BLACK: |
if (gen3_blend_op[blend].src_alpha) |
gen3_fs_mov(out_reg, |
gen3_fs_operand_reg(mask_reg)); |
else |
gen3_fs_mov(out_reg, |
gen3_fs_operand(mask_reg, ZERO, ZERO, ZERO, W)); |
break; |
case SHADER_WHITE: |
gen3_fs_mov(out_reg, |
gen3_fs_operand_reg(mask_reg)); |
break; |
default: |
if (gen3_blend_op[blend].src_alpha) |
gen3_fs_mul(out_reg, |
gen3_fs_operand(src_reg, W, W, W, W), |
gen3_fs_operand_reg(mask_reg)); |
else |
gen3_fs_mul(out_reg, |
gen3_fs_operand_reg(src_reg), |
gen3_fs_operand_reg(mask_reg)); |
break; |
} |
} else { |
switch (src->u.gen3.type) { |
case SHADER_WHITE: |
gen3_fs_mov(out_reg, |
gen3_fs_operand(mask_reg, W, W, W, W)); |
break; |
case SHADER_BLACK: |
gen3_fs_mov(out_reg, |
gen3_fs_operand(mask_reg, ZERO, ZERO, ZERO, W)); |
break; |
default: |
gen3_fs_mul(out_reg, |
gen3_fs_operand_reg(src_reg), |
gen3_fs_operand(mask_reg, W, W, W, W)); |
break; |
} |
} |
} |
mask_done: |
if (op->rb_reversed) |
gen3_fs_mov(FS_OC, gen3_fs_operand(FS_U0, Z, Y, X, W)); |
} |
done: |
length = sna->kgem.nbatch - shader_offset; |
sna->kgem.batch[shader_offset] = |
_3DSTATE_PIXEL_SHADER_PROGRAM | (length - 2); |
} |
static uint32_t gen3_ms_tiling(uint32_t tiling) |
{ |
uint32_t v = 0; |
switch (tiling) { |
case I915_TILING_Y: v |= MS3_TILE_WALK; |
case I915_TILING_X: v |= MS3_TILED_SURFACE; |
case I915_TILING_NONE: break; |
} |
return v; |
} |
static void gen3_emit_invariant(struct sna *sna) |
{ |
/* Disable independent alpha blend */ |
OUT_BATCH(_3DSTATE_INDEPENDENT_ALPHA_BLEND_CMD | IAB_MODIFY_ENABLE | |
IAB_MODIFY_FUNC | BLENDFUNC_ADD << IAB_FUNC_SHIFT | |
IAB_MODIFY_SRC_FACTOR | BLENDFACT_ONE << IAB_SRC_FACTOR_SHIFT | |
IAB_MODIFY_DST_FACTOR | BLENDFACT_ZERO << IAB_DST_FACTOR_SHIFT); |
OUT_BATCH(_3DSTATE_COORD_SET_BINDINGS | |
CSB_TCB(0, 0) | |
CSB_TCB(1, 1) | |
CSB_TCB(2, 2) | |
CSB_TCB(3, 3) | |
CSB_TCB(4, 4) | |
CSB_TCB(5, 5) | |
CSB_TCB(6, 6) | |
CSB_TCB(7, 7)); |
OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 | I1_LOAD_S(3) | I1_LOAD_S(4) | I1_LOAD_S(5) | I1_LOAD_S(6) | 3); |
OUT_BATCH(0); /* Disable texture coordinate wrap-shortest */ |
OUT_BATCH((1 << S4_POINT_WIDTH_SHIFT) | |
S4_LINE_WIDTH_ONE | |
S4_CULLMODE_NONE | |
S4_VFMT_XY); |
OUT_BATCH(0); /* Disable fog/stencil. *Enable* write mask. */ |
OUT_BATCH(S6_COLOR_WRITE_ONLY); /* Disable blending, depth */ |
OUT_BATCH(_3DSTATE_SCISSOR_ENABLE_CMD | DISABLE_SCISSOR_RECT); |
OUT_BATCH(_3DSTATE_DEPTH_SUBRECT_DISABLE); |
OUT_BATCH(_3DSTATE_LOAD_INDIRECT); |
OUT_BATCH(0x00000000); |
OUT_BATCH(_3DSTATE_STIPPLE); |
OUT_BATCH(0x00000000); |
sna->render_state.gen3.need_invariant = false; |
} |
#define MAX_OBJECTS 3 /* worst case: dst + src + mask */ |
static void |
gen3_get_batch(struct sna *sna, const struct sna_composite_op *op) |
{ |
kgem_set_mode(&sna->kgem, KGEM_RENDER, op->dst.bo); |
if (!kgem_check_batch(&sna->kgem, 200)) { |
DBG(("%s: flushing batch: size %d > %d\n", |
__FUNCTION__, 200, |
sna->kgem.surface-sna->kgem.nbatch)); |
kgem_submit(&sna->kgem); |
_kgem_set_mode(&sna->kgem, KGEM_RENDER); |
} |
if (!kgem_check_reloc(&sna->kgem, MAX_OBJECTS)) { |
DBG(("%s: flushing batch: reloc %d >= %d\n", |
__FUNCTION__, |
sna->kgem.nreloc, |
(int)KGEM_RELOC_SIZE(&sna->kgem) - MAX_OBJECTS)); |
kgem_submit(&sna->kgem); |
_kgem_set_mode(&sna->kgem, KGEM_RENDER); |
} |
if (!kgem_check_exec(&sna->kgem, MAX_OBJECTS)) { |
DBG(("%s: flushing batch: exec %d >= %d\n", |
__FUNCTION__, |
sna->kgem.nexec, |
(int)KGEM_EXEC_SIZE(&sna->kgem) - MAX_OBJECTS - 1)); |
kgem_submit(&sna->kgem); |
_kgem_set_mode(&sna->kgem, KGEM_RENDER); |
} |
if (sna->render_state.gen3.need_invariant) |
gen3_emit_invariant(sna); |
#undef MAX_OBJECTS |
} |
static void gen3_emit_target(struct sna *sna, |
struct kgem_bo *bo, |
int width, |
int height, |
int format) |
{ |
struct gen3_render_state *state = &sna->render_state.gen3; |
assert(!too_large(width, height)); |
/* BUF_INFO is an implicit flush, so skip if the target is unchanged. */ |
assert(bo->unique_id != 0); |
if (bo->unique_id != state->current_dst) { |
uint32_t v; |
DBG(("%s: setting new target id=%d, handle=%d\n", |
__FUNCTION__, bo->unique_id, bo->handle)); |
OUT_BATCH(_3DSTATE_BUF_INFO_CMD); |
OUT_BATCH(BUF_3D_ID_COLOR_BACK | |
gen3_buf_tiling(bo->tiling) | |
bo->pitch); |
OUT_BATCH(kgem_add_reloc(&sna->kgem, sna->kgem.nbatch, |
bo, |
I915_GEM_DOMAIN_RENDER << 16 | |
I915_GEM_DOMAIN_RENDER, |
0)); |
OUT_BATCH(_3DSTATE_DST_BUF_VARS_CMD); |
OUT_BATCH(gen3_get_dst_format(format)); |
v = DRAW_YMAX(height - 1) | DRAW_XMAX(width - 1); |
if (v != state->last_drawrect_limit) { |
OUT_BATCH(_3DSTATE_DRAW_RECT_CMD); |
OUT_BATCH(0); /* XXX dither origin? */ |
OUT_BATCH(0); |
OUT_BATCH(v); |
OUT_BATCH(0); |
state->last_drawrect_limit = v; |
} |
state->current_dst = bo->unique_id; |
} |
assert(bo->exec); |
kgem_bo_mark_dirty(bo); |
} |
static void gen3_emit_composite_state(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
struct gen3_render_state *state = &sna->render_state.gen3; |
uint32_t map[4]; |
uint32_t sampler[4]; |
struct kgem_bo *bo[2]; |
unsigned int tex_count, n; |
uint32_t ss2; |
gen3_get_batch(sna, op); |
if (kgem_bo_is_dirty(op->src.bo) || kgem_bo_is_dirty(op->mask.bo)) { |
if (op->src.bo == op->dst.bo || op->mask.bo == op->dst.bo) |
OUT_BATCH(MI_FLUSH | MI_INVALIDATE_MAP_CACHE); |
else |
OUT_BATCH(_3DSTATE_MODES_5_CMD | |
PIPELINE_FLUSH_RENDER_CACHE | |
PIPELINE_FLUSH_TEXTURE_CACHE); |
kgem_clear_dirty(&sna->kgem); |
} |
gen3_emit_target(sna, |
op->dst.bo, |
op->dst.width, |
op->dst.height, |
op->dst.format); |
ss2 = ~0; |
tex_count = 0; |
switch (op->src.u.gen3.type) { |
case SHADER_OPACITY: |
case SHADER_NONE: |
assert(0); |
case SHADER_ZERO: |
case SHADER_BLACK: |
case SHADER_WHITE: |
break; |
case SHADER_CONSTANT: |
if (op->src.u.gen3.mode != state->last_diffuse) { |
OUT_BATCH(_3DSTATE_DFLT_DIFFUSE_CMD); |
OUT_BATCH(op->src.u.gen3.mode); |
state->last_diffuse = op->src.u.gen3.mode; |
} |
break; |
case SHADER_LINEAR: |
case SHADER_RADIAL: |
case SHADER_TEXTURE: |
ss2 &= ~S2_TEXCOORD_FMT(tex_count, TEXCOORDFMT_NOT_PRESENT); |
ss2 |= S2_TEXCOORD_FMT(tex_count, |
op->src.is_affine ? TEXCOORDFMT_2D : TEXCOORDFMT_4D); |
assert(op->src.card_format); |
map[tex_count * 2 + 0] = |
op->src.card_format | |
gen3_ms_tiling(op->src.bo->tiling) | |
(op->src.height - 1) << MS3_HEIGHT_SHIFT | |
(op->src.width - 1) << MS3_WIDTH_SHIFT; |
map[tex_count * 2 + 1] = |
(op->src.bo->pitch / 4 - 1) << MS4_PITCH_SHIFT; |
sampler[tex_count * 2 + 0] = op->src.filter; |
sampler[tex_count * 2 + 1] = |
op->src.repeat | |
tex_count << SS3_TEXTUREMAP_INDEX_SHIFT; |
bo[tex_count] = op->src.bo; |
tex_count++; |
break; |
} |
switch (op->mask.u.gen3.type) { |
case SHADER_NONE: |
case SHADER_ZERO: |
case SHADER_BLACK: |
case SHADER_WHITE: |
break; |
case SHADER_CONSTANT: |
if (op->mask.u.gen3.mode != state->last_specular) { |
OUT_BATCH(_3DSTATE_DFLT_SPEC_CMD); |
OUT_BATCH(op->mask.u.gen3.mode); |
state->last_specular = op->mask.u.gen3.mode; |
} |
break; |
case SHADER_LINEAR: |
case SHADER_RADIAL: |
case SHADER_TEXTURE: |
ss2 &= ~S2_TEXCOORD_FMT(tex_count, TEXCOORDFMT_NOT_PRESENT); |
ss2 |= S2_TEXCOORD_FMT(tex_count, |
op->mask.is_affine ? TEXCOORDFMT_2D : TEXCOORDFMT_4D); |
assert(op->mask.card_format); |
map[tex_count * 2 + 0] = |
op->mask.card_format | |
gen3_ms_tiling(op->mask.bo->tiling) | |
(op->mask.height - 1) << MS3_HEIGHT_SHIFT | |
(op->mask.width - 1) << MS3_WIDTH_SHIFT; |
map[tex_count * 2 + 1] = |
(op->mask.bo->pitch / 4 - 1) << MS4_PITCH_SHIFT; |
sampler[tex_count * 2 + 0] = op->mask.filter; |
sampler[tex_count * 2 + 1] = |
op->mask.repeat | |
tex_count << SS3_TEXTUREMAP_INDEX_SHIFT; |
bo[tex_count] = op->mask.bo; |
tex_count++; |
break; |
case SHADER_OPACITY: |
ss2 &= ~S2_TEXCOORD_FMT(tex_count, TEXCOORDFMT_NOT_PRESENT); |
ss2 |= S2_TEXCOORD_FMT(tex_count, TEXCOORDFMT_1D); |
break; |
} |
{ |
uint32_t blend_offset = sna->kgem.nbatch; |
OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 | I1_LOAD_S(2) | I1_LOAD_S(6) | 1); |
OUT_BATCH(ss2); |
OUT_BATCH(gen3_get_blend_cntl(op->op, |
op->has_component_alpha, |
op->dst.format)); |
if (memcmp(sna->kgem.batch + state->last_blend + 1, |
sna->kgem.batch + blend_offset + 1, |
2 * 4) == 0) |
sna->kgem.nbatch = blend_offset; |
else |
state->last_blend = blend_offset; |
} |
if (op->u.gen3.num_constants) { |
int count = op->u.gen3.num_constants; |
if (state->last_constants) { |
int last = sna->kgem.batch[state->last_constants+1]; |
if (last == (1 << (count >> 2)) - 1 && |
memcmp(&sna->kgem.batch[state->last_constants+2], |
op->u.gen3.constants, |
count * sizeof(uint32_t)) == 0) |
count = 0; |
} |
if (count) { |
state->last_constants = sna->kgem.nbatch; |
OUT_BATCH(_3DSTATE_PIXEL_SHADER_CONSTANTS | count); |
OUT_BATCH((1 << (count >> 2)) - 1); |
memcpy(sna->kgem.batch + sna->kgem.nbatch, |
op->u.gen3.constants, |
count * sizeof(uint32_t)); |
sna->kgem.nbatch += count; |
} |
} |
if (tex_count != 0) { |
uint32_t rewind; |
n = 0; |
if (tex_count == state->tex_count) { |
for (; n < tex_count; n++) { |
if (map[2*n+0] != state->tex_map[2*n+0] || |
map[2*n+1] != state->tex_map[2*n+1] || |
state->tex_handle[n] != bo[n]->handle || |
state->tex_delta[n] != bo[n]->delta) |
break; |
} |
} |
if (n < tex_count) { |
OUT_BATCH(_3DSTATE_MAP_STATE | (3 * tex_count)); |
OUT_BATCH((1 << tex_count) - 1); |
for (n = 0; n < tex_count; n++) { |
OUT_BATCH(kgem_add_reloc(&sna->kgem, |
sna->kgem.nbatch, |
bo[n], |
I915_GEM_DOMAIN_SAMPLER<< 16, |
0)); |
OUT_BATCH(map[2*n + 0]); |
OUT_BATCH(map[2*n + 1]); |
state->tex_map[2*n+0] = map[2*n+0]; |
state->tex_map[2*n+1] = map[2*n+1]; |
state->tex_handle[n] = bo[n]->handle; |
state->tex_delta[n] = bo[n]->delta; |
} |
state->tex_count = n; |
} |
rewind = sna->kgem.nbatch; |
OUT_BATCH(_3DSTATE_SAMPLER_STATE | (3 * tex_count)); |
OUT_BATCH((1 << tex_count) - 1); |
for (n = 0; n < tex_count; n++) { |
OUT_BATCH(sampler[2*n + 0]); |
OUT_BATCH(sampler[2*n + 1]); |
OUT_BATCH(0); |
} |
if (state->last_sampler && |
memcmp(&sna->kgem.batch[state->last_sampler+1], |
&sna->kgem.batch[rewind + 1], |
(3*tex_count + 1)*sizeof(uint32_t)) == 0) |
sna->kgem.nbatch = rewind; |
else |
state->last_sampler = rewind; |
} |
gen3_composite_emit_shader(sna, op, op->op); |
} |
static bool gen3_magic_ca_pass(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
if (!op->need_magic_ca_pass) |
return false; |
DBG(("%s(%d)\n", __FUNCTION__, |
sna->render.vertex_index - sna->render.vertex_start)); |
OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 | I1_LOAD_S(6) | 0); |
OUT_BATCH(gen3_get_blend_cntl(PictOpAdd, true, op->dst.format)); |
gen3_composite_emit_shader(sna, op, PictOpAdd); |
OUT_BATCH(PRIM3D_RECTLIST | PRIM3D_INDIRECT_SEQUENTIAL | |
(sna->render.vertex_index - sna->render.vertex_start)); |
OUT_BATCH(sna->render.vertex_start); |
sna->render_state.gen3.last_blend = 0; |
return true; |
} |
static void gen3_vertex_flush(struct sna *sna) |
{ |
assert(sna->render.vertex_offset); |
DBG(("%s[%x] = %d\n", __FUNCTION__, |
4*sna->render.vertex_offset, |
sna->render.vertex_index - sna->render.vertex_start)); |
sna->kgem.batch[sna->render.vertex_offset] = |
PRIM3D_RECTLIST | PRIM3D_INDIRECT_SEQUENTIAL | |
(sna->render.vertex_index - sna->render.vertex_start); |
sna->kgem.batch[sna->render.vertex_offset + 1] = |
sna->render.vertex_start; |
sna->render.vertex_offset = 0; |
} |
static int gen3_vertex_finish(struct sna *sna) |
{ |
struct kgem_bo *bo; |
DBG(("%s: used=%d/%d, vbo active? %d\n", |
__FUNCTION__, sna->render.vertex_used, sna->render.vertex_size, |
sna->render.vbo ? sna->render.vbo->handle : 0)); |
assert(sna->render.vertex_offset == 0); |
assert(sna->render.vertex_used); |
assert(sna->render.vertex_used <= sna->render.vertex_size); |
sna_vertex_wait__locked(&sna->render); |
bo = sna->render.vbo; |
if (bo) { |
DBG(("%s: reloc = %d\n", __FUNCTION__, |
sna->render.vertex_reloc[0])); |
if (sna->render.vertex_reloc[0]) { |
sna->kgem.batch[sna->render.vertex_reloc[0]] = |
kgem_add_reloc(&sna->kgem, sna->render.vertex_reloc[0], |
bo, I915_GEM_DOMAIN_VERTEX << 16, 0); |
sna->render.vertex_reloc[0] = 0; |
} |
sna->render.vertex_used = 0; |
sna->render.vertex_index = 0; |
sna->render.vbo = NULL; |
kgem_bo_destroy(&sna->kgem, bo); |
} |
sna->render.vertices = NULL; |
sna->render.vbo = kgem_create_linear(&sna->kgem, |
256*1024, CREATE_GTT_MAP); |
if (sna->render.vbo) |
sna->render.vertices = kgem_bo_map(&sna->kgem, sna->render.vbo); |
if (sna->render.vertices == NULL) { |
if (sna->render.vbo) |
kgem_bo_destroy(&sna->kgem, sna->render.vbo); |
sna->render.vbo = NULL; |
return 0; |
} |
assert(sna->render.vbo->snoop == false); |
if (sna->render.vertex_used) { |
memcpy(sna->render.vertices, |
sna->render.vertex_data, |
sizeof(float)*sna->render.vertex_used); |
} |
sna->render.vertex_size = 64 * 1024 - 1; |
return sna->render.vertex_size - sna->render.vertex_used; |
} |
static void gen3_vertex_close(struct sna *sna) |
{ |
struct kgem_bo *bo, *free_bo = NULL; |
unsigned int delta = 0; |
assert(sna->render.vertex_offset == 0); |
if (sna->render.vertex_reloc[0] == 0) |
return; |
DBG(("%s: used=%d/%d, vbo active? %d\n", |
__FUNCTION__, sna->render.vertex_used, sna->render.vertex_size, |
sna->render.vbo ? sna->render.vbo->handle : 0)); |
bo = sna->render.vbo; |
if (bo) { |
if (sna->render.vertex_size - sna->render.vertex_used < 64) { |
DBG(("%s: discarding full vbo\n", __FUNCTION__)); |
sna->render.vbo = NULL; |
sna->render.vertices = sna->render.vertex_data; |
sna->render.vertex_size = ARRAY_SIZE(sna->render.vertex_data); |
free_bo = bo; |
} else if (IS_CPU_MAP(bo->map)) { |
DBG(("%s: converting CPU map to GTT\n", __FUNCTION__)); |
sna->render.vertices = kgem_bo_map__gtt(&sna->kgem, bo); |
if (sna->render.vertices == NULL) { |
DBG(("%s: discarding non-mappable vertices\n",__FUNCTION__)); |
sna->render.vbo = NULL; |
sna->render.vertices = sna->render.vertex_data; |
sna->render.vertex_size = ARRAY_SIZE(sna->render.vertex_data); |
free_bo = bo; |
} |
} |
} else { |
if (sna->kgem.nbatch + sna->render.vertex_used <= sna->kgem.surface) { |
DBG(("%s: copy to batch: %d @ %d\n", __FUNCTION__, |
sna->render.vertex_used, sna->kgem.nbatch)); |
memcpy(sna->kgem.batch + sna->kgem.nbatch, |
sna->render.vertex_data, |
sna->render.vertex_used * 4); |
delta = sna->kgem.nbatch * 4; |
bo = NULL; |
sna->kgem.nbatch += sna->render.vertex_used; |
} else { |
DBG(("%s: new vbo: %d\n", __FUNCTION__, |
sna->render.vertex_used)); |
bo = kgem_create_linear(&sna->kgem, |
4*sna->render.vertex_used, |
CREATE_NO_THROTTLE); |
if (bo) { |
assert(bo->snoop == false); |
kgem_bo_write(&sna->kgem, bo, |
sna->render.vertex_data, |
4*sna->render.vertex_used); |
} |
free_bo = bo; |
} |
} |
DBG(("%s: reloc = %d\n", __FUNCTION__, sna->render.vertex_reloc[0])); |
sna->kgem.batch[sna->render.vertex_reloc[0]] = |
kgem_add_reloc(&sna->kgem, sna->render.vertex_reloc[0], |
bo, I915_GEM_DOMAIN_VERTEX << 16, delta); |
sna->render.vertex_reloc[0] = 0; |
if (sna->render.vbo == NULL) { |
DBG(("%s: resetting vbo\n", __FUNCTION__)); |
sna->render.vertex_used = 0; |
sna->render.vertex_index = 0; |
assert(sna->render.vertices == sna->render.vertex_data); |
assert(sna->render.vertex_size == ARRAY_SIZE(sna->render.vertex_data)); |
} |
if (free_bo) |
kgem_bo_destroy(&sna->kgem, free_bo); |
} |
static bool gen3_rectangle_begin(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
struct gen3_render_state *state = &sna->render_state.gen3; |
int ndwords, i1_cmd = 0, i1_len = 0; |
if (sna_vertex_wait__locked(&sna->render) && sna->render.vertex_offset) |
return true; |
ndwords = 2; |
if (op->need_magic_ca_pass) |
ndwords += 100; |
if (sna->render.vertex_reloc[0] == 0) |
i1_len++, i1_cmd |= I1_LOAD_S(0), ndwords++; |
if (state->floats_per_vertex != op->floats_per_vertex) |
i1_len++, i1_cmd |= I1_LOAD_S(1), ndwords++; |
if (!kgem_check_batch(&sna->kgem, ndwords+1)) |
return false; |
if (i1_cmd) { |
OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 | i1_cmd | (i1_len - 1)); |
if (sna->render.vertex_reloc[0] == 0) |
sna->render.vertex_reloc[0] = sna->kgem.nbatch++; |
if (state->floats_per_vertex != op->floats_per_vertex) { |
state->floats_per_vertex = op->floats_per_vertex; |
OUT_BATCH(state->floats_per_vertex << S1_VERTEX_WIDTH_SHIFT | |
state->floats_per_vertex << S1_VERTEX_PITCH_SHIFT); |
} |
} |
if (sna->kgem.nbatch == 2 + state->last_vertex_offset && |
!op->need_magic_ca_pass) { |
sna->render.vertex_offset = state->last_vertex_offset; |
} else { |
sna->render.vertex_offset = sna->kgem.nbatch; |
OUT_BATCH(MI_NOOP); /* to be filled later */ |
OUT_BATCH(MI_NOOP); |
sna->render.vertex_start = sna->render.vertex_index; |
state->last_vertex_offset = sna->render.vertex_offset; |
} |
return true; |
} |
static int gen3_get_rectangles__flush(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
/* Preventing discarding new vbo after lock contention */ |
if (sna_vertex_wait__locked(&sna->render)) { |
int rem = vertex_space(sna); |
if (rem > op->floats_per_rect) |
return rem; |
} |
if (!kgem_check_batch(&sna->kgem, op->need_magic_ca_pass ? 105: 5)) |
return 0; |
if (!kgem_check_reloc_and_exec(&sna->kgem, 1)) |
return 0; |
if (sna->render.vertex_offset) { |
gen3_vertex_flush(sna); |
if (gen3_magic_ca_pass(sna, op)) { |
OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 | I1_LOAD_S(6) | 0); |
OUT_BATCH(gen3_get_blend_cntl(op->op, |
op->has_component_alpha, |
op->dst.format)); |
gen3_composite_emit_shader(sna, op, op->op); |
} |
} |
return gen3_vertex_finish(sna); |
} |
inline static int gen3_get_rectangles(struct sna *sna, |
const struct sna_composite_op *op, |
int want) |
{ |
int rem; |
DBG(("%s: want=%d, rem=%d\n", |
__FUNCTION__, want*op->floats_per_rect, vertex_space(sna))); |
assert(want); |
assert(sna->render.vertex_index * op->floats_per_vertex == sna->render.vertex_used); |
start: |
rem = vertex_space(sna); |
if (unlikely(op->floats_per_rect > rem)) { |
DBG(("flushing vbo for %s: %d < %d\n", |
__FUNCTION__, rem, op->floats_per_rect)); |
rem = gen3_get_rectangles__flush(sna, op); |
if (unlikely(rem == 0)) |
goto flush; |
} |
if (unlikely(sna->render.vertex_offset == 0)) { |
if (!gen3_rectangle_begin(sna, op)) |
goto flush; |
else |
goto start; |
} |
assert(rem <= vertex_space(sna)); |
assert(op->floats_per_rect <= rem); |
if (want > 1 && want * op->floats_per_rect > rem) |
want = rem / op->floats_per_rect; |
sna->render.vertex_index += 3*want; |
assert(want); |
assert(sna->render.vertex_index * op->floats_per_vertex <= sna->render.vertex_size); |
return want; |
flush: |
DBG(("%s: flushing batch\n", __FUNCTION__)); |
if (sna->render.vertex_offset) { |
gen3_vertex_flush(sna); |
gen3_magic_ca_pass(sna, op); |
} |
sna_vertex_wait__locked(&sna->render); |
_kgem_submit(&sna->kgem); |
gen3_emit_composite_state(sna, op); |
assert(sna->render.vertex_offset == 0); |
assert(sna->render.vertex_reloc[0] == 0); |
goto start; |
} |
fastcall static void |
gen3_render_composite_blt(struct sna *sna, |
const struct sna_composite_op *op, |
const struct sna_composite_rectangles *r) |
{ |
DBG(("%s: src=(%d, %d)+(%d, %d), mask=(%d, %d)+(%d, %d), dst=(%d, %d)+(%d, %d), size=(%d, %d)\n", __FUNCTION__, |
r->src.x, r->src.y, op->src.offset[0], op->src.offset[1], |
r->mask.x, r->mask.y, op->mask.offset[0], op->mask.offset[1], |
r->dst.x, r->dst.y, op->dst.x, op->dst.y, |
r->width, r->height)); |
gen3_get_rectangles(sna, op, 1); |
op->prim_emit(sna, op, r); |
} |
static void |
gen3_render_composite_done(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
DBG(("%s()\n", __FUNCTION__)); |
if (sna->render.vertex_offset) { |
gen3_vertex_flush(sna); |
gen3_magic_ca_pass(sna, op); |
} |
} |
static void |
discard_vbo(struct sna *sna) |
{ |
kgem_bo_destroy(&sna->kgem, sna->render.vbo); |
sna->render.vbo = NULL; |
sna->render.vertices = sna->render.vertex_data; |
sna->render.vertex_size = ARRAY_SIZE(sna->render.vertex_data); |
sna->render.vertex_used = 0; |
sna->render.vertex_index = 0; |
} |
static void |
gen3_render_reset(struct sna *sna) |
{ |
struct gen3_render_state *state = &sna->render_state.gen3; |
state->need_invariant = true; |
state->current_dst = 0; |
state->tex_count = 0; |
state->last_drawrect_limit = ~0U; |
state->last_target = 0; |
state->last_blend = 0; |
state->last_constants = 0; |
state->last_sampler = 0; |
state->last_shader = 0x7fffffff; |
state->last_diffuse = 0xcc00ffee; |
state->last_specular = 0xcc00ffee; |
state->floats_per_vertex = 0; |
state->last_floats_per_vertex = 0; |
state->last_vertex_offset = 0; |
if (sna->render.vbo != NULL && |
!kgem_bo_is_mappable(&sna->kgem, sna->render.vbo)) { |
DBG(("%s: discarding vbo as next access will stall: %d\n", |
__FUNCTION__, sna->render.vbo->presumed_offset)); |
discard_vbo(sna); |
} |
sna->render.vertex_reloc[0] = 0; |
sna->render.vertex_offset = 0; |
} |
static void |
gen3_render_retire(struct kgem *kgem) |
{ |
struct sna *sna; |
sna = container_of(kgem, struct sna, kgem); |
if (sna->render.vertex_reloc[0] == 0 && |
sna->render.vbo && !kgem_bo_is_busy(sna->render.vbo)) { |
DBG(("%s: resetting idle vbo\n", __FUNCTION__)); |
sna->render.vertex_used = 0; |
sna->render.vertex_index = 0; |
} |
} |
static void |
gen3_render_expire(struct kgem *kgem) |
{ |
struct sna *sna; |
sna = container_of(kgem, struct sna, kgem); |
if (sna->render.vbo && !sna->render.vertex_used) { |
DBG(("%s: discarding vbo\n", __FUNCTION__)); |
discard_vbo(sna); |
} |
} |
static bool gen3_composite_channel_set_format(struct sna_composite_channel *channel, |
CARD32 format) |
{ |
unsigned int i; |
for (i = 0; i < ARRAY_SIZE(gen3_tex_formats); i++) { |
if (gen3_tex_formats[i].fmt == format) { |
channel->card_format = gen3_tex_formats[i].card_fmt; |
channel->rb_reversed = gen3_tex_formats[i].rb_reversed; |
return true; |
} |
} |
return false; |
} |
#if 0 |
static int |
gen3_composite_picture(struct sna *sna, |
PicturePtr picture, |
struct sna_composite_op *op, |
struct sna_composite_channel *channel, |
int16_t x, int16_t y, |
int16_t w, int16_t h, |
int16_t dst_x, int16_t dst_y, |
bool precise) |
{ |
PixmapPtr pixmap; |
uint32_t color; |
int16_t dx, dy; |
DBG(("%s: (%d, %d)x(%d, %d), dst=(%d, %d)\n", |
__FUNCTION__, x, y, w, h, dst_x, dst_y)); |
channel->card_format = 0; |
if (picture->pDrawable == NULL) { |
SourcePict *source = picture->pSourcePict; |
int ret = -1; |
switch (source->type) { |
case SourcePictTypeSolidFill: |
DBG(("%s: solid fill [%08x], format %08x\n", |
__FUNCTION__, |
(unsigned)source->solidFill.color, |
(unsigned)picture->format)); |
ret = gen3_init_solid(channel, source->solidFill.color); |
break; |
case SourcePictTypeLinear: |
ret = gen3_init_linear(sna, picture, op, channel, |
x - dst_x, y - dst_y); |
break; |
case SourcePictTypeRadial: |
ret = gen3_init_radial(sna, picture, op, channel, |
x - dst_x, y - dst_y); |
break; |
} |
if (ret == -1) { |
if (!precise) |
ret = sna_render_picture_approximate_gradient(sna, picture, channel, |
x, y, w, h, dst_x, dst_y); |
if (ret == -1) |
ret = sna_render_picture_fixup(sna, picture, channel, |
x, y, w, h, dst_x, dst_y); |
} |
return ret; |
} |
if (picture->alphaMap) { |
DBG(("%s -- fallback, alphamap\n", __FUNCTION__)); |
return sna_render_picture_fixup(sna, picture, channel, |
x, y, w, h, dst_x, dst_y); |
} |
if (sna_picture_is_solid(picture, &color)) { |
DBG(("%s: solid drawable [%08x]\n", __FUNCTION__, color)); |
return gen3_init_solid(channel, color); |
} |
if (sna_picture_is_clear(picture, x, y, w, h, &color)) { |
DBG(("%s: clear drawable [%08x]\n", __FUNCTION__, color)); |
return gen3_init_solid(channel, color_convert(color, picture->format, PICT_a8r8g8b8)); |
} |
if (!gen3_check_repeat(picture)) |
return sna_render_picture_fixup(sna, picture, channel, |
x, y, w, h, dst_x, dst_y); |
if (!gen3_check_filter(picture)) |
return sna_render_picture_fixup(sna, picture, channel, |
x, y, w, h, dst_x, dst_y); |
channel->repeat = picture->repeat ? picture->repeatType : RepeatNone; |
channel->filter = picture->filter; |
channel->pict_format = picture->format; |
pixmap = get_drawable_pixmap(picture->pDrawable); |
get_drawable_deltas(picture->pDrawable, pixmap, &dx, &dy); |
x += dx + picture->pDrawable->x; |
y += dy + picture->pDrawable->y; |
if (sna_transform_is_integer_translation(picture->transform, &dx, &dy)) { |
DBG(("%s: integer translation (%d, %d), removing\n", |
__FUNCTION__, dx, dy)); |
x += dx; |
y += dy; |
channel->transform = NULL; |
channel->filter = PictFilterNearest; |
} else { |
channel->transform = picture->transform; |
channel->is_affine = sna_transform_is_affine(picture->transform); |
} |
if (!gen3_composite_channel_set_format(channel, picture->format) && |
!gen3_composite_channel_set_xformat(picture, channel, x, y, w, h)) |
return sna_render_picture_convert(sna, picture, channel, pixmap, |
x, y, w, h, dst_x, dst_y, |
false); |
assert(channel->card_format); |
if (too_large(pixmap->drawable.width, pixmap->drawable.height)) { |
DBG(("%s: pixmap too large (%dx%d), extracting (%d, %d)x(%d,%d)\n", |
__FUNCTION__, |
pixmap->drawable.width, pixmap->drawable.height, |
x, y, w, h)); |
return sna_render_picture_extract(sna, picture, channel, |
x, y, w, h, dst_x, dst_y); |
} |
return sna_render_pixmap_bo(sna, channel, pixmap, |
x, y, w, h, dst_x, dst_y); |
} |
static inline bool |
source_use_blt(struct sna *sna, PicturePtr picture) |
{ |
/* If it is a solid, try to use the BLT paths */ |
if (!picture->pDrawable) |
return picture->pSourcePict->type == SourcePictTypeSolidFill; |
if (picture->pDrawable->width == 1 && |
picture->pDrawable->height == 1 && |
picture->repeat) |
return true; |
if (too_large(picture->pDrawable->width, picture->pDrawable->height)) |
return true; |
return !is_gpu(sna, picture->pDrawable, PREFER_GPU_RENDER); |
} |
static bool |
try_blt(struct sna *sna, |
PicturePtr dst, |
PicturePtr src, |
int width, int height) |
{ |
if (sna->kgem.mode != KGEM_RENDER) { |
DBG(("%s: already performing BLT\n", __FUNCTION__)); |
return true; |
} |
if (too_large(width, height)) { |
DBG(("%s: operation too large for 3D pipe (%d, %d)\n", |
__FUNCTION__, width, height)); |
return true; |
} |
if (too_large(dst->pDrawable->width, dst->pDrawable->height)) { |
DBG(("%s: target too large for 3D pipe (%d, %d)\n", |
__FUNCTION__, |
dst->pDrawable->width, dst->pDrawable->height)); |
return true; |
} |
/* is the source picture only in cpu memory e.g. a shm pixmap? */ |
return source_use_blt(sna, src); |
} |
#endif |
static void |
gen3_align_vertex(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
if (op->floats_per_vertex != sna->render_state.gen3.last_floats_per_vertex) { |
if (sna->render.vertex_size - sna->render.vertex_used < 2*op->floats_per_rect) |
gen3_vertex_finish(sna); |
DBG(("aligning vertex: was %d, now %d floats per vertex, %d->%d\n", |
sna->render_state.gen3.last_floats_per_vertex, |
op->floats_per_vertex, |
sna->render.vertex_index, |
(sna->render.vertex_used + op->floats_per_vertex - 1) / op->floats_per_vertex)); |
sna->render.vertex_index = (sna->render.vertex_used + op->floats_per_vertex - 1) / op->floats_per_vertex; |
sna->render.vertex_used = sna->render.vertex_index * op->floats_per_vertex; |
assert(sna->render.vertex_used < sna->render.vertex_size - op->floats_per_rect); |
sna->render_state.gen3.last_floats_per_vertex = op->floats_per_vertex; |
} |
} |
static inline bool is_constant_ps(uint32_t type) |
{ |
switch (type) { |
case SHADER_NONE: /* be warned! */ |
case SHADER_ZERO: |
case SHADER_BLACK: |
case SHADER_WHITE: |
case SHADER_CONSTANT: |
return true; |
default: |
return false; |
} |
} |
#if 0 |
static bool |
gen3_composite_fallback(struct sna *sna, |
uint8_t op, |
PicturePtr src, |
PicturePtr mask, |
PicturePtr dst) |
{ |
PixmapPtr src_pixmap; |
PixmapPtr mask_pixmap; |
PixmapPtr dst_pixmap; |
bool src_fallback, mask_fallback; |
if (!gen3_check_dst_format(dst->format)) { |
DBG(("%s: unknown destination format: %d\n", |
__FUNCTION__, dst->format)); |
return true; |
} |
dst_pixmap = get_drawable_pixmap(dst->pDrawable); |
src_pixmap = src->pDrawable ? get_drawable_pixmap(src->pDrawable) : NULL; |
src_fallback = source_fallback(src, src_pixmap, |
dst->polyMode == PolyModePrecise); |
if (mask) { |
mask_pixmap = mask->pDrawable ? get_drawable_pixmap(mask->pDrawable) : NULL; |
mask_fallback = source_fallback(mask, mask_pixmap, |
dst->polyMode == PolyModePrecise); |
} else { |
mask_pixmap = NULL; |
mask_fallback = false; |
} |
/* If we are using the destination as a source and need to |
* readback in order to upload the source, do it all |
* on the cpu. |
*/ |
if (src_pixmap == dst_pixmap && src_fallback) { |
DBG(("%s: src is dst and will fallback\n",__FUNCTION__)); |
return true; |
} |
if (mask_pixmap == dst_pixmap && mask_fallback) { |
DBG(("%s: mask is dst and will fallback\n",__FUNCTION__)); |
return true; |
} |
if (mask && |
mask->componentAlpha && PICT_FORMAT_RGB(mask->format) && |
gen3_blend_op[op].src_alpha && |
gen3_blend_op[op].src_blend != BLENDFACT_ZERO && |
op != PictOpOver) { |
DBG(("%s: component-alpha mask with op=%d, should fallback\n", |
__FUNCTION__, op)); |
return true; |
} |
/* If anything is on the GPU, push everything out to the GPU */ |
if (dst_use_gpu(dst_pixmap)) { |
DBG(("%s: dst is already on the GPU, try to use GPU\n", |
__FUNCTION__)); |
return false; |
} |
if (src_pixmap && !src_fallback) { |
DBG(("%s: src is already on the GPU, try to use GPU\n", |
__FUNCTION__)); |
return false; |
} |
if (mask_pixmap && !mask_fallback) { |
DBG(("%s: mask is already on the GPU, try to use GPU\n", |
__FUNCTION__)); |
return false; |
} |
/* However if the dst is not on the GPU and we need to |
* render one of the sources using the CPU, we may |
* as well do the entire operation in place onthe CPU. |
*/ |
if (src_fallback) { |
DBG(("%s: dst is on the CPU and src will fallback\n", |
__FUNCTION__)); |
return true; |
} |
if (mask && mask_fallback) { |
DBG(("%s: dst is on the CPU and mask will fallback\n", |
__FUNCTION__)); |
return true; |
} |
if (too_large(dst_pixmap->drawable.width, |
dst_pixmap->drawable.height) && |
dst_is_cpu(dst_pixmap)) { |
DBG(("%s: dst is on the CPU and too large\n", __FUNCTION__)); |
return true; |
} |
DBG(("%s: dst is not on the GPU and the operation should not fallback: use-cpu? %d\n", |
__FUNCTION__, dst_use_cpu(dst_pixmap))); |
return dst_use_cpu(dst_pixmap); |
} |
static bool |
gen3_render_composite(struct sna *sna, |
uint8_t op, |
PicturePtr src, |
PicturePtr mask, |
PicturePtr dst, |
int16_t src_x, int16_t src_y, |
int16_t mask_x, int16_t mask_y, |
int16_t dst_x, int16_t dst_y, |
int16_t width, int16_t height, |
struct sna_composite_op *tmp) |
{ |
DBG(("%s()\n", __FUNCTION__)); |
if (op >= ARRAY_SIZE(gen3_blend_op)) { |
DBG(("%s: fallback due to unhandled blend op: %d\n", |
__FUNCTION__, op)); |
return false; |
} |
/* Try to use the BLT engine unless it implies a |
* 3D -> 2D context switch. |
*/ |
if (mask == NULL && |
try_blt(sna, dst, src, width, height) && |
sna_blt_composite(sna, |
op, src, dst, |
src_x, src_y, |
dst_x, dst_y, |
width, height, |
tmp, false)) |
return true; |
if (gen3_composite_fallback(sna, op, src, mask, dst)) |
return false; |
if (need_tiling(sna, width, height)) |
return sna_tiling_composite(op, src, mask, dst, |
src_x, src_y, |
mask_x, mask_y, |
dst_x, dst_y, |
width, height, |
tmp); |
if (!gen3_composite_set_target(sna, tmp, dst, |
dst_x, dst_y, width, height)) { |
DBG(("%s: unable to set render target\n", |
__FUNCTION__)); |
return false; |
} |
tmp->op = op; |
tmp->rb_reversed = gen3_dst_rb_reversed(tmp->dst.format); |
if (too_large(tmp->dst.width, tmp->dst.height) || |
!gen3_check_pitch_3d(tmp->dst.bo)) { |
if (!sna_render_composite_redirect(sna, tmp, |
dst_x, dst_y, width, height, |
op > PictOpSrc || dst->pCompositeClip->data)) |
return false; |
} |
tmp->u.gen3.num_constants = 0; |
tmp->src.u.gen3.type = SHADER_TEXTURE; |
tmp->src.is_affine = true; |
DBG(("%s: preparing source\n", __FUNCTION__)); |
switch (gen3_composite_picture(sna, src, tmp, &tmp->src, |
src_x, src_y, |
width, height, |
dst_x, dst_y, |
dst->polyMode == PolyModePrecise)) { |
case -1: |
goto cleanup_dst; |
case 0: |
tmp->src.u.gen3.type = SHADER_ZERO; |
break; |
case 1: |
if (mask == NULL && tmp->src.bo && |
sna_blt_composite__convert(sna, |
dst_x, dst_y, width, height, |
tmp)) |
return true; |
gen3_composite_channel_convert(&tmp->src); |
break; |
} |
DBG(("%s: source type=%d\n", __FUNCTION__, tmp->src.u.gen3.type)); |
tmp->mask.u.gen3.type = SHADER_NONE; |
tmp->mask.is_affine = true; |
tmp->need_magic_ca_pass = false; |
tmp->has_component_alpha = false; |
if (mask && tmp->src.u.gen3.type != SHADER_ZERO) { |
if (!reuse_source(sna, |
src, &tmp->src, src_x, src_y, |
mask, &tmp->mask, mask_x, mask_y)) { |
tmp->mask.u.gen3.type = SHADER_TEXTURE; |
DBG(("%s: preparing mask\n", __FUNCTION__)); |
switch (gen3_composite_picture(sna, mask, tmp, &tmp->mask, |
mask_x, mask_y, |
width, height, |
dst_x, dst_y, |
dst->polyMode == PolyModePrecise)) { |
case -1: |
goto cleanup_src; |
case 0: |
tmp->mask.u.gen3.type = SHADER_ZERO; |
break; |
case 1: |
gen3_composite_channel_convert(&tmp->mask); |
break; |
} |
} |
DBG(("%s: mask type=%d\n", __FUNCTION__, tmp->mask.u.gen3.type)); |
if (tmp->mask.u.gen3.type == SHADER_ZERO) { |
if (tmp->src.bo) { |
kgem_bo_destroy(&sna->kgem, |
tmp->src.bo); |
tmp->src.bo = NULL; |
} |
tmp->src.u.gen3.type = SHADER_ZERO; |
tmp->mask.u.gen3.type = SHADER_NONE; |
} |
if (tmp->mask.u.gen3.type != SHADER_NONE) { |
if (mask->componentAlpha && PICT_FORMAT_RGB(mask->format)) { |
/* Check if it's component alpha that relies on a source alpha |
* and on the source value. We can only get one of those |
* into the single source value that we get to blend with. |
*/ |
DBG(("%s: component-alpha mask: %d\n", |
__FUNCTION__, tmp->mask.u.gen3.type)); |
tmp->has_component_alpha = true; |
if (tmp->mask.u.gen3.type == SHADER_WHITE) { |
tmp->mask.u.gen3.type = SHADER_NONE; |
tmp->has_component_alpha = false; |
} else if (gen3_blend_op[op].src_alpha && |
gen3_blend_op[op].src_blend != BLENDFACT_ZERO) { |
if (op != PictOpOver) |
goto cleanup_mask; |
tmp->need_magic_ca_pass = true; |
tmp->op = PictOpOutReverse; |
} |
} else { |
if (tmp->mask.is_opaque) { |
tmp->mask.u.gen3.type = SHADER_NONE; |
} else if (is_constant_ps(tmp->src.u.gen3.type) && |
is_constant_ps(tmp->mask.u.gen3.type)) { |
uint32_t v; |
v = multa(tmp->src.u.gen3.mode, |
tmp->mask.u.gen3.mode, |
24); |
v |= multa(tmp->src.u.gen3.mode, |
tmp->mask.u.gen3.mode, |
16); |
v |= multa(tmp->src.u.gen3.mode, |
tmp->mask.u.gen3.mode, |
8); |
v |= multa(tmp->src.u.gen3.mode, |
tmp->mask.u.gen3.mode, |
0); |
DBG(("%s: combining constant source/mask: %x x %x -> %x\n", |
__FUNCTION__, |
tmp->src.u.gen3.mode, |
tmp->mask.u.gen3.mode, |
v)); |
tmp->src.u.gen3.type = SHADER_CONSTANT; |
tmp->src.u.gen3.mode = v; |
tmp->src.is_opaque = false; |
tmp->mask.u.gen3.type = SHADER_NONE; |
} |
} |
} |
} |
DBG(("%s: final src/mask type=%d/%d, affine=%d/%d\n", __FUNCTION__, |
tmp->src.u.gen3.type, tmp->mask.u.gen3.type, |
tmp->src.is_affine, tmp->mask.is_affine)); |
tmp->prim_emit = gen3_emit_composite_primitive; |
if (is_constant_ps(tmp->mask.u.gen3.type)) { |
switch (tmp->src.u.gen3.type) { |
case SHADER_NONE: |
case SHADER_ZERO: |
case SHADER_BLACK: |
case SHADER_WHITE: |
case SHADER_CONSTANT: |
#if defined(sse2) && !defined(__x86_64__) |
if (sna->cpu_features & SSE2) { |
tmp->prim_emit = gen3_emit_composite_primitive_constant__sse2; |
tmp->emit_boxes = gen3_emit_composite_boxes_constant__sse2; |
} else |
#endif |
{ |
tmp->prim_emit = gen3_emit_composite_primitive_constant; |
tmp->emit_boxes = gen3_emit_composite_boxes_constant; |
} |
break; |
case SHADER_LINEAR: |
case SHADER_RADIAL: |
if (tmp->src.transform == NULL) { |
#if defined(sse2) && !defined(__x86_64__) |
if (sna->cpu_features & SSE2) { |
tmp->prim_emit = gen3_emit_composite_primitive_identity_gradient__sse2; |
tmp->emit_boxes = gen3_emit_composite_boxes_identity_gradient__sse2; |
} else |
#endif |
{ |
tmp->prim_emit = gen3_emit_composite_primitive_identity_gradient; |
tmp->emit_boxes = gen3_emit_composite_boxes_identity_gradient; |
} |
} else if (tmp->src.is_affine) { |
tmp->src.scale[1] = tmp->src.scale[0] = 1. / tmp->src.transform->matrix[2][2]; |
#if defined(sse2) && !defined(__x86_64__) |
if (sna->cpu_features & SSE2) { |
tmp->prim_emit = gen3_emit_composite_primitive_affine_gradient__sse2; |
tmp->emit_boxes = gen3_emit_composite_boxes_affine_gradient__sse2; |
} else |
#endif |
{ |
tmp->prim_emit = gen3_emit_composite_primitive_affine_gradient; |
tmp->emit_boxes = gen3_emit_composite_boxes_affine_gradient; |
} |
} |
break; |
case SHADER_TEXTURE: |
if (tmp->src.transform == NULL) { |
if ((tmp->src.offset[0]|tmp->src.offset[1]|tmp->dst.x|tmp->dst.y) == 0) { |
#if defined(sse2) && !defined(__x86_64__) |
if (sna->cpu_features & SSE2) { |
tmp->prim_emit = gen3_emit_composite_primitive_identity_source_no_offset__sse2; |
tmp->emit_boxes = gen3_emit_composite_boxes_identity_source_no_offset__sse2; |
} else |
#endif |
{ |
tmp->prim_emit = gen3_emit_composite_primitive_identity_source_no_offset; |
tmp->emit_boxes = gen3_emit_composite_boxes_identity_source_no_offset; |
} |
} else { |
#if defined(sse2) && !defined(__x86_64__) |
if (sna->cpu_features & SSE2) { |
tmp->prim_emit = gen3_emit_composite_primitive_identity_source__sse2; |
tmp->emit_boxes = gen3_emit_composite_boxes_identity_source__sse2; |
} else |
#endif |
{ |
tmp->prim_emit = gen3_emit_composite_primitive_identity_source; |
tmp->emit_boxes = gen3_emit_composite_boxes_identity_source; |
} |
} |
} else if (tmp->src.is_affine) { |
tmp->src.scale[0] /= tmp->src.transform->matrix[2][2]; |
tmp->src.scale[1] /= tmp->src.transform->matrix[2][2]; |
#if defined(sse2) && !defined(__x86_64__) |
if (sna->cpu_features & SSE2) { |
tmp->prim_emit = gen3_emit_composite_primitive_affine_source__sse2; |
tmp->emit_boxes = gen3_emit_composite_boxes_affine_source__sse2; |
} else |
#endif |
{ |
tmp->prim_emit = gen3_emit_composite_primitive_affine_source; |
tmp->emit_boxes = gen3_emit_composite_boxes_affine_source; |
} |
} |
break; |
} |
} else if (tmp->mask.u.gen3.type == SHADER_TEXTURE) { |
if (tmp->mask.transform == NULL) { |
if (is_constant_ps(tmp->src.u.gen3.type)) { |
if ((tmp->mask.offset[0]|tmp->mask.offset[1]|tmp->dst.x|tmp->dst.y) == 0) { |
#if defined(sse2) && !defined(__x86_64__) |
if (sna->cpu_features & SSE2) { |
tmp->prim_emit = gen3_emit_composite_primitive_constant_identity_mask_no_offset__sse2; |
} else |
#endif |
{ |
tmp->prim_emit = gen3_emit_composite_primitive_constant_identity_mask_no_offset; |
} |
} else { |
#if defined(sse2) && !defined(__x86_64__) |
if (sna->cpu_features & SSE2) { |
tmp->prim_emit = gen3_emit_composite_primitive_constant_identity_mask__sse2; |
} else |
#endif |
{ |
tmp->prim_emit = gen3_emit_composite_primitive_constant_identity_mask; |
} |
} |
} else if (tmp->src.transform == NULL) { |
#if defined(sse2) && !defined(__x86_64__) |
if (sna->cpu_features & SSE2) { |
tmp->prim_emit = gen3_emit_composite_primitive_identity_source_mask__sse2; |
} else |
#endif |
{ |
tmp->prim_emit = gen3_emit_composite_primitive_identity_source_mask; |
} |
} else if (tmp->src.is_affine) { |
tmp->src.scale[0] /= tmp->src.transform->matrix[2][2]; |
tmp->src.scale[1] /= tmp->src.transform->matrix[2][2]; |
#if defined(sse2) && !defined(__x86_64__) |
if (sna->cpu_features & SSE2) { |
tmp->prim_emit = gen3_emit_composite_primitive_affine_source_mask__sse2; |
} else |
#endif |
{ |
tmp->prim_emit = gen3_emit_composite_primitive_affine_source_mask; |
} |
} |
} |
} |
tmp->floats_per_vertex = 2; |
if (!is_constant_ps(tmp->src.u.gen3.type)) |
tmp->floats_per_vertex += tmp->src.is_affine ? 2 : 4; |
if (!is_constant_ps(tmp->mask.u.gen3.type)) |
tmp->floats_per_vertex += tmp->mask.is_affine ? 2 : 4; |
DBG(("%s: floats_per_vertex = 2 + %d + %d = %d [specialised emitter? %d]\n", __FUNCTION__, |
!is_constant_ps(tmp->src.u.gen3.type) ? tmp->src.is_affine ? 2 : 4 : 0, |
!is_constant_ps(tmp->mask.u.gen3.type) ? tmp->mask.is_affine ? 2 : 4 : 0, |
tmp->floats_per_vertex, |
tmp->prim_emit != gen3_emit_composite_primitive)); |
tmp->floats_per_rect = 3 * tmp->floats_per_vertex; |
tmp->blt = gen3_render_composite_blt; |
tmp->box = gen3_render_composite_box; |
tmp->boxes = gen3_render_composite_boxes__blt; |
if (tmp->emit_boxes) { |
tmp->boxes = gen3_render_composite_boxes; |
tmp->thread_boxes = gen3_render_composite_boxes__thread; |
} |
tmp->done = gen3_render_composite_done; |
if (!kgem_check_bo(&sna->kgem, |
tmp->dst.bo, tmp->src.bo, tmp->mask.bo, |
NULL)) { |
kgem_submit(&sna->kgem); |
if (!kgem_check_bo(&sna->kgem, |
tmp->dst.bo, tmp->src.bo, tmp->mask.bo, |
NULL)) |
goto cleanup_mask; |
} |
gen3_emit_composite_state(sna, tmp); |
gen3_align_vertex(sna, tmp); |
return true; |
cleanup_mask: |
if (tmp->mask.bo) |
kgem_bo_destroy(&sna->kgem, tmp->mask.bo); |
cleanup_src: |
if (tmp->src.bo) |
kgem_bo_destroy(&sna->kgem, tmp->src.bo); |
cleanup_dst: |
if (tmp->redirect.real_bo) |
kgem_bo_destroy(&sna->kgem, tmp->dst.bo); |
return false; |
} |
#endif |
static void gen3_render_flush(struct sna *sna) |
{ |
gen3_vertex_close(sna); |
assert(sna->render.vertex_reloc[0] == 0); |
assert(sna->render.vertex_offset == 0); |
} |
static void |
gen3_render_fini(struct sna *sna) |
{ |
} |
const char *gen3_render_init(struct sna *sna, const char *backend) |
{ |
struct sna_render *render = &sna->render; |
#if 0 |
#if !NO_COMPOSITE |
render->composite = gen3_render_composite; |
render->prefer_gpu |= PREFER_GPU_RENDER; |
#endif |
#if !NO_COMPOSITE_SPANS |
render->check_composite_spans = gen3_check_composite_spans; |
render->composite_spans = gen3_render_composite_spans; |
render->prefer_gpu |= PREFER_GPU_SPANS; |
#endif |
render->video = gen3_render_video; |
render->copy_boxes = gen3_render_copy_boxes; |
render->copy = gen3_render_copy; |
render->fill_boxes = gen3_render_fill_boxes; |
render->fill = gen3_render_fill; |
render->fill_one = gen3_render_fill_one; |
#endif |
render->blit_tex = gen3_blit_tex; |
render->caps = HW_BIT_BLIT | HW_TEX_BLIT; |
render->reset = gen3_render_reset; |
render->flush = gen3_render_flush; |
render->fini = gen3_render_fini; |
render->max_3d_size = MAX_3D_SIZE; |
render->max_3d_pitch = MAX_3D_PITCH; |
sna->kgem.retire = gen3_render_retire; |
sna->kgem.expire = gen3_render_expire; |
return "Alviso (gen3)"; |
} |
static bool |
gen3_blit_tex(struct sna *sna, |
uint8_t op, bool scale, |
PixmapPtr src, struct kgem_bo *src_bo, |
PixmapPtr mask,struct kgem_bo *mask_bo, |
PixmapPtr dst, struct kgem_bo *dst_bo, |
int32_t src_x, int32_t src_y, |
int32_t msk_x, int32_t msk_y, |
int32_t dst_x, int32_t dst_y, |
int32_t width, int32_t height, |
struct sna_composite_op *tmp) |
{ |
DBG(("%s: %dx%d, current mode=%d\n", __FUNCTION__, |
width, height, sna->kgem.ring)); |
tmp->op = PictOpSrc; |
tmp->dst.pixmap = dst; |
tmp->dst.bo = dst_bo; |
tmp->dst.width = dst->drawable.width; |
tmp->dst.height = dst->drawable.height; |
tmp->dst.format = PICT_x8r8g8b8; |
tmp->rb_reversed = gen3_dst_rb_reversed(tmp->dst.format); |
tmp->u.gen3.num_constants = 0; |
tmp->src.u.gen3.type = SHADER_TEXTURE; |
tmp->src.is_affine = true; |
tmp->src.repeat = RepeatNone; |
tmp->src.filter = PictFilterNearest; |
tmp->src.bo = src_bo; |
tmp->src.pict_format = PICT_x8r8g8b8; |
gen3_composite_channel_set_format(&tmp->src, tmp->src.pict_format); |
tmp->src.width = src->drawable.width; |
tmp->src.height = src->drawable.height; |
tmp->mask.u.gen3.type = SHADER_TEXTURE; |
tmp->mask.is_affine = true; |
tmp->need_magic_ca_pass = false; |
tmp->has_component_alpha = false; |
tmp->mask.repeat = RepeatNone; |
tmp->mask.filter = PictFilterNearest; |
tmp->mask.is_affine = true; |
tmp->mask.bo = mask_bo; |
tmp->mask.pict_format = PIXMAN_a8; |
gen3_composite_channel_set_format(&tmp->mask, tmp->mask.pict_format); |
tmp->mask.width = mask->drawable.width; |
tmp->mask.height = mask->drawable.height; |
if( scale ) |
{ |
tmp->src.scale[0] = 1.f/width; |
tmp->src.scale[1] = 1.f/height; |
} |
else |
{ |
tmp->src.scale[0] = 1.f/src->drawable.width; |
tmp->src.scale[1] = 1.f/src->drawable.height; |
} |
tmp->mask.scale[0] = 1.f/mask->drawable.width; |
tmp->mask.scale[1] = 1.f/mask->drawable.height; |
tmp->prim_emit = gen3_emit_composite_primitive_identity_source_mask; |
tmp->floats_per_vertex = 2; |
if (!is_constant_ps(tmp->src.u.gen3.type)) |
tmp->floats_per_vertex += tmp->src.is_affine ? 2 : 4; |
if (!is_constant_ps(tmp->mask.u.gen3.type)) |
tmp->floats_per_vertex += tmp->mask.is_affine ? 2 : 4; |
// DBG(("%s: floats_per_vertex = 2 + %d + %d = %d [specialised emitter? %d]\n", __FUNCTION__, |
// !is_constant_ps(tmp->src.u.gen3.type) ? tmp->src.is_affine ? 2 : 4 : 0, |
// !is_constant_ps(tmp->mask.u.gen3.type) ? tmp->mask.is_affine ? 2 : 4 : 0, |
// tmp->floats_per_vertex, |
// tmp->prim_emit != gen3_emit_composite_primitive)); |
tmp->floats_per_rect = 3 * tmp->floats_per_vertex; |
tmp->blt = gen3_render_composite_blt; |
tmp->done = gen3_render_composite_done; |
if (!kgem_check_bo(&sna->kgem, |
tmp->dst.bo, tmp->src.bo, tmp->mask.bo, |
NULL)) { |
kgem_submit(&sna->kgem); |
} |
gen3_emit_composite_state(sna, tmp); |
gen3_align_vertex(sna, tmp); |
return true; |
} |
/drivers/video/Intel-2D/sna/gen3_render.h |
---|
0,0 → 1,1479 |
/************************************************************************** |
* |
* Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. |
* All Rights Reserved. |
* |
* Permission is hereby granted, free of charge, to any person obtaining a |
* copy of this software and associated documentation files (the |
* "Software"), to deal in the Software without restriction, including |
* without limitation the rights to use, copy, modify, merge, publish, |
* distribute, sub license, and/or sell copies of the Software, and to |
* permit persons to whom the Software is furnished to do so, subject to |
* the following conditions: |
* |
* The above copyright notice and this permission notice (including the |
* next paragraph) shall be included in all copies or substantial portions |
* of the Software. |
* |
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS |
* OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF |
* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. |
* IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR |
* ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, |
* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE |
* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. |
* |
**************************************************************************/ |
#ifndef _I915_REG_H_ |
#define _I915_REG_H_ |
#define CMD_3D (3 << 29) |
#define I915_SET_FIELD( var, mask, value ) (var &= ~(mask), var |= value) |
#define PRIM3D (CMD_3D | (0x1f<<24)) |
#define PRIM3D_INDIRECT_SEQUENTIAL ((1<<23) | (0<<17)) |
#define PRIM3D_TRILIST (PRIM3D | (0x0<<18)) |
#define PRIM3D_TRISTRIP (PRIM3D | (0x1<<18)) |
#define PRIM3D_TRISTRIP_RVRSE (PRIM3D | (0x2<<18)) |
#define PRIM3D_TRIFAN (PRIM3D | (0x3<<18)) |
#define PRIM3D_POLY (PRIM3D | (0x4<<18)) |
#define PRIM3D_LINELIST (PRIM3D | (0x5<<18)) |
#define PRIM3D_LINESTRIP (PRIM3D | (0x6<<18)) |
#define PRIM3D_RECTLIST (PRIM3D | (0x7<<18)) |
#define PRIM3D_POINTLIST (PRIM3D | (0x8<<18)) |
#define PRIM3D_DIB (PRIM3D | (0x9<<18)) |
#define PRIM3D_CLEAR_RECT (PRIM3D | (0xa<<18)) |
#define PRIM3D_ZONE_INIT (PRIM3D | (0xd<<18)) |
#define PRIM3D_MASK (0x1f<<18) |
/* p137 */ |
#define _3DSTATE_AA_CMD (CMD_3D | (0x06<<24)) |
#define AA_LINE_ECAAR_WIDTH_ENABLE (1<<16) |
#define AA_LINE_ECAAR_WIDTH_0_5 0 |
#define AA_LINE_ECAAR_WIDTH_1_0 (1<<14) |
#define AA_LINE_ECAAR_WIDTH_2_0 (2<<14) |
#define AA_LINE_ECAAR_WIDTH_4_0 (3<<14) |
#define AA_LINE_REGION_WIDTH_ENABLE (1<<8) |
#define AA_LINE_REGION_WIDTH_0_5 0 |
#define AA_LINE_REGION_WIDTH_1_0 (1<<6) |
#define AA_LINE_REGION_WIDTH_2_0 (2<<6) |
#define AA_LINE_REGION_WIDTH_4_0 (3<<6) |
/* 3DSTATE_BACKFACE_STENCIL_OPS, p138*/ |
#define _3DSTATE_BACKFACE_STENCIL_OPS (CMD_3D | (0x8<<24)) |
#define BFO_ENABLE_STENCIL_REF (1<<23) |
#define BFO_STENCIL_REF_SHIFT 15 |
#define BFO_STENCIL_REF_MASK (0xff<<15) |
#define BFO_ENABLE_STENCIL_FUNCS (1<<14) |
#define BFO_STENCIL_TEST_SHIFT 11 |
#define BFO_STENCIL_TEST_MASK (0x7<<11) |
#define BFO_STENCIL_FAIL_SHIFT 8 |
#define BFO_STENCIL_FAIL_MASK (0x7<<8) |
#define BFO_STENCIL_PASS_Z_FAIL_SHIFT 5 |
#define BFO_STENCIL_PASS_Z_FAIL_MASK (0x7<<5) |
#define BFO_STENCIL_PASS_Z_PASS_SHIFT 2 |
#define BFO_STENCIL_PASS_Z_PASS_MASK (0x7<<2) |
#define BFO_ENABLE_STENCIL_TWO_SIDE (1<<1) |
#define BFO_STENCIL_TWO_SIDE (1<<0) |
/* 3DSTATE_BACKFACE_STENCIL_MASKS, p140 */ |
#define _3DSTATE_BACKFACE_STENCIL_MASKS (CMD_3D | (0x9<<24)) |
#define BFM_ENABLE_STENCIL_TEST_MASK (1<<17) |
#define BFM_ENABLE_STENCIL_WRITE_MASK (1<<16) |
#define BFM_STENCIL_TEST_MASK_SHIFT 8 |
#define BFM_STENCIL_TEST_MASK_MASK (0xff<<8) |
#define BFM_STENCIL_WRITE_MASK_SHIFT 0 |
#define BFM_STENCIL_WRITE_MASK_MASK (0xff<<0) |
/* 3DSTATE_BIN_CONTROL p141 */ |
/* p143 */ |
#define _3DSTATE_BUF_INFO_CMD (CMD_3D | (0x1d<<24) | (0x8e<<16) | 1) |
/* Dword 1 */ |
#define BUF_3D_ID_COLOR_BACK (0x3<<24) |
#define BUF_3D_ID_DEPTH (0x7<<24) |
#define BUF_3D_USE_FENCE (1<<23) |
#define BUF_3D_TILED_SURFACE (1<<22) |
#define BUF_3D_TILE_WALK_X 0 |
#define BUF_3D_TILE_WALK_Y (1<<21) |
/* Dword 2 */ |
#define BUF_3D_ADDR(x) ((x) & ~0x3) |
/* 3DSTATE_CHROMA_KEY */ |
/* 3DSTATE_CLEAR_PARAMETERS, p150 */ |
#define _3DSTATE_CLEAR_PARAMETERS (CMD_3D | (0x1d<<24) | (0x9c<<16) | 5) |
/* Dword 1 */ |
#define CLEARPARAM_CLEAR_RECT (1 << 16) |
#define CLEARPARAM_ZONE_INIT (0 << 16) |
#define CLEARPARAM_WRITE_COLOR (1 << 2) |
#define CLEARPARAM_WRITE_DEPTH (1 << 1) |
#define CLEARPARAM_WRITE_STENCIL (1 << 0) |
/* 3DSTATE_CONSTANT_BLEND_COLOR, p153 */ |
#define _3DSTATE_CONST_BLEND_COLOR_CMD (CMD_3D | (0x1d<<24) | (0x88<<16)) |
/* 3DSTATE_COORD_SET_BINDINGS, p154 */ |
#define _3DSTATE_COORD_SET_BINDINGS (CMD_3D | (0x16<<24)) |
#define CSB_TCB(iunit, eunit) ((eunit)<<(iunit*3)) |
/* p156 */ |
#define _3DSTATE_DFLT_DIFFUSE_CMD (CMD_3D | (0x1d<<24) | (0x99<<16)) |
/* p157 */ |
#define _3DSTATE_DFLT_SPEC_CMD (CMD_3D | (0x1d<<24) | (0x9a<<16)) |
/* p158 */ |
#define _3DSTATE_DFLT_Z_CMD (CMD_3D | (0x1d<<24) | (0x98<<16)) |
/* 3DSTATE_DEPTH_OFFSET_SCALE, p159 */ |
#define _3DSTATE_DEPTH_OFFSET_SCALE (CMD_3D | (0x1d<<24) | (0x97<<16)) |
/* scale in dword 1 */ |
/* The depth subrectangle is not supported, but must be disabled. */ |
/* 3DSTATE_DEPTH_SUBRECT_DISABLE, p160 */ |
#define _3DSTATE_DEPTH_SUBRECT_DISABLE (CMD_3D | (0x1c<<24) | (0x11<<19) | (1 << 1) | (0 << 0)) |
/* p161 */ |
#define _3DSTATE_DST_BUF_VARS_CMD (CMD_3D | (0x1d<<24) | (0x85<<16)) |
/* Dword 1 */ |
#define TEX_DEFAULT_COLOR_OGL (0<<30) |
#define TEX_DEFAULT_COLOR_D3D (1<<30) |
#define ZR_EARLY_DEPTH (1<<29) |
#define LOD_PRECLAMP_OGL (1<<28) |
#define LOD_PRECLAMP_D3D (0<<28) |
#define DITHER_FULL_ALWAYS (0<<26) |
#define DITHER_FULL_ON_FB_BLEND (1<<26) |
#define DITHER_CLAMPED_ALWAYS (2<<26) |
#define LINEAR_GAMMA_BLEND_32BPP (1<<25) |
#define DEBUG_DISABLE_ENH_DITHER (1<<24) |
#define DSTORG_HORT_BIAS(x) ((x)<<20) |
#define DSTORG_VERT_BIAS(x) ((x)<<16) |
#define COLOR_4_2_2_CHNL_WRT_ALL 0 |
#define COLOR_4_2_2_CHNL_WRT_Y (1<<12) |
#define COLOR_4_2_2_CHNL_WRT_CR (2<<12) |
#define COLOR_4_2_2_CHNL_WRT_CB (3<<12) |
#define COLOR_4_2_2_CHNL_WRT_CRCB (4<<12) |
#define COLR_BUF_8BIT 0 |
#define COLR_BUF_RGB555 (1<<8) |
#define COLR_BUF_RGB565 (2<<8) |
#define COLR_BUF_ARGB8888 (3<<8) |
#define COLR_BUF_ARGB4444 (8<<8) |
#define COLR_BUF_ARGB1555 (9<<8) |
#define COLR_BUF_ARGB2AAA (0xa<<8) |
#define DEPTH_IS_Z 0 |
#define DEPTH_IS_W (1<<6) |
#define DEPTH_FRMT_16_FIXED 0 |
#define DEPTH_FRMT_16_FLOAT (1<<2) |
#define DEPTH_FRMT_24_FIXED_8_OTHER (2<<2) |
#define DEPTH_FRMT_24_FLOAT_8_OTHER (3<<2) |
#define VERT_LINE_STRIDE_1 (1<<1) |
#define VERT_LINE_STRIDE_0 0 |
#define VERT_LINE_STRIDE_OFS_1 1 |
#define VERT_LINE_STRIDE_OFS_0 0 |
/* p166 */ |
#define _3DSTATE_DRAW_RECT_CMD (CMD_3D|(0x1d<<24)|(0x80<<16)|3) |
/* Dword 1 */ |
#define DRAW_RECT_DIS_DEPTH_OFS (1<<30) |
#define DRAW_DITHER_OFS_X(x) ((x)<<26) |
#define DRAW_DITHER_OFS_Y(x) ((x)<<24) |
/* Dword 2 */ |
#define DRAW_YMIN(x) ((uint16_t)(x)<<16) |
#define DRAW_XMIN(x) ((uint16_t)(x)) |
/* Dword 3 */ |
#define DRAW_YMAX(x) ((uint16_t)(x)<<16) |
#define DRAW_XMAX(x) ((uint16_t)(x)) |
/* Dword 4 */ |
#define DRAW_YORG(x) ((uint16_t)(x)<<16) |
#define DRAW_XORG(x) ((uint16_t)(x)) |
/* 3DSTATE_FILTER_COEFFICIENTS_4X4, p170 */ |
/* 3DSTATE_FILTER_COEFFICIENTS_6X5, p172 */ |
/* _3DSTATE_FOG_COLOR, p173 */ |
#define _3DSTATE_FOG_COLOR_CMD (CMD_3D|(0x15<<24)) |
#define FOG_COLOR_RED(x) ((x)<<16) |
#define FOG_COLOR_GREEN(x) ((x)<<8) |
#define FOG_COLOR_BLUE(x) (x) |
/* _3DSTATE_FOG_MODE, p174 */ |
#define _3DSTATE_FOG_MODE_CMD (CMD_3D|(0x1d<<24)|(0x89<<16)|2) |
/* Dword 1 */ |
#define FMC1_FOGFUNC_MODIFY_ENABLE (1<<31) |
#define FMC1_FOGFUNC_VERTEX (0<<28) |
#define FMC1_FOGFUNC_PIXEL_EXP (1<<28) |
#define FMC1_FOGFUNC_PIXEL_EXP2 (2<<28) |
#define FMC1_FOGFUNC_PIXEL_LINEAR (3<<28) |
#define FMC1_FOGFUNC_MASK (3<<28) |
#define FMC1_FOGINDEX_MODIFY_ENABLE (1<<27) |
#define FMC1_FOGINDEX_Z (0<<25) |
#define FMC1_FOGINDEX_W (1<<25) |
#define FMC1_C1_C2_MODIFY_ENABLE (1<<24) |
#define FMC1_DENSITY_MODIFY_ENABLE (1<<23) |
#define FMC1_C1_ONE (1<<13) |
#define FMC1_C1_MASK (0xffff<<4) |
/* Dword 2 */ |
#define FMC2_C2_ONE (1<<16) |
/* Dword 3 */ |
#define FMC3_D_ONE (1<<16) |
/* _3DSTATE_INDEPENDENT_ALPHA_BLEND, p177 */ |
#define _3DSTATE_INDEPENDENT_ALPHA_BLEND_CMD (CMD_3D|(0x0b<<24)) |
#define IAB_MODIFY_ENABLE (1<<23) |
#define IAB_ENABLE (1<<22) |
#define IAB_MODIFY_FUNC (1<<21) |
#define IAB_FUNC_SHIFT 16 |
#define IAB_MODIFY_SRC_FACTOR (1<<11) |
#define IAB_SRC_FACTOR_SHIFT 6 |
#define IAB_SRC_FACTOR_MASK (BLENDFACT_MASK<<6) |
#define IAB_MODIFY_DST_FACTOR (1<<5) |
#define IAB_DST_FACTOR_SHIFT 0 |
#define IAB_DST_FACTOR_MASK (BLENDFACT_MASK<<0) |
#define BLENDFACT_ZERO 0x01 |
#define BLENDFACT_ONE 0x02 |
#define BLENDFACT_SRC_COLR 0x03 |
#define BLENDFACT_INV_SRC_COLR 0x04 |
#define BLENDFACT_SRC_ALPHA 0x05 |
#define BLENDFACT_INV_SRC_ALPHA 0x06 |
#define BLENDFACT_DST_ALPHA 0x07 |
#define BLENDFACT_INV_DST_ALPHA 0x08 |
#define BLENDFACT_DST_COLR 0x09 |
#define BLENDFACT_INV_DST_COLR 0x0a |
#define BLENDFACT_SRC_ALPHA_SATURATE 0x0b |
#define BLENDFACT_CONST_COLOR 0x0c |
#define BLENDFACT_INV_CONST_COLOR 0x0d |
#define BLENDFACT_CONST_ALPHA 0x0e |
#define BLENDFACT_INV_CONST_ALPHA 0x0f |
#define BLENDFACT_MASK 0x0f |
#define BLENDFUNC_ADD 0x0 |
#define BLENDFUNC_SUBTRACT 0x1 |
#define BLENDFUNC_REVERSE_SUBTRACT 0x2 |
#define BLENDFUNC_MIN 0x3 |
#define BLENDFUNC_MAX 0x4 |
#define BLENDFUNC_MASK 0x7 |
/* 3DSTATE_LOAD_INDIRECT, p180 */ |
#define _3DSTATE_LOAD_INDIRECT (CMD_3D|(0x1d<<24)|(0x7<<16)) |
#define LI0_STATE_STATIC_INDIRECT (0x01<<8) |
#define LI0_STATE_DYNAMIC_INDIRECT (0x02<<8) |
#define LI0_STATE_SAMPLER (0x04<<8) |
#define LI0_STATE_MAP (0x08<<8) |
#define LI0_STATE_PROGRAM (0x10<<8) |
#define LI0_STATE_CONSTANTS (0x20<<8) |
#define SIS0_BUFFER_ADDRESS(x) ((x)&~0x3) |
#define SIS0_FORCE_LOAD (1<<1) |
#define SIS0_BUFFER_VALID (1<<0) |
#define SIS1_BUFFER_LENGTH(x) ((x)&0xff) |
#define DIS0_BUFFER_ADDRESS(x) ((x)&~0x3) |
#define DIS0_BUFFER_RESET (1<<1) |
#define DIS0_BUFFER_VALID (1<<0) |
#define SSB0_BUFFER_ADDRESS(x) ((x)&~0x3) |
#define SSB0_FORCE_LOAD (1<<1) |
#define SSB0_BUFFER_VALID (1<<0) |
#define SSB1_BUFFER_LENGTH(x) ((x)&0xff) |
#define MSB0_BUFFER_ADDRESS(x) ((x)&~0x3) |
#define MSB0_FORCE_LOAD (1<<1) |
#define MSB0_BUFFER_VALID (1<<0) |
#define MSB1_BUFFER_LENGTH(x) ((x)&0xff) |
#define PSP0_BUFFER_ADDRESS(x) ((x)&~0x3) |
#define PSP0_FORCE_LOAD (1<<1) |
#define PSP0_BUFFER_VALID (1<<0) |
#define PSP1_BUFFER_LENGTH(x) ((x)&0xff) |
#define PSC0_BUFFER_ADDRESS(x) ((x)&~0x3) |
#define PSC0_FORCE_LOAD (1<<1) |
#define PSC0_BUFFER_VALID (1<<0) |
#define PSC1_BUFFER_LENGTH(x) ((x)&0xff) |
/* _3DSTATE_RASTERIZATION_RULES */ |
#define _3DSTATE_RASTER_RULES_CMD (CMD_3D|(0x07<<24)) |
#define ENABLE_POINT_RASTER_RULE (1<<15) |
#define OGL_POINT_RASTER_RULE (1<<13) |
#define ENABLE_TEXKILL_3D_4D (1<<10) |
#define TEXKILL_3D (0<<9) |
#define TEXKILL_4D (1<<9) |
#define ENABLE_LINE_STRIP_PROVOKE_VRTX (1<<8) |
#define ENABLE_TRI_FAN_PROVOKE_VRTX (1<<5) |
#define LINE_STRIP_PROVOKE_VRTX(x) ((x)<<6) |
#define TRI_FAN_PROVOKE_VRTX(x) ((x)<<3) |
/* _3DSTATE_SCISSOR_ENABLE, p256 */ |
#define _3DSTATE_SCISSOR_ENABLE_CMD (CMD_3D|(0x1c<<24)|(0x10<<19)) |
#define ENABLE_SCISSOR_RECT ((1<<1) | 1) |
#define DISABLE_SCISSOR_RECT (1<<1) |
/* _3DSTATE_SCISSOR_RECTANGLE_0, p257 */ |
#define _3DSTATE_SCISSOR_RECT_0_CMD (CMD_3D|(0x1d<<24)|(0x81<<16)|1) |
/* Dword 1 */ |
#define SCISSOR_RECT_0_YMIN(x) ((x)<<16) |
#define SCISSOR_RECT_0_XMIN(x) (x) |
/* Dword 2 */ |
#define SCISSOR_RECT_0_YMAX(x) ((x)<<16) |
#define SCISSOR_RECT_0_XMAX(x) (x) |
/* p189 */ |
#define _3DSTATE_LOAD_STATE_IMMEDIATE_1 ((0x3<<29)|(0x1d<<24)|(0x04<<16)) |
#define I1_LOAD_S(n) (1<<(4+n)) |
#define S0_VB_OFFSET_MASK 0xffffffc |
#define S0_AUTO_CACHE_INV_DISABLE (1<<0) |
#define S1_VERTEX_WIDTH_SHIFT 24 |
#define S1_VERTEX_WIDTH_MASK (0x3f<<24) |
#define S1_VERTEX_PITCH_SHIFT 16 |
#define S1_VERTEX_PITCH_MASK (0x3f<<16) |
#define TEXCOORDFMT_2D 0x0 |
#define TEXCOORDFMT_3D 0x1 |
#define TEXCOORDFMT_4D 0x2 |
#define TEXCOORDFMT_1D 0x3 |
#define TEXCOORDFMT_2D_16 0x4 |
#define TEXCOORDFMT_4D_16 0x5 |
#define TEXCOORDFMT_NOT_PRESENT 0xf |
#define S2_TEXCOORD_FMT0_MASK 0xf |
#define S2_TEXCOORD_FMT1_SHIFT 4 |
#define S2_TEXCOORD_FMT(unit, type) ((type)<<(unit*4)) |
#define S2_TEXCOORD_NONE (~0) |
#define TEXCOORD_WRAP_SHORTEST_TCX 8 |
#define TEXCOORD_WRAP_SHORTEST_TCY 4 |
#define TEXCOORD_WRAP_SHORTEST_TCZ 2 |
#define TEXCOORD_PERSPECTIVE_DISABLE 1 |
#define S3_WRAP_SHORTEST_TCX(unit) (TEXCOORD_WRAP_SHORTEST_TCX << ((unit) * 4)) |
#define S3_WRAP_SHORTEST_TCY(unit) (TEXCOORD_WRAP_SHORTEST_TCY << ((unit) * 4)) |
#define S3_WRAP_SHORTEST_TCZ(unit) (TEXCOORD_WRAP_SHORTEST_TCZ << ((unit) * 4)) |
#define S3_PERSPECTIVE_DISABLE(unit) (TEXCOORD_PERSPECTIVE_DISABLE << ((unit) * 4)) |
/* S3 not interesting */ |
#define S4_POINT_WIDTH_SHIFT 23 |
#define S4_POINT_WIDTH_MASK (0x1ff<<23) |
#define S4_LINE_WIDTH_SHIFT 19 |
#define S4_LINE_WIDTH_ONE (0x2<<19) |
#define S4_LINE_WIDTH_MASK (0xf<<19) |
#define S4_FLATSHADE_ALPHA (1<<18) |
#define S4_FLATSHADE_FOG (1<<17) |
#define S4_FLATSHADE_SPECULAR (1<<16) |
#define S4_FLATSHADE_COLOR (1<<15) |
#define S4_CULLMODE_BOTH (0<<13) |
#define S4_CULLMODE_NONE (1<<13) |
#define S4_CULLMODE_CW (2<<13) |
#define S4_CULLMODE_CCW (3<<13) |
#define S4_CULLMODE_MASK (3<<13) |
#define S4_VFMT_POINT_WIDTH (1<<12) |
#define S4_VFMT_SPEC_FOG (1<<11) |
#define S4_VFMT_COLOR (1<<10) |
#define S4_VFMT_DEPTH_OFFSET (1<<9) |
#define S4_VFMT_XYZ (1<<6) |
#define S4_VFMT_XYZW (2<<6) |
#define S4_VFMT_XY (3<<6) |
#define S4_VFMT_XYW (4<<6) |
#define S4_VFMT_XYZW_MASK (7<<6) |
#define S4_FORCE_DEFAULT_DIFFUSE (1<<5) |
#define S4_FORCE_DEFAULT_SPECULAR (1<<4) |
#define S4_LOCAL_DEPTH_OFFSET_ENABLE (1<<3) |
#define S4_VFMT_FOG_PARAM (1<<2) |
#define S4_SPRITE_POINT_ENABLE (1<<1) |
#define S4_LINE_ANTIALIAS_ENABLE (1<<0) |
#define S4_VFMT_MASK (S4_VFMT_POINT_WIDTH | \ |
S4_VFMT_SPEC_FOG | \ |
S4_VFMT_COLOR | \ |
S4_VFMT_DEPTH_OFFSET | \ |
S4_VFMT_XYZW_MASK | \ |
S4_VFMT_FOG_PARAM) |
#define S5_WRITEDISABLE_ALPHA (1<<31) |
#define S5_WRITEDISABLE_RED (1<<30) |
#define S5_WRITEDISABLE_GREEN (1<<29) |
#define S5_WRITEDISABLE_BLUE (1<<28) |
#define S5_WRITEDISABLE_MASK (0xf<<28) |
#define S5_FORCE_DEFAULT_POINT_SIZE (1<<27) |
#define S5_LAST_PIXEL_ENABLE (1<<26) |
#define S5_GLOBAL_DEPTH_OFFSET_ENABLE (1<<25) |
#define S5_FOG_ENABLE (1<<24) |
#define S5_STENCIL_REF_SHIFT 16 |
#define S5_STENCIL_REF_MASK (0xff<<16) |
#define S5_STENCIL_TEST_FUNC_SHIFT 13 |
#define S5_STENCIL_TEST_FUNC_MASK (0x7<<13) |
#define S5_STENCIL_FAIL_SHIFT 10 |
#define S5_STENCIL_FAIL_MASK (0x7<<10) |
#define S5_STENCIL_PASS_Z_FAIL_SHIFT 7 |
#define S5_STENCIL_PASS_Z_FAIL_MASK (0x7<<7) |
#define S5_STENCIL_PASS_Z_PASS_SHIFT 4 |
#define S5_STENCIL_PASS_Z_PASS_MASK (0x7<<4) |
#define S5_STENCIL_WRITE_ENABLE (1<<3) |
#define S5_STENCIL_TEST_ENABLE (1<<2) |
#define S5_COLOR_DITHER_ENABLE (1<<1) |
#define S5_LOGICOP_ENABLE (1<<0) |
#define S6_ALPHA_TEST_ENABLE (1<<31) |
#define S6_ALPHA_TEST_FUNC_SHIFT 28 |
#define S6_ALPHA_TEST_FUNC_MASK (0x7<<28) |
#define S6_ALPHA_REF_SHIFT 20 |
#define S6_ALPHA_REF_MASK (0xff<<20) |
#define S6_DEPTH_TEST_ENABLE (1<<19) |
#define S6_DEPTH_TEST_FUNC_SHIFT 16 |
#define S6_DEPTH_TEST_FUNC_MASK (0x7<<16) |
#define S6_CBUF_BLEND_ENABLE (1<<15) |
#define S6_CBUF_BLEND_FUNC_SHIFT 12 |
#define S6_CBUF_BLEND_FUNC_MASK (0x7<<12) |
#define S6_CBUF_SRC_BLEND_FACT_SHIFT 8 |
#define S6_CBUF_SRC_BLEND_FACT_MASK (0xf<<8) |
#define S6_CBUF_DST_BLEND_FACT_SHIFT 4 |
#define S6_CBUF_DST_BLEND_FACT_MASK (0xf<<4) |
#define S6_DEPTH_WRITE_ENABLE (1<<3) |
#define S6_COLOR_WRITE_ENABLE (1<<2) |
#define S6_TRISTRIP_PV_SHIFT 0 |
#define S6_TRISTRIP_PV_MASK (0x3<<0) |
#define S7_DEPTH_OFFSET_CONST_MASK ~0 |
/* 3DSTATE_MAP_DEINTERLACER_PARAMETERS */ |
/* 3DSTATE_MAP_PALETTE_LOAD_32, p206 */ |
/* _3DSTATE_MODES_4, p218 */ |
#define _3DSTATE_MODES_4_CMD (CMD_3D|(0x0d<<24)) |
#define ENABLE_LOGIC_OP_FUNC (1<<23) |
#define LOGIC_OP_FUNC(x) ((x)<<18) |
#define LOGICOP_MASK (0xf<<18) |
#define LOGICOP_COPY 0xc |
#define MODE4_ENABLE_STENCIL_TEST_MASK ((1<<17)|(0xff00)) |
#define ENABLE_STENCIL_TEST_MASK (1<<17) |
#define STENCIL_TEST_MASK(x) ((x)<<8) |
#define MODE4_ENABLE_STENCIL_WRITE_MASK ((1<<16)|(0x00ff)) |
#define ENABLE_STENCIL_WRITE_MASK (1<<16) |
#define STENCIL_WRITE_MASK(x) ((x)&0xff) |
/* _3DSTATE_MODES_5, p220 */ |
#define _3DSTATE_MODES_5_CMD (CMD_3D|(0x0c<<24)) |
#define PIPELINE_FLUSH_RENDER_CACHE (1<<18) |
#define PIPELINE_FLUSH_TEXTURE_CACHE (1<<16) |
/* p221 */ |
#define _3DSTATE_PIXEL_SHADER_CONSTANTS (CMD_3D|(0x1d<<24)|(0x6<<16)) |
#define PS1_REG(n) (1<<(n)) |
#define PS2_CONST_X(n) (n) |
#define PS3_CONST_Y(n) (n) |
#define PS4_CONST_Z(n) (n) |
#define PS5_CONST_W(n) (n) |
/* p222 */ |
#define I915_MAX_TEX_INDIRECT 4 |
#define I915_MAX_TEX_INSN 32 |
#define I915_MAX_ALU_INSN 64 |
#define I915_MAX_DECL_INSN 27 |
#define I915_MAX_TEMPORARY 16 |
/* Each instruction is 3 dwords long, though most don't require all |
* this space. Maximum of 123 instructions. Smaller maxes per insn |
* type. |
*/ |
#define _3DSTATE_PIXEL_SHADER_PROGRAM (CMD_3D|(0x1d<<24)|(0x5<<16)) |
#define REG_TYPE_R 0 /* temporary regs, no need to |
* dcl, must be written before |
* read -- Preserved between |
* phases. |
*/ |
#define REG_TYPE_T 1 /* Interpolated values, must be |
* dcl'ed before use. |
* |
* 0..7: texture coord, |
* 8: diffuse spec, |
* 9: specular color, |
* 10: fog parameter in w. |
*/ |
#define REG_TYPE_CONST 2 /* Restriction: only one const |
* can be referenced per |
* instruction, though it may be |
* selected for multiple inputs. |
* Constants not initialized |
* default to zero. |
*/ |
#define REG_TYPE_S 3 /* sampler */ |
#define REG_TYPE_OC 4 /* output color (rgba) */ |
#define REG_TYPE_OD 5 /* output depth (w), xyz are |
* temporaries. If not written, |
* interpolated depth is used? |
*/ |
#define REG_TYPE_U 6 /* unpreserved temporaries */ |
#define REG_TYPE_MASK 0x7 |
#define REG_NR_MASK 0xf |
/* REG_TYPE_T: |
*/ |
#define T_TEX0 0 |
#define T_TEX1 1 |
#define T_TEX2 2 |
#define T_TEX3 3 |
#define T_TEX4 4 |
#define T_TEX5 5 |
#define T_TEX6 6 |
#define T_TEX7 7 |
#define T_DIFFUSE 8 |
#define T_SPECULAR 9 |
#define T_FOG_W 10 /* interpolated fog is in W coord */ |
/* Arithmetic instructions */ |
/* .replicate_swizzle == selection and replication of a particular |
* scalar channel, ie., .xxxx, .yyyy, .zzzz or .wwww |
*/ |
#define A0_NOP (0x0<<24) /* no operation */ |
#define A0_ADD (0x1<<24) /* dst = src0 + src1 */ |
#define A0_MOV (0x2<<24) /* dst = src0 */ |
#define A0_MUL (0x3<<24) /* dst = src0 * src1 */ |
#define A0_MAD (0x4<<24) /* dst = src0 * src1 + src2 */ |
#define A0_DP2ADD (0x5<<24) /* dst.xyzw = src0.xy dot src1.xy + src2.replicate_swizzle */ |
#define A0_DP3 (0x6<<24) /* dst.xyzw = src0.xyz dot src1.xyz */ |
#define A0_DP4 (0x7<<24) /* dst.xyzw = src0.xyzw dot src1.xyzw */ |
#define A0_FRC (0x8<<24) /* dst = src0 - floor(src0) */ |
#define A0_RCP (0x9<<24) /* dst.xyzw = 1/(src0.replicate_swizzle) */ |
#define A0_RSQ (0xa<<24) /* dst.xyzw = 1/(sqrt(abs(src0.replicate_swizzle))) */ |
#define A0_EXP (0xb<<24) /* dst.xyzw = exp2(src0.replicate_swizzle) */ |
#define A0_LOG (0xc<<24) /* dst.xyzw = log2(abs(src0.replicate_swizzle)) */ |
#define A0_CMP (0xd<<24) /* dst = (src0 >= 0.0) ? src1 : src2 */ |
#define A0_MIN (0xe<<24) /* dst = (src0 < src1) ? src0 : src1 */ |
#define A0_MAX (0xf<<24) /* dst = (src0 >= src1) ? src0 : src1 */ |
#define A0_FLR (0x10<<24) /* dst = floor(src0) */ |
#define A0_MOD (0x11<<24) /* dst = src0 fmod 1.0 */ |
#define A0_TRC (0x12<<24) /* dst = int(src0) */ |
#define A0_SGE (0x13<<24) /* dst = src0 >= src1 ? 1.0 : 0.0 */ |
#define A0_SLT (0x14<<24) /* dst = src0 < src1 ? 1.0 : 0.0 */ |
#define A0_DEST_SATURATE (1<<22) |
#define A0_DEST_TYPE_SHIFT 19 |
/* Allow: R, OC, OD, U */ |
#define A0_DEST_NR_SHIFT 14 |
/* Allow R: 0..15, OC,OD: 0..0, U: 0..2 */ |
#define A0_DEST_CHANNEL_X (1<<10) |
#define A0_DEST_CHANNEL_Y (2<<10) |
#define A0_DEST_CHANNEL_Z (4<<10) |
#define A0_DEST_CHANNEL_W (8<<10) |
#define A0_DEST_CHANNEL_ALL (0xf<<10) |
#define A0_DEST_CHANNEL_SHIFT 10 |
#define A0_SRC0_TYPE_SHIFT 7 |
#define A0_SRC0_NR_SHIFT 2 |
#define A0_DEST_CHANNEL_XY (A0_DEST_CHANNEL_X|A0_DEST_CHANNEL_Y) |
#define A0_DEST_CHANNEL_XYZ (A0_DEST_CHANNEL_XY|A0_DEST_CHANNEL_Z) |
#define SRC_X 0 |
#define SRC_Y 1 |
#define SRC_Z 2 |
#define SRC_W 3 |
#define SRC_ZERO 4 |
#define SRC_ONE 5 |
#define A1_SRC0_CHANNEL_X_NEGATE (1<<31) |
#define A1_SRC0_CHANNEL_X_SHIFT 28 |
#define A1_SRC0_CHANNEL_Y_NEGATE (1<<27) |
#define A1_SRC0_CHANNEL_Y_SHIFT 24 |
#define A1_SRC0_CHANNEL_Z_NEGATE (1<<23) |
#define A1_SRC0_CHANNEL_Z_SHIFT 20 |
#define A1_SRC0_CHANNEL_W_NEGATE (1<<19) |
#define A1_SRC0_CHANNEL_W_SHIFT 16 |
#define A1_SRC1_TYPE_SHIFT 13 |
#define A1_SRC1_NR_SHIFT 8 |
#define A1_SRC1_CHANNEL_X_NEGATE (1<<7) |
#define A1_SRC1_CHANNEL_X_SHIFT 4 |
#define A1_SRC1_CHANNEL_Y_NEGATE (1<<3) |
#define A1_SRC1_CHANNEL_Y_SHIFT 0 |
#define A2_SRC1_CHANNEL_Z_NEGATE (1<<31) |
#define A2_SRC1_CHANNEL_Z_SHIFT 28 |
#define A2_SRC1_CHANNEL_W_NEGATE (1<<27) |
#define A2_SRC1_CHANNEL_W_SHIFT 24 |
#define A2_SRC2_TYPE_SHIFT 21 |
#define A2_SRC2_NR_SHIFT 16 |
#define A2_SRC2_CHANNEL_X_NEGATE (1<<15) |
#define A2_SRC2_CHANNEL_X_SHIFT 12 |
#define A2_SRC2_CHANNEL_Y_NEGATE (1<<11) |
#define A2_SRC2_CHANNEL_Y_SHIFT 8 |
#define A2_SRC2_CHANNEL_Z_NEGATE (1<<7) |
#define A2_SRC2_CHANNEL_Z_SHIFT 4 |
#define A2_SRC2_CHANNEL_W_NEGATE (1<<3) |
#define A2_SRC2_CHANNEL_W_SHIFT 0 |
/* Texture instructions */ |
#define T0_TEXLD (0x15<<24) /* Sample texture using predeclared |
* sampler and address, and output |
* filtered texel data to destination |
* register */ |
#define T0_TEXLDP (0x16<<24) /* Same as texld but performs a |
* perspective divide of the texture |
* coordinate .xyz values by .w before |
* sampling. */ |
#define T0_TEXLDB (0x17<<24) /* Same as texld but biases the |
* computed LOD by w. Only S4.6 two's |
* comp is used. This implies that a |
* float to fixed conversion is |
* done. */ |
#define T0_TEXKILL (0x18<<24) /* Does not perform a sampling |
* operation. Simply kills the pixel |
* if any channel of the address |
* register is < 0.0. */ |
#define T0_DEST_TYPE_SHIFT 19 |
/* Allow: R, OC, OD, U */ |
/* Note: U (unpreserved) regs do not retain their values between |
* phases (cannot be used for feedback) |
* |
* Note: oC and OD registers can only be used as the destination of a |
* texture instruction once per phase (this is an implementation |
* restriction). |
*/ |
#define T0_DEST_NR_SHIFT 14 |
/* Allow R: 0..15, OC,OD: 0..0, U: 0..2 */ |
#define T0_SAMPLER_NR_SHIFT 0 /* This field ignored for TEXKILL */ |
#define T0_SAMPLER_NR_MASK (0xf<<0) |
#define T1_ADDRESS_REG_TYPE_SHIFT 24 /* Reg to use as texture coord */ |
/* Allow R, T, OC, OD -- R, OC, OD are 'dependent' reads, new program phase */ |
#define T1_ADDRESS_REG_NR_SHIFT 17 |
#define T2_MBZ 0 |
/* Declaration instructions */ |
#define D0_DCL (0x19<<24) /* Declare a t (interpolated attrib) |
* register or an s (sampler) |
* register. */ |
#define D0_SAMPLE_TYPE_SHIFT 22 |
#define D0_SAMPLE_TYPE_2D (0x0<<22) |
#define D0_SAMPLE_TYPE_CUBE (0x1<<22) |
#define D0_SAMPLE_TYPE_VOLUME (0x2<<22) |
#define D0_SAMPLE_TYPE_MASK (0x3<<22) |
#define D0_TYPE_SHIFT 19 |
/* Allow: T, S */ |
#define D0_NR_SHIFT 14 |
/* Allow T: 0..10, S: 0..15 */ |
#define D0_CHANNEL_X (1<<10) |
#define D0_CHANNEL_Y (2<<10) |
#define D0_CHANNEL_Z (4<<10) |
#define D0_CHANNEL_W (8<<10) |
#define D0_CHANNEL_ALL (0xf<<10) |
#define D0_CHANNEL_NONE (0<<10) |
#define D0_CHANNEL_XY (D0_CHANNEL_X|D0_CHANNEL_Y) |
#define D0_CHANNEL_XYZ (D0_CHANNEL_XY|D0_CHANNEL_Z) |
/* I915 Errata: Do not allow (xz), (xw), (xzw) combinations for diffuse |
* or specular declarations. |
* |
* For T dcls, only allow: (x), (xy), (xyz), (w), (xyzw) |
* |
* Must be zero for S (sampler) dcls |
*/ |
#define D1_MBZ 0 |
#define D2_MBZ 0 |
/* p207. |
* The DWORD count is 3 times the number of bits set in MS1_MAPMASK_MASK |
*/ |
#define _3DSTATE_MAP_STATE (CMD_3D|(0x1d<<24)|(0x0<<16)) |
#define MS1_MAPMASK_SHIFT 0 |
#define MS1_MAPMASK_MASK (0x8fff<<0) |
#define MS2_UNTRUSTED_SURFACE (1<<31) |
#define MS2_ADDRESS_MASK 0xfffffffc |
#define MS2_VERTICAL_LINE_STRIDE (1<<1) |
#define MS2_VERTICAL_OFFSET (1<<1) |
#define MS3_HEIGHT_SHIFT 21 |
#define MS3_WIDTH_SHIFT 10 |
#define MS3_PALETTE_SELECT (1<<9) |
#define MS3_MAPSURF_FORMAT_SHIFT 7 |
#define MS3_MAPSURF_FORMAT_MASK (0x7<<7) |
#define MAPSURF_8BIT (1<<7) |
#define MAPSURF_16BIT (2<<7) |
#define MAPSURF_32BIT (3<<7) |
#define MAPSURF_422 (5<<7) |
#define MAPSURF_COMPRESSED (6<<7) |
#define MAPSURF_4BIT_INDEXED (7<<7) |
#define MS3_MT_FORMAT_MASK (0x7 << 3) |
#define MS3_MT_FORMAT_SHIFT 3 |
#define MT_4BIT_IDX_ARGB8888 (7<<3) /* SURFACE_4BIT_INDEXED */ |
#define MT_8BIT_I8 (0<<3) /* SURFACE_8BIT */ |
#define MT_8BIT_L8 (1<<3) |
#define MT_8BIT_A8 (4<<3) |
#define MT_8BIT_MONO8 (5<<3) |
#define MT_16BIT_RGB565 (0<<3) /* SURFACE_16BIT */ |
#define MT_16BIT_ARGB1555 (1<<3) |
#define MT_16BIT_ARGB4444 (2<<3) |
#define MT_16BIT_AY88 (3<<3) |
#define MT_16BIT_88DVDU (5<<3) |
#define MT_16BIT_BUMP_655LDVDU (6<<3) |
#define MT_16BIT_I16 (7<<3) |
#define MT_16BIT_L16 (8<<3) |
#define MT_16BIT_A16 (9<<3) |
#define MT_32BIT_ARGB8888 (0<<3) /* SURFACE_32BIT */ |
#define MT_32BIT_ABGR8888 (1<<3) |
#define MT_32BIT_XRGB8888 (2<<3) |
#define MT_32BIT_XBGR8888 (3<<3) |
#define MT_32BIT_QWVU8888 (4<<3) |
#define MT_32BIT_AXVU8888 (5<<3) |
#define MT_32BIT_LXVU8888 (6<<3) |
#define MT_32BIT_XLVU8888 (7<<3) |
#define MT_32BIT_ARGB2101010 (8<<3) |
#define MT_32BIT_ABGR2101010 (9<<3) |
#define MT_32BIT_AWVU2101010 (0xA<<3) |
#define MT_32BIT_GR1616 (0xB<<3) |
#define MT_32BIT_VU1616 (0xC<<3) |
#define MT_32BIT_xI824 (0xD<<3) |
#define MT_32BIT_xA824 (0xE<<3) |
#define MT_32BIT_xL824 (0xF<<3) |
#define MT_422_YCRCB_SWAPY (0<<3) /* SURFACE_422 */ |
#define MT_422_YCRCB_NORMAL (1<<3) |
#define MT_422_YCRCB_SWAPUV (2<<3) |
#define MT_422_YCRCB_SWAPUVY (3<<3) |
#define MT_COMPRESS_DXT1 (0<<3) /* SURFACE_COMPRESSED */ |
#define MT_COMPRESS_DXT2_3 (1<<3) |
#define MT_COMPRESS_DXT4_5 (2<<3) |
#define MT_COMPRESS_FXT1 (3<<3) |
#define MT_COMPRESS_DXT1_RGB (4<<3) |
#define MS3_USE_FENCE_REGS (1<<2) |
#define MS3_TILED_SURFACE (1<<1) |
#define MS3_TILE_WALK (1<<0) |
/* The pitch is the pitch measured in DWORDS, minus 1 */ |
#define MS4_PITCH_SHIFT 21 |
#define MS4_CUBE_FACE_ENA_NEGX (1<<20) |
#define MS4_CUBE_FACE_ENA_POSX (1<<19) |
#define MS4_CUBE_FACE_ENA_NEGY (1<<18) |
#define MS4_CUBE_FACE_ENA_POSY (1<<17) |
#define MS4_CUBE_FACE_ENA_NEGZ (1<<16) |
#define MS4_CUBE_FACE_ENA_POSZ (1<<15) |
#define MS4_CUBE_FACE_ENA_MASK (0x3f<<15) |
#define MS4_MAX_LOD_SHIFT 9 |
#define MS4_MAX_LOD_MASK (0x3f<<9) |
#define MS4_MIP_LAYOUT_LEGACY (0<<8) |
#define MS4_MIP_LAYOUT_BELOW_LPT (0<<8) |
#define MS4_MIP_LAYOUT_RIGHT_LPT (1<<8) |
#define MS4_VOLUME_DEPTH_SHIFT 0 |
#define MS4_VOLUME_DEPTH_MASK (0xff<<0) |
/* p244. |
* The DWORD count is 3 times the number of bits set in SS1_MAPMASK_MASK. |
*/ |
#define _3DSTATE_SAMPLER_STATE (CMD_3D|(0x1d<<24)|(0x1<<16)) |
#define SS1_MAPMASK_SHIFT 0 |
#define SS1_MAPMASK_MASK (0x8fff<<0) |
#define SS2_REVERSE_GAMMA_ENABLE (1<<31) |
#define SS2_PACKED_TO_PLANAR_ENABLE (1<<30) |
#define SS2_COLORSPACE_CONVERSION (1<<29) |
#define SS2_CHROMAKEY_SHIFT 27 |
#define SS2_BASE_MIP_LEVEL_SHIFT 22 |
#define SS2_BASE_MIP_LEVEL_MASK (0x1f<<22) |
#define SS2_MIP_FILTER_SHIFT 20 |
#define SS2_MIP_FILTER_MASK (0x3<<20) |
#define MIPFILTER_NONE 0 |
#define MIPFILTER_NEAREST 1 |
#define MIPFILTER_LINEAR 3 |
#define SS2_MAG_FILTER_SHIFT 17 |
#define SS2_MAG_FILTER_MASK (0x7<<17) |
#define FILTER_NEAREST 0 |
#define FILTER_LINEAR 1 |
#define FILTER_ANISOTROPIC 2 |
#define FILTER_4X4_1 3 |
#define FILTER_4X4_2 4 |
#define FILTER_4X4_FLAT 5 |
#define FILTER_6X5_MONO 6 /* XXX - check */ |
#define SS2_MIN_FILTER_SHIFT 14 |
#define SS2_MIN_FILTER_MASK (0x7<<14) |
#define SS2_LOD_BIAS_SHIFT 5 |
#define SS2_LOD_BIAS_ONE (0x10<<5) |
#define SS2_LOD_BIAS_MASK (0x1ff<<5) |
/* Shadow requires: |
* MT_X8{I,L,A}24 or MT_{I,L,A}16 texture format |
* FILTER_4X4_x MIN and MAG filters |
*/ |
#define SS2_SHADOW_ENABLE (1<<4) |
#define SS2_MAX_ANISO_MASK (1<<3) |
#define SS2_MAX_ANISO_2 (0<<3) |
#define SS2_MAX_ANISO_4 (1<<3) |
#define SS2_SHADOW_FUNC_SHIFT 0 |
#define SS2_SHADOW_FUNC_MASK (0x7<<0) |
/* SS2_SHADOW_FUNC values: see COMPAREFUNC_* */ |
#define SS3_MIN_LOD_SHIFT 24 |
#define SS3_MIN_LOD_ONE (0x10<<24) |
#define SS3_MIN_LOD_MASK (0xff<<24) |
#define SS3_KILL_PIXEL_ENABLE (1<<17) |
#define SS3_TCX_ADDR_MODE_SHIFT 12 |
#define SS3_TCX_ADDR_MODE_MASK (0x7<<12) |
#define TEXCOORDMODE_WRAP 0 |
#define TEXCOORDMODE_MIRROR 1 |
#define TEXCOORDMODE_CLAMP_EDGE 2 |
#define TEXCOORDMODE_CUBE 3 |
#define TEXCOORDMODE_CLAMP_BORDER 4 |
#define TEXCOORDMODE_MIRROR_ONCE 5 |
#define SS3_TCY_ADDR_MODE_SHIFT 9 |
#define SS3_TCY_ADDR_MODE_MASK (0x7<<9) |
#define SS3_TCZ_ADDR_MODE_SHIFT 6 |
#define SS3_TCZ_ADDR_MODE_MASK (0x7<<6) |
#define SS3_NORMALIZED_COORDS (1<<5) |
#define SS3_TEXTUREMAP_INDEX_SHIFT 1 |
#define SS3_TEXTUREMAP_INDEX_MASK (0xf<<1) |
#define SS3_DEINTERLACER_ENABLE (1<<0) |
#define SS4_BORDER_COLOR_MASK (~0) |
/* 3DSTATE_SPAN_STIPPLE, p258 |
*/ |
#define _3DSTATE_STIPPLE ((0x3<<29)|(0x1d<<24)|(0x83<<16)) |
#define ST1_ENABLE (1<<16) |
#define ST1_MASK (0xffff) |
#define FLUSH_MAP_CACHE (1<<0) |
#define FLUSH_RENDER_CACHE (1<<1) |
#endif |
/* -*- c-basic-offset: 4 -*- */ |
/* |
* Copyright © 2006,2010 Intel Corporation |
* |
* Permission is hereby granted, free of charge, to any person obtaining a |
* copy of this software and associated documentation files (the "Software"), |
* to deal in the Software without restriction, including without limitation |
* the rights to use, copy, modify, merge, publish, distribute, sublicense, |
* and/or sell copies of the Software, and to permit persons to whom the |
* Software is furnished to do so, subject to the following conditions: |
* |
* The above copyright notice and this permission notice (including the next |
* paragraph) shall be included in all copies or substantial portions of the |
* Software. |
* |
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL |
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, |
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE |
* SOFTWARE. |
* |
* Authors: |
* Eric Anholt <eric@anholt.net> |
* Chris Wilson <chris@chris-wilson.co.uk> |
* |
*/ |
/* Each instruction is 3 dwords long, though most don't require all |
* this space. Maximum of 123 instructions. Smaller maxes per insn |
* type. |
*/ |
#define _3DSTATE_PIXEL_SHADER_PROGRAM (CMD_3D|(0x1d<<24)|(0x5<<16)) |
#define REG_TYPE_R 0 /* temporary regs, no need to |
* dcl, must be written before |
* read -- Preserved between |
* phases. |
*/ |
#define REG_TYPE_T 1 /* Interpolated values, must be |
* dcl'ed before use. |
* |
* 0..7: texture coord, |
* 8: diffuse spec, |
* 9: specular color, |
* 10: fog parameter in w. |
*/ |
#define REG_TYPE_CONST 2 /* Restriction: only one const |
* can be referenced per |
* instruction, though it may be |
* selected for multiple inputs. |
* Constants not initialized |
* default to zero. |
*/ |
#define REG_TYPE_S 3 /* sampler */ |
#define REG_TYPE_OC 4 /* output color (rgba) */ |
#define REG_TYPE_OD 5 /* output depth (w), xyz are |
* temporaries. If not written, |
* interpolated depth is used? |
*/ |
#define REG_TYPE_U 6 /* unpreserved temporaries */ |
#define REG_TYPE_MASK 0x7 |
#define REG_TYPE_SHIFT 4 |
#define REG_NR_MASK 0xf |
/* REG_TYPE_T: |
*/ |
#define T_TEX0 0 |
#define T_TEX1 1 |
#define T_TEX2 2 |
#define T_TEX3 3 |
#define T_TEX4 4 |
#define T_TEX5 5 |
#define T_TEX6 6 |
#define T_TEX7 7 |
#define T_DIFFUSE 8 |
#define T_SPECULAR 9 |
#define T_FOG_W 10 /* interpolated fog is in W coord */ |
/* Arithmetic instructions */ |
/* .replicate_swizzle == selection and replication of a particular |
* scalar channel, ie., .xxxx, .yyyy, .zzzz or .wwww |
*/ |
#define A0_NOP (0x0<<24) /* no operation */ |
#define A0_ADD (0x1<<24) /* dst = src0 + src1 */ |
#define A0_MOV (0x2<<24) /* dst = src0 */ |
#define A0_MUL (0x3<<24) /* dst = src0 * src1 */ |
#define A0_MAD (0x4<<24) /* dst = src0 * src1 + src2 */ |
#define A0_DP2ADD (0x5<<24) /* dst.xyzw = src0.xy dot src1.xy + src2.replicate_swizzle */ |
#define A0_DP3 (0x6<<24) /* dst.xyzw = src0.xyz dot src1.xyz */ |
#define A0_DP4 (0x7<<24) /* dst.xyzw = src0.xyzw dot src1.xyzw */ |
#define A0_FRC (0x8<<24) /* dst = src0 - floor(src0) */ |
#define A0_RCP (0x9<<24) /* dst.xyzw = 1/(src0.replicate_swizzle) */ |
#define A0_RSQ (0xa<<24) /* dst.xyzw = 1/(sqrt(abs(src0.replicate_swizzle))) */ |
#define A0_EXP (0xb<<24) /* dst.xyzw = exp2(src0.replicate_swizzle) */ |
#define A0_LOG (0xc<<24) /* dst.xyzw = log2(abs(src0.replicate_swizzle)) */ |
#define A0_CMP (0xd<<24) /* dst = (src0 >= 0.0) ? src1 : src2 */ |
#define A0_MIN (0xe<<24) /* dst = (src0 < src1) ? src0 : src1 */ |
#define A0_MAX (0xf<<24) /* dst = (src0 >= src1) ? src0 : src1 */ |
#define A0_FLR (0x10<<24) /* dst = floor(src0) */ |
#define A0_MOD (0x11<<24) /* dst = src0 fmod 1.0 */ |
#define A0_TRC (0x12<<24) /* dst = int(src0) */ |
#define A0_SGE (0x13<<24) /* dst = src0 >= src1 ? 1.0 : 0.0 */ |
#define A0_SLT (0x14<<24) /* dst = src0 < src1 ? 1.0 : 0.0 */ |
#define A0_DEST_SATURATE (1<<22) |
#define A0_DEST_TYPE_SHIFT 19 |
/* Allow: R, OC, OD, U */ |
#define A0_DEST_NR_SHIFT 14 |
/* Allow R: 0..15, OC,OD: 0..0, U: 0..2 */ |
#define A0_DEST_CHANNEL_X (1<<10) |
#define A0_DEST_CHANNEL_Y (2<<10) |
#define A0_DEST_CHANNEL_Z (4<<10) |
#define A0_DEST_CHANNEL_W (8<<10) |
#define A0_DEST_CHANNEL_ALL (0xf<<10) |
#define A0_DEST_CHANNEL_SHIFT 10 |
#define A0_SRC0_TYPE_SHIFT 7 |
#define A0_SRC0_NR_SHIFT 2 |
#define A0_DEST_CHANNEL_XY (A0_DEST_CHANNEL_X|A0_DEST_CHANNEL_Y) |
#define A0_DEST_CHANNEL_XYZ (A0_DEST_CHANNEL_XY|A0_DEST_CHANNEL_Z) |
#define SRC_X 0 |
#define SRC_Y 1 |
#define SRC_Z 2 |
#define SRC_W 3 |
#define SRC_ZERO 4 |
#define SRC_ONE 5 |
#define A1_SRC0_CHANNEL_X_NEGATE (1<<31) |
#define A1_SRC0_CHANNEL_X_SHIFT 28 |
#define A1_SRC0_CHANNEL_Y_NEGATE (1<<27) |
#define A1_SRC0_CHANNEL_Y_SHIFT 24 |
#define A1_SRC0_CHANNEL_Z_NEGATE (1<<23) |
#define A1_SRC0_CHANNEL_Z_SHIFT 20 |
#define A1_SRC0_CHANNEL_W_NEGATE (1<<19) |
#define A1_SRC0_CHANNEL_W_SHIFT 16 |
#define A1_SRC1_TYPE_SHIFT 13 |
#define A1_SRC1_NR_SHIFT 8 |
#define A1_SRC1_CHANNEL_X_NEGATE (1<<7) |
#define A1_SRC1_CHANNEL_X_SHIFT 4 |
#define A1_SRC1_CHANNEL_Y_NEGATE (1<<3) |
#define A1_SRC1_CHANNEL_Y_SHIFT 0 |
#define A2_SRC1_CHANNEL_Z_NEGATE (1<<31) |
#define A2_SRC1_CHANNEL_Z_SHIFT 28 |
#define A2_SRC1_CHANNEL_W_NEGATE (1<<27) |
#define A2_SRC1_CHANNEL_W_SHIFT 24 |
#define A2_SRC2_TYPE_SHIFT 21 |
#define A2_SRC2_NR_SHIFT 16 |
#define A2_SRC2_CHANNEL_X_NEGATE (1<<15) |
#define A2_SRC2_CHANNEL_X_SHIFT 12 |
#define A2_SRC2_CHANNEL_Y_NEGATE (1<<11) |
#define A2_SRC2_CHANNEL_Y_SHIFT 8 |
#define A2_SRC2_CHANNEL_Z_NEGATE (1<<7) |
#define A2_SRC2_CHANNEL_Z_SHIFT 4 |
#define A2_SRC2_CHANNEL_W_NEGATE (1<<3) |
#define A2_SRC2_CHANNEL_W_SHIFT 0 |
/* Texture instructions */ |
#define T0_TEXLD (0x15<<24) /* Sample texture using predeclared |
* sampler and address, and output |
* filtered texel data to destination |
* register */ |
#define T0_TEXLDP (0x16<<24) /* Same as texld but performs a |
* perspective divide of the texture |
* coordinate .xyz values by .w before |
* sampling. */ |
#define T0_TEXLDB (0x17<<24) /* Same as texld but biases the |
* computed LOD by w. Only S4.6 two's |
* comp is used. This implies that a |
* float to fixed conversion is |
* done. */ |
#define T0_TEXKILL (0x18<<24) /* Does not perform a sampling |
* operation. Simply kills the pixel |
* if any channel of the address |
* register is < 0.0. */ |
#define T0_DEST_TYPE_SHIFT 19 |
/* Allow: R, OC, OD, U */ |
/* Note: U (unpreserved) regs do not retain their values between |
* phases (cannot be used for feedback) |
* |
* Note: oC and OD registers can only be used as the destination of a |
* texture instruction once per phase (this is an implementation |
* restriction). |
*/ |
#define T0_DEST_NR_SHIFT 14 |
/* Allow R: 0..15, OC,OD: 0..0, U: 0..2 */ |
#define T0_SAMPLER_NR_SHIFT 0 /* This field ignored for TEXKILL */ |
#define T0_SAMPLER_NR_MASK (0xf<<0) |
#define T1_ADDRESS_REG_TYPE_SHIFT 24 /* Reg to use as texture coord */ |
/* Allow R, T, OC, OD -- R, OC, OD are 'dependent' reads, new program phase */ |
#define T1_ADDRESS_REG_NR_SHIFT 17 |
#define T2_MBZ 0 |
/* Declaration instructions */ |
#define D0_DCL (0x19<<24) /* Declare a t (interpolated attrib) |
* register or an s (sampler) |
* register. */ |
#define D0_SAMPLE_TYPE_SHIFT 22 |
#define D0_SAMPLE_TYPE_2D (0x0<<22) |
#define D0_SAMPLE_TYPE_CUBE (0x1<<22) |
#define D0_SAMPLE_TYPE_VOLUME (0x2<<22) |
#define D0_SAMPLE_TYPE_MASK (0x3<<22) |
#define D0_TYPE_SHIFT 19 |
/* Allow: T, S */ |
#define D0_NR_SHIFT 14 |
/* Allow T: 0..10, S: 0..15 */ |
#define D0_CHANNEL_X (1<<10) |
#define D0_CHANNEL_Y (2<<10) |
#define D0_CHANNEL_Z (4<<10) |
#define D0_CHANNEL_W (8<<10) |
#define D0_CHANNEL_ALL (0xf<<10) |
#define D0_CHANNEL_NONE (0<<10) |
#define D0_CHANNEL_XY (D0_CHANNEL_X|D0_CHANNEL_Y) |
#define D0_CHANNEL_XYZ (D0_CHANNEL_XY|D0_CHANNEL_Z) |
/* I915 Errata: Do not allow (xz), (xw), (xzw) combinations for diffuse |
* or specular declarations. |
* |
* For T dcls, only allow: (x), (xy), (xyz), (w), (xyzw) |
* |
* Must be zero for S (sampler) dcls |
*/ |
#define D1_MBZ 0 |
#define D2_MBZ 0 |
/* MASK_* are the unshifted bitmasks of the destination mask in arithmetic |
* operations |
*/ |
#define MASK_X 0x1 |
#define MASK_Y 0x2 |
#define MASK_Z 0x4 |
#define MASK_W 0x8 |
#define MASK_XYZ (MASK_X | MASK_Y | MASK_Z) |
#define MASK_XYZW (MASK_XYZ | MASK_W) |
#define MASK_SATURATE 0x10 |
/* Temporary, undeclared regs. Preserved between phases */ |
#define FS_R0 ((REG_TYPE_R << REG_TYPE_SHIFT) | 0) |
#define FS_R1 ((REG_TYPE_R << REG_TYPE_SHIFT) | 1) |
#define FS_R2 ((REG_TYPE_R << REG_TYPE_SHIFT) | 2) |
#define FS_R3 ((REG_TYPE_R << REG_TYPE_SHIFT) | 3) |
/* Texture coordinate regs. Must be declared. */ |
#define FS_T0 ((REG_TYPE_T << REG_TYPE_SHIFT) | 0) |
#define FS_T1 ((REG_TYPE_T << REG_TYPE_SHIFT) | 1) |
#define FS_T2 ((REG_TYPE_T << REG_TYPE_SHIFT) | 2) |
#define FS_T3 ((REG_TYPE_T << REG_TYPE_SHIFT) | 3) |
#define FS_T4 ((REG_TYPE_T << REG_TYPE_SHIFT) | 4) |
#define FS_T5 ((REG_TYPE_T << REG_TYPE_SHIFT) | 5) |
#define FS_T6 ((REG_TYPE_T << REG_TYPE_SHIFT) | 6) |
#define FS_T7 ((REG_TYPE_T << REG_TYPE_SHIFT) | 7) |
#define FS_T8 ((REG_TYPE_T << REG_TYPE_SHIFT) | 8) |
#define FS_T9 ((REG_TYPE_T << REG_TYPE_SHIFT) | 9) |
#define FS_T10 ((REG_TYPE_T << REG_TYPE_SHIFT) | 10) |
/* Constant values */ |
#define FS_C0 ((REG_TYPE_CONST << REG_TYPE_SHIFT) | 0) |
#define FS_C1 ((REG_TYPE_CONST << REG_TYPE_SHIFT) | 1) |
#define FS_C2 ((REG_TYPE_CONST << REG_TYPE_SHIFT) | 2) |
#define FS_C3 ((REG_TYPE_CONST << REG_TYPE_SHIFT) | 3) |
#define FS_C4 ((REG_TYPE_CONST << REG_TYPE_SHIFT) | 4) |
#define FS_C5 ((REG_TYPE_CONST << REG_TYPE_SHIFT) | 5) |
#define FS_C6 ((REG_TYPE_CONST << REG_TYPE_SHIFT) | 6) |
#define FS_C7 ((REG_TYPE_CONST << REG_TYPE_SHIFT) | 7) |
/* Sampler regs */ |
#define FS_S0 ((REG_TYPE_S << REG_TYPE_SHIFT) | 0) |
#define FS_S1 ((REG_TYPE_S << REG_TYPE_SHIFT) | 1) |
#define FS_S2 ((REG_TYPE_S << REG_TYPE_SHIFT) | 2) |
#define FS_S3 ((REG_TYPE_S << REG_TYPE_SHIFT) | 3) |
/* Output color */ |
#define FS_OC ((REG_TYPE_OC << REG_TYPE_SHIFT) | 0) |
/* Output depth */ |
#define FS_OD ((REG_TYPE_OD << REG_TYPE_SHIFT) | 0) |
/* Unpreserved temporary regs */ |
#define FS_U0 ((REG_TYPE_U << REG_TYPE_SHIFT) | 0) |
#define FS_U1 ((REG_TYPE_U << REG_TYPE_SHIFT) | 1) |
#define FS_U2 ((REG_TYPE_U << REG_TYPE_SHIFT) | 2) |
#define FS_U3 ((REG_TYPE_U << REG_TYPE_SHIFT) | 3) |
#define X_CHANNEL_SHIFT (REG_TYPE_SHIFT + 3) |
#define Y_CHANNEL_SHIFT (X_CHANNEL_SHIFT + 4) |
#define Z_CHANNEL_SHIFT (Y_CHANNEL_SHIFT + 4) |
#define W_CHANNEL_SHIFT (Z_CHANNEL_SHIFT + 4) |
#define REG_CHANNEL_MASK 0xf |
#define REG_NR(reg) ((reg) & REG_NR_MASK) |
#define REG_TYPE(reg) (((reg) >> REG_TYPE_SHIFT) & REG_TYPE_MASK) |
#define REG_X(reg) (((reg) >> X_CHANNEL_SHIFT) & REG_CHANNEL_MASK) |
#define REG_Y(reg) (((reg) >> Y_CHANNEL_SHIFT) & REG_CHANNEL_MASK) |
#define REG_Z(reg) (((reg) >> Z_CHANNEL_SHIFT) & REG_CHANNEL_MASK) |
#define REG_W(reg) (((reg) >> W_CHANNEL_SHIFT) & REG_CHANNEL_MASK) |
enum gen3_fs_channel { |
X_CHANNEL_VAL = 0, |
Y_CHANNEL_VAL, |
Z_CHANNEL_VAL, |
W_CHANNEL_VAL, |
ZERO_CHANNEL_VAL, |
ONE_CHANNEL_VAL, |
NEG_X_CHANNEL_VAL = X_CHANNEL_VAL | 0x8, |
NEG_Y_CHANNEL_VAL = Y_CHANNEL_VAL | 0x8, |
NEG_Z_CHANNEL_VAL = Z_CHANNEL_VAL | 0x8, |
NEG_W_CHANNEL_VAL = W_CHANNEL_VAL | 0x8, |
NEG_ONE_CHANNEL_VAL = ONE_CHANNEL_VAL | 0x8 |
}; |
#define gen3_fs_operand(reg, x, y, z, w) \ |
(reg) | \ |
(x##_CHANNEL_VAL << X_CHANNEL_SHIFT) | \ |
(y##_CHANNEL_VAL << Y_CHANNEL_SHIFT) | \ |
(z##_CHANNEL_VAL << Z_CHANNEL_SHIFT) | \ |
(w##_CHANNEL_VAL << W_CHANNEL_SHIFT) |
/** |
* Construct an operand description for using a register with no swizzling |
*/ |
#define gen3_fs_operand_reg(reg) \ |
gen3_fs_operand(reg, X, Y, Z, W) |
#define gen3_fs_operand_reg_negate(reg) \ |
gen3_fs_operand(reg, NEG_X, NEG_Y, NEG_Z, NEG_W) |
/** |
* Returns an operand containing (0.0, 0.0, 0.0, 0.0). |
*/ |
#define gen3_fs_operand_zero() gen3_fs_operand(FS_R0, ZERO, ZERO, ZERO, ZERO) |
/** |
* Returns an unused operand |
*/ |
#define gen3_fs_operand_none() gen3_fs_operand_zero() |
/** |
* Returns an operand containing (1.0, 1.0, 1.0, 1.0). |
*/ |
#define gen3_fs_operand_one() gen3_fs_operand(FS_R0, ONE, ONE, ONE, ONE) |
#define gen3_get_hardware_channel_val(val, shift, negate) \ |
(((val & 0x7) << shift) | ((val & 0x8) ? negate : 0)) |
/** |
* Outputs a fragment shader command to declare a sampler or texture register. |
*/ |
#define gen3_fs_dcl(reg) \ |
do { \ |
OUT_BATCH(D0_DCL | \ |
(REG_TYPE(reg) << D0_TYPE_SHIFT) | \ |
(REG_NR(reg) << D0_NR_SHIFT) | \ |
((REG_TYPE(reg) != REG_TYPE_S) ? D0_CHANNEL_ALL : 0)); \ |
OUT_BATCH(0); \ |
OUT_BATCH(0); \ |
} while (0) |
#define gen3_fs_texld(dest_reg, sampler_reg, address_reg) \ |
do { \ |
OUT_BATCH(T0_TEXLD | \ |
(REG_TYPE(dest_reg) << T0_DEST_TYPE_SHIFT) | \ |
(REG_NR(dest_reg) << T0_DEST_NR_SHIFT) | \ |
(REG_NR(sampler_reg) << T0_SAMPLER_NR_SHIFT)); \ |
OUT_BATCH((REG_TYPE(address_reg) << T1_ADDRESS_REG_TYPE_SHIFT) | \ |
(REG_NR(address_reg) << T1_ADDRESS_REG_NR_SHIFT)); \ |
OUT_BATCH(0); \ |
} while (0) |
#define gen3_fs_texldp(dest_reg, sampler_reg, address_reg) \ |
do { \ |
OUT_BATCH(T0_TEXLDP | \ |
(REG_TYPE(dest_reg) << T0_DEST_TYPE_SHIFT) | \ |
(REG_NR(dest_reg) << T0_DEST_NR_SHIFT) | \ |
(REG_NR(sampler_reg) << T0_SAMPLER_NR_SHIFT)); \ |
OUT_BATCH((REG_TYPE(address_reg) << T1_ADDRESS_REG_TYPE_SHIFT) | \ |
(REG_NR(address_reg) << T1_ADDRESS_REG_NR_SHIFT)); \ |
OUT_BATCH(0); \ |
} while (0) |
#define gen3_fs_arith_masked(op, dest_reg, dest_mask, operand0, operand1, operand2) \ |
_gen3_fs_arith_masked(A0_##op, dest_reg, dest_mask, operand0, operand1, operand2) |
#define gen3_fs_arith(op, dest_reg, operand0, operand1, operand2) \ |
_gen3_fs_arith(A0_##op, dest_reg, operand0, operand1, operand2) |
#define _gen3_fs_arith_masked(cmd, dest_reg, dest_mask, operand0, operand1, operand2) \ |
do { \ |
/* Set up destination register and write mask */ \ |
OUT_BATCH(cmd | \ |
(REG_TYPE(dest_reg) << A0_DEST_TYPE_SHIFT) | \ |
(REG_NR(dest_reg) << A0_DEST_NR_SHIFT) | \ |
(((dest_mask) & ~MASK_SATURATE) << A0_DEST_CHANNEL_SHIFT) | \ |
(((dest_mask) & MASK_SATURATE) ? A0_DEST_SATURATE : 0) | \ |
/* Set up operand 0 */ \ |
(REG_TYPE(operand0) << A0_SRC0_TYPE_SHIFT) | \ |
(REG_NR(operand0) << A0_SRC0_NR_SHIFT)); \ |
OUT_BATCH(gen3_get_hardware_channel_val(REG_X(operand0), \ |
A1_SRC0_CHANNEL_X_SHIFT, \ |
A1_SRC0_CHANNEL_X_NEGATE) | \ |
gen3_get_hardware_channel_val(REG_Y(operand0), \ |
A1_SRC0_CHANNEL_Y_SHIFT, \ |
A1_SRC0_CHANNEL_Y_NEGATE) | \ |
gen3_get_hardware_channel_val(REG_Z(operand0), \ |
A1_SRC0_CHANNEL_Z_SHIFT, \ |
A1_SRC0_CHANNEL_Z_NEGATE) | \ |
gen3_get_hardware_channel_val(REG_W(operand0), \ |
A1_SRC0_CHANNEL_W_SHIFT, \ |
A1_SRC0_CHANNEL_W_NEGATE) | \ |
/* Set up operand 1 */ \ |
(REG_TYPE(operand1) << A1_SRC1_TYPE_SHIFT) | \ |
(REG_NR(operand1) << A1_SRC1_NR_SHIFT) | \ |
gen3_get_hardware_channel_val(REG_X(operand1), \ |
A1_SRC1_CHANNEL_X_SHIFT, \ |
A1_SRC1_CHANNEL_X_NEGATE) | \ |
gen3_get_hardware_channel_val(REG_Y(operand1), \ |
A1_SRC1_CHANNEL_Y_SHIFT, \ |
A1_SRC1_CHANNEL_Y_NEGATE)); \ |
OUT_BATCH(gen3_get_hardware_channel_val(REG_Z(operand1), \ |
A2_SRC1_CHANNEL_Z_SHIFT, \ |
A2_SRC1_CHANNEL_Z_NEGATE) | \ |
gen3_get_hardware_channel_val(REG_W(operand1), \ |
A2_SRC1_CHANNEL_W_SHIFT, \ |
A2_SRC1_CHANNEL_W_NEGATE) | \ |
/* Set up operand 2 */ \ |
(REG_TYPE(operand2) << A2_SRC2_TYPE_SHIFT) | \ |
(REG_NR(operand2) << A2_SRC2_NR_SHIFT) | \ |
gen3_get_hardware_channel_val(REG_X(operand2), \ |
A2_SRC2_CHANNEL_X_SHIFT, \ |
A2_SRC2_CHANNEL_X_NEGATE) | \ |
gen3_get_hardware_channel_val(REG_Y(operand2), \ |
A2_SRC2_CHANNEL_Y_SHIFT, \ |
A2_SRC2_CHANNEL_Y_NEGATE) | \ |
gen3_get_hardware_channel_val(REG_Z(operand2), \ |
A2_SRC2_CHANNEL_Z_SHIFT, \ |
A2_SRC2_CHANNEL_Z_NEGATE) | \ |
gen3_get_hardware_channel_val(REG_W(operand2), \ |
A2_SRC2_CHANNEL_W_SHIFT, \ |
A2_SRC2_CHANNEL_W_NEGATE)); \ |
} while (0) |
#define _gen3_fs_arith(cmd, dest_reg, operand0, operand1, operand2) do {\ |
/* Set up destination register and write mask */ \ |
OUT_BATCH(cmd | \ |
(REG_TYPE(dest_reg) << A0_DEST_TYPE_SHIFT) | \ |
(REG_NR(dest_reg) << A0_DEST_NR_SHIFT) | \ |
(A0_DEST_CHANNEL_ALL) | \ |
/* Set up operand 0 */ \ |
(REG_TYPE(operand0) << A0_SRC0_TYPE_SHIFT) | \ |
(REG_NR(operand0) << A0_SRC0_NR_SHIFT)); \ |
OUT_BATCH(gen3_get_hardware_channel_val(REG_X(operand0), \ |
A1_SRC0_CHANNEL_X_SHIFT, \ |
A1_SRC0_CHANNEL_X_NEGATE) | \ |
gen3_get_hardware_channel_val(REG_Y(operand0), \ |
A1_SRC0_CHANNEL_Y_SHIFT, \ |
A1_SRC0_CHANNEL_Y_NEGATE) | \ |
gen3_get_hardware_channel_val(REG_Z(operand0), \ |
A1_SRC0_CHANNEL_Z_SHIFT, \ |
A1_SRC0_CHANNEL_Z_NEGATE) | \ |
gen3_get_hardware_channel_val(REG_W(operand0), \ |
A1_SRC0_CHANNEL_W_SHIFT, \ |
A1_SRC0_CHANNEL_W_NEGATE) | \ |
/* Set up operand 1 */ \ |
(REG_TYPE(operand1) << A1_SRC1_TYPE_SHIFT) | \ |
(REG_NR(operand1) << A1_SRC1_NR_SHIFT) | \ |
gen3_get_hardware_channel_val(REG_X(operand1), \ |
A1_SRC1_CHANNEL_X_SHIFT, \ |
A1_SRC1_CHANNEL_X_NEGATE) | \ |
gen3_get_hardware_channel_val(REG_Y(operand1), \ |
A1_SRC1_CHANNEL_Y_SHIFT, \ |
A1_SRC1_CHANNEL_Y_NEGATE)); \ |
OUT_BATCH(gen3_get_hardware_channel_val(REG_Z(operand1), \ |
A2_SRC1_CHANNEL_Z_SHIFT, \ |
A2_SRC1_CHANNEL_Z_NEGATE) | \ |
gen3_get_hardware_channel_val(REG_W(operand1), \ |
A2_SRC1_CHANNEL_W_SHIFT, \ |
A2_SRC1_CHANNEL_W_NEGATE) | \ |
/* Set up operand 2 */ \ |
(REG_TYPE(operand2) << A2_SRC2_TYPE_SHIFT) | \ |
(REG_NR(operand2) << A2_SRC2_NR_SHIFT) | \ |
gen3_get_hardware_channel_val(REG_X(operand2), \ |
A2_SRC2_CHANNEL_X_SHIFT, \ |
A2_SRC2_CHANNEL_X_NEGATE) | \ |
gen3_get_hardware_channel_val(REG_Y(operand2), \ |
A2_SRC2_CHANNEL_Y_SHIFT, \ |
A2_SRC2_CHANNEL_Y_NEGATE) | \ |
gen3_get_hardware_channel_val(REG_Z(operand2), \ |
A2_SRC2_CHANNEL_Z_SHIFT, \ |
A2_SRC2_CHANNEL_Z_NEGATE) | \ |
gen3_get_hardware_channel_val(REG_W(operand2), \ |
A2_SRC2_CHANNEL_W_SHIFT, \ |
A2_SRC2_CHANNEL_W_NEGATE)); \ |
} while (0) |
#define gen3_fs_mov(dest_reg, operand0) \ |
gen3_fs_arith(MOV, dest_reg, \ |
operand0, \ |
gen3_fs_operand_none(), \ |
gen3_fs_operand_none()) |
#define gen3_fs_mov_masked(dest_reg, dest_mask, operand0) \ |
gen3_fs_arith_masked (MOV, dest_reg, dest_mask, \ |
operand0, \ |
gen3_fs_operand_none(), \ |
gen3_fs_operand_none()) |
#define gen3_fs_frc(dest_reg, operand0) \ |
gen3_fs_arith (FRC, dest_reg, \ |
operand0, \ |
gen3_fs_operand_none(), \ |
gen3_fs_operand_none()) |
/** Add operand0 and operand1 and put the result in dest_reg */ |
#define gen3_fs_add(dest_reg, operand0, operand1) \ |
gen3_fs_arith (ADD, dest_reg, \ |
operand0, operand1, \ |
gen3_fs_operand_none()) |
/** Multiply operand0 and operand1 and put the result in dest_reg */ |
#define gen3_fs_mul(dest_reg, operand0, operand1) \ |
gen3_fs_arith (MUL, dest_reg, \ |
operand0, operand1, \ |
gen3_fs_operand_none()) |
/** Computes 1/(operand0.replicate_swizzle) puts the result in dest_reg */ |
#define gen3_fs_rcp(dest_reg, dest_mask, operand0) \ |
do { \ |
if (dest_mask) { \ |
gen3_fs_arith_masked (RCP, dest_reg, dest_mask, \ |
operand0, \ |
gen3_fs_operand_none (), \ |
gen3_fs_operand_none ()); \ |
} else { \ |
gen3_fs_arith (RCP, dest_reg, \ |
operand0, \ |
gen3_fs_operand_none (), \ |
gen3_fs_operand_none ()); \ |
} \ |
} while (0) |
/** Computes 1/sqrt(operand0.replicate_swizzle) puts the result in dest_reg */ |
#define gen3_fs_rsq(dest_reg, dest_mask, operand0) \ |
do { \ |
if (dest_mask) { \ |
gen3_fs_arith_masked (RSQ, dest_reg, dest_mask, \ |
operand0, \ |
gen3_fs_operand_none (), \ |
gen3_fs_operand_none ()); \ |
} else { \ |
gen3_fs_arith (RSQ, dest_reg, \ |
operand0, \ |
gen3_fs_operand_none (), \ |
gen3_fs_operand_none ()); \ |
} \ |
} while (0) |
/** Puts the minimum of operand0 and operand1 in dest_reg */ |
#define gen3_fs_min(dest_reg, operand0, operand1) \ |
gen3_fs_arith (MIN, dest_reg, \ |
operand0, operand1, \ |
gen3_fs_operand_none()) |
/** Puts the maximum of operand0 and operand1 in dest_reg */ |
#define gen3_fs_max(dest_reg, operand0, operand1) \ |
gen3_fs_arith (MAX, dest_reg, \ |
operand0, operand1, \ |
gen3_fs_operand_none()) |
#define gen3_fs_cmp(dest_reg, operand0, operand1, operand2) \ |
gen3_fs_arith (CMP, dest_reg, operand0, operand1, operand2) |
/** Perform operand0 * operand1 + operand2 and put the result in dest_reg */ |
#define gen3_fs_mad(dest_reg, dest_mask, op0, op1, op2) \ |
do { \ |
if (dest_mask) { \ |
gen3_fs_arith_masked (MAD, dest_reg, dest_mask, op0, op1, op2); \ |
} else { \ |
gen3_fs_arith (MAD, dest_reg, op0, op1, op2); \ |
} \ |
} while (0) |
#define gen3_fs_dp2add(dest_reg, dest_mask, op0, op1, op2) \ |
do { \ |
if (dest_mask) { \ |
gen3_fs_arith_masked (DP2ADD, dest_reg, dest_mask, op0, op1, op2); \ |
} else { \ |
gen3_fs_arith (DP2ADD, dest_reg, op0, op1, op2); \ |
} \ |
} while (0) |
/** |
* Perform a 3-component dot-product of operand0 and operand1 and put the |
* resulting scalar in the channels of dest_reg specified by the dest_mask. |
*/ |
#define gen3_fs_dp3(dest_reg, dest_mask, op0, op1) \ |
do { \ |
if (dest_mask) { \ |
gen3_fs_arith_masked (DP3, dest_reg, dest_mask, \ |
op0, op1,\ |
gen3_fs_operand_none()); \ |
} else { \ |
gen3_fs_arith (DP3, dest_reg, op0, op1,\ |
gen3_fs_operand_none()); \ |
} \ |
} while (0) |
/** |
* Perform a 4-component dot-product of operand0 and operand1 and put the |
* resulting scalar in the channels of dest_reg specified by the dest_mask. |
*/ |
#define gen3_fs_dp4(dest_reg, dest_mask, op0, op1) \ |
do { \ |
if (dest_mask) { \ |
gen3_fs_arith_masked (DP4, dest_reg, dest_mask, \ |
op0, op1,\ |
gen3_fs_operand_none()); \ |
} else { \ |
gen3_fs_arith (DP4, dest_reg, op0, op1,\ |
gen3_fs_operand_none()); \ |
} \ |
} while (0) |
#define SHADER_TRAPEZOIDS (1 << 24) |
/drivers/video/Intel-2D/sna/gen4_render.c |
---|
0,0 → 1,2414 |
/* |
* Copyright © 2006,2008,2011 Intel Corporation |
* Copyright © 2007 Red Hat, Inc. |
* |
* Permission is hereby granted, free of charge, to any person obtaining a |
* copy of this software and associated documentation files (the "Software"), |
* to deal in the Software without restriction, including without limitation |
* the rights to use, copy, modify, merge, publish, distribute, sublicense, |
* and/or sell copies of the Software, and to permit persons to whom the |
* Software is furnished to do so, subject to the following conditions: |
* |
* The above copyright notice and this permission notice (including the next |
* paragraph) shall be included in all copies or substantial portions of the |
* Software. |
* |
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL |
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, |
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE |
* SOFTWARE. |
* |
* Authors: |
* Wang Zhenyu <zhenyu.z.wang@sna.com> |
* Eric Anholt <eric@anholt.net> |
* Carl Worth <cworth@redhat.com> |
* Keith Packard <keithp@keithp.com> |
* Chris Wilson <chris@chris-wilson.co.uk> |
* |
*/ |
#ifdef HAVE_CONFIG_H |
#include "config.h" |
#endif |
#include "sna.h" |
#include "sna_reg.h" |
#include "sna_render.h" |
#include "sna_render_inline.h" |
//#include "sna_video.h" |
#include "brw/brw.h" |
#include "gen4_render.h" |
#include "gen4_source.h" |
#include "gen4_vertex.h" |
/* gen4 has a serious issue with its shaders that we need to flush |
* after every rectangle... So until that is resolved, prefer |
* the BLT engine. |
*/ |
#define FORCE_SPANS 0 |
#define FORCE_NONRECTILINEAR_SPANS -1 |
#define FORCE_FLUSH 1 /* https://bugs.freedesktop.org/show_bug.cgi?id=55500 */ |
#define NO_COMPOSITE 0 |
#define NO_COMPOSITE_SPANS 0 |
#define NO_COPY 0 |
#define NO_COPY_BOXES 0 |
#define NO_FILL 0 |
#define NO_FILL_ONE 0 |
#define NO_FILL_BOXES 0 |
#define NO_VIDEO 0 |
#define MAX_FLUSH_VERTICES 6 |
#define GEN4_GRF_BLOCKS(nreg) ((nreg + 15) / 16 - 1) |
/* Set up a default static partitioning of the URB, which is supposed to |
* allow anything we would want to do, at potentially lower performance. |
*/ |
#define URB_CS_ENTRY_SIZE 1 |
#define URB_CS_ENTRIES 0 |
#define URB_VS_ENTRY_SIZE 1 |
#define URB_VS_ENTRIES 32 |
#define URB_GS_ENTRY_SIZE 0 |
#define URB_GS_ENTRIES 0 |
#define URB_CLIP_ENTRY_SIZE 0 |
#define URB_CLIP_ENTRIES 0 |
#define URB_SF_ENTRY_SIZE 2 |
#define URB_SF_ENTRIES 64 |
/* |
* this program computes dA/dx and dA/dy for the texture coordinates along |
* with the base texture coordinate. It was extracted from the Mesa driver |
*/ |
#define SF_KERNEL_NUM_GRF 16 |
#define PS_KERNEL_NUM_GRF 32 |
#define GEN4_MAX_SF_THREADS 24 |
#define GEN4_MAX_WM_THREADS 32 |
#define G4X_MAX_WM_THREADS 50 |
static const uint32_t ps_kernel_packed_static[][4] = { |
#include "exa_wm_xy.g4b" |
#include "exa_wm_src_affine.g4b" |
#include "exa_wm_src_sample_argb.g4b" |
#include "exa_wm_yuv_rgb.g4b" |
#include "exa_wm_write.g4b" |
}; |
static const uint32_t ps_kernel_planar_static[][4] = { |
#include "exa_wm_xy.g4b" |
#include "exa_wm_src_affine.g4b" |
#include "exa_wm_src_sample_planar.g4b" |
#include "exa_wm_yuv_rgb.g4b" |
#include "exa_wm_write.g4b" |
}; |
#define NOKERNEL(kernel_enum, func, masked) \ |
[kernel_enum] = {func, 0, masked} |
#define KERNEL(kernel_enum, kernel, masked) \ |
[kernel_enum] = {&kernel, sizeof(kernel), masked} |
static const struct wm_kernel_info { |
const void *data; |
unsigned int size; |
bool has_mask; |
} wm_kernels[] = { |
NOKERNEL(WM_KERNEL, brw_wm_kernel__affine, false), |
NOKERNEL(WM_KERNEL_P, brw_wm_kernel__projective, false), |
NOKERNEL(WM_KERNEL_MASK, brw_wm_kernel__affine_mask, true), |
NOKERNEL(WM_KERNEL_MASK_P, brw_wm_kernel__projective_mask, true), |
NOKERNEL(WM_KERNEL_MASKCA, brw_wm_kernel__affine_mask_ca, true), |
NOKERNEL(WM_KERNEL_MASKCA_P, brw_wm_kernel__projective_mask_ca, true), |
NOKERNEL(WM_KERNEL_MASKSA, brw_wm_kernel__affine_mask_sa, true), |
NOKERNEL(WM_KERNEL_MASKSA_P, brw_wm_kernel__projective_mask_sa, true), |
NOKERNEL(WM_KERNEL_OPACITY, brw_wm_kernel__affine_opacity, true), |
NOKERNEL(WM_KERNEL_OPACITY_P, brw_wm_kernel__projective_opacity, true), |
KERNEL(WM_KERNEL_VIDEO_PLANAR, ps_kernel_planar_static, false), |
KERNEL(WM_KERNEL_VIDEO_PACKED, ps_kernel_packed_static, false), |
}; |
#undef KERNEL |
static const struct blendinfo { |
bool src_alpha; |
uint32_t src_blend; |
uint32_t dst_blend; |
} gen4_blend_op[] = { |
/* Clear */ {0, GEN4_BLENDFACTOR_ZERO, GEN4_BLENDFACTOR_ZERO}, |
/* Src */ {0, GEN4_BLENDFACTOR_ONE, GEN4_BLENDFACTOR_ZERO}, |
/* Dst */ {0, GEN4_BLENDFACTOR_ZERO, GEN4_BLENDFACTOR_ONE}, |
/* Over */ {1, GEN4_BLENDFACTOR_ONE, GEN4_BLENDFACTOR_INV_SRC_ALPHA}, |
/* OverReverse */ {0, GEN4_BLENDFACTOR_INV_DST_ALPHA, GEN4_BLENDFACTOR_ONE}, |
/* In */ {0, GEN4_BLENDFACTOR_DST_ALPHA, GEN4_BLENDFACTOR_ZERO}, |
/* InReverse */ {1, GEN4_BLENDFACTOR_ZERO, GEN4_BLENDFACTOR_SRC_ALPHA}, |
/* Out */ {0, GEN4_BLENDFACTOR_INV_DST_ALPHA, GEN4_BLENDFACTOR_ZERO}, |
/* OutReverse */ {1, GEN4_BLENDFACTOR_ZERO, GEN4_BLENDFACTOR_INV_SRC_ALPHA}, |
/* Atop */ {1, GEN4_BLENDFACTOR_DST_ALPHA, GEN4_BLENDFACTOR_INV_SRC_ALPHA}, |
/* AtopReverse */ {1, GEN4_BLENDFACTOR_INV_DST_ALPHA, GEN4_BLENDFACTOR_SRC_ALPHA}, |
/* Xor */ {1, GEN4_BLENDFACTOR_INV_DST_ALPHA, GEN4_BLENDFACTOR_INV_SRC_ALPHA}, |
/* Add */ {0, GEN4_BLENDFACTOR_ONE, GEN4_BLENDFACTOR_ONE}, |
}; |
/** |
* Highest-valued BLENDFACTOR used in gen4_blend_op. |
* |
* This leaves out GEN4_BLENDFACTOR_INV_DST_COLOR, |
* GEN4_BLENDFACTOR_INV_CONST_{COLOR,ALPHA}, |
* GEN4_BLENDFACTOR_INV_SRC1_{COLOR,ALPHA} |
*/ |
#define GEN4_BLENDFACTOR_COUNT (GEN4_BLENDFACTOR_INV_DST_ALPHA + 1) |
#define BLEND_OFFSET(s, d) \ |
(((s) * GEN4_BLENDFACTOR_COUNT + (d)) * 64) |
#define SAMPLER_OFFSET(sf, se, mf, me, k) \ |
((((((sf) * EXTEND_COUNT + (se)) * FILTER_COUNT + (mf)) * EXTEND_COUNT + (me)) * KERNEL_COUNT + (k)) * 64) |
static void |
gen4_emit_pipelined_pointers(struct sna *sna, |
const struct sna_composite_op *op, |
int blend, int kernel); |
#define OUT_BATCH(v) batch_emit(sna, v) |
#define OUT_VERTEX(x,y) vertex_emit_2s(sna, x,y) |
#define OUT_VERTEX_F(v) vertex_emit(sna, v) |
#define GEN4_MAX_3D_SIZE 8192 |
static inline bool too_large(int width, int height) |
{ |
return width > GEN4_MAX_3D_SIZE || height > GEN4_MAX_3D_SIZE; |
} |
static int |
gen4_choose_composite_kernel(int op, bool has_mask, bool is_ca, bool is_affine) |
{ |
int base; |
if (has_mask) { |
if (is_ca) { |
if (gen4_blend_op[op].src_alpha) |
base = WM_KERNEL_MASKSA; |
else |
base = WM_KERNEL_MASKCA; |
} else |
base = WM_KERNEL_MASK; |
} else |
base = WM_KERNEL; |
return base + !is_affine; |
} |
static bool gen4_magic_ca_pass(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
struct gen4_render_state *state = &sna->render_state.gen4; |
if (!op->need_magic_ca_pass) |
return false; |
assert(sna->render.vertex_index > sna->render.vertex_start); |
DBG(("%s: CA fixup\n", __FUNCTION__)); |
assert(op->mask.bo != NULL); |
assert(op->has_component_alpha); |
gen4_emit_pipelined_pointers(sna, op, PictOpAdd, |
gen4_choose_composite_kernel(PictOpAdd, |
true, true, op->is_affine)); |
OUT_BATCH(GEN4_3DPRIMITIVE | |
GEN4_3DPRIMITIVE_VERTEX_SEQUENTIAL | |
(_3DPRIM_RECTLIST << GEN4_3DPRIMITIVE_TOPOLOGY_SHIFT) | |
(0 << 9) | |
4); |
OUT_BATCH(sna->render.vertex_index - sna->render.vertex_start); |
OUT_BATCH(sna->render.vertex_start); |
OUT_BATCH(1); /* single instance */ |
OUT_BATCH(0); /* start instance location */ |
OUT_BATCH(0); /* index buffer offset, ignored */ |
state->last_primitive = sna->kgem.nbatch; |
return true; |
} |
static uint32_t gen4_get_blend(int op, |
bool has_component_alpha, |
uint32_t dst_format) |
{ |
uint32_t src, dst; |
src = GEN4_BLENDFACTOR_ONE; //gen4_blend_op[op].src_blend; |
dst = GEN4_BLENDFACTOR_INV_SRC_ALPHA; //gen6_blend_op[op].dst_blend; |
#if 0 |
/* If there's no dst alpha channel, adjust the blend op so that we'll treat |
* it as always 1. |
*/ |
if (PICT_FORMAT_A(dst_format) == 0) { |
if (src == GEN4_BLENDFACTOR_DST_ALPHA) |
src = GEN4_BLENDFACTOR_ONE; |
else if (src == GEN4_BLENDFACTOR_INV_DST_ALPHA) |
src = GEN4_BLENDFACTOR_ZERO; |
} |
/* If the source alpha is being used, then we should only be in a |
* case where the source blend factor is 0, and the source blend |
* value is the mask channels multiplied by the source picture's alpha. |
*/ |
if (has_component_alpha && gen4_blend_op[op].src_alpha) { |
if (dst == GEN4_BLENDFACTOR_SRC_ALPHA) |
dst = GEN4_BLENDFACTOR_SRC_COLOR; |
else if (dst == GEN4_BLENDFACTOR_INV_SRC_ALPHA) |
dst = GEN4_BLENDFACTOR_INV_SRC_COLOR; |
} |
#endif |
DBG(("blend op=%d, dst=%x [A=%d] => src=%d, dst=%d => offset=%x\n", |
op, dst_format, PICT_FORMAT_A(dst_format), |
src, dst, BLEND_OFFSET(src, dst))); |
return BLEND_OFFSET(src, dst); |
} |
static uint32_t gen4_get_card_format(PictFormat format) |
{ |
switch (format) { |
default: |
return -1; |
case PICT_a8r8g8b8: |
return GEN4_SURFACEFORMAT_B8G8R8A8_UNORM; |
case PICT_x8r8g8b8: |
return GEN4_SURFACEFORMAT_B8G8R8X8_UNORM; |
case PICT_a8b8g8r8: |
return GEN4_SURFACEFORMAT_R8G8B8A8_UNORM; |
case PICT_x8b8g8r8: |
return GEN4_SURFACEFORMAT_R8G8B8X8_UNORM; |
case PICT_a2r10g10b10: |
return GEN4_SURFACEFORMAT_B10G10R10A2_UNORM; |
case PICT_x2r10g10b10: |
return GEN4_SURFACEFORMAT_B10G10R10X2_UNORM; |
case PICT_r8g8b8: |
return GEN4_SURFACEFORMAT_R8G8B8_UNORM; |
case PICT_r5g6b5: |
return GEN4_SURFACEFORMAT_B5G6R5_UNORM; |
case PICT_a1r5g5b5: |
return GEN4_SURFACEFORMAT_B5G5R5A1_UNORM; |
case PICT_a8: |
return GEN4_SURFACEFORMAT_A8_UNORM; |
case PICT_a4r4g4b4: |
return GEN4_SURFACEFORMAT_B4G4R4A4_UNORM; |
} |
} |
static uint32_t gen4_get_dest_format(PictFormat format) |
{ |
switch (format) { |
default: |
return -1; |
case PICT_a8r8g8b8: |
case PICT_x8r8g8b8: |
return GEN4_SURFACEFORMAT_B8G8R8A8_UNORM; |
case PICT_a8b8g8r8: |
case PICT_x8b8g8r8: |
return GEN4_SURFACEFORMAT_R8G8B8A8_UNORM; |
case PICT_a2r10g10b10: |
case PICT_x2r10g10b10: |
return GEN4_SURFACEFORMAT_B10G10R10A2_UNORM; |
case PICT_r5g6b5: |
return GEN4_SURFACEFORMAT_B5G6R5_UNORM; |
case PICT_x1r5g5b5: |
case PICT_a1r5g5b5: |
return GEN4_SURFACEFORMAT_B5G5R5A1_UNORM; |
case PICT_a8: |
return GEN4_SURFACEFORMAT_A8_UNORM; |
case PICT_a4r4g4b4: |
case PICT_x4r4g4b4: |
return GEN4_SURFACEFORMAT_B4G4R4A4_UNORM; |
} |
} |
typedef struct gen4_surface_state_padded { |
struct gen4_surface_state state; |
char pad[32 - sizeof(struct gen4_surface_state)]; |
} gen4_surface_state_padded; |
static void null_create(struct sna_static_stream *stream) |
{ |
/* A bunch of zeros useful for legacy border color and depth-stencil */ |
sna_static_stream_map(stream, 64, 64); |
} |
static void |
sampler_state_init(struct gen4_sampler_state *sampler_state, |
sampler_filter_t filter, |
sampler_extend_t extend) |
{ |
sampler_state->ss0.lod_preclamp = 1; /* GL mode */ |
/* We use the legacy mode to get the semantics specified by |
* the Render extension. */ |
sampler_state->ss0.border_color_mode = GEN4_BORDER_COLOR_MODE_LEGACY; |
switch (filter) { |
default: |
case SAMPLER_FILTER_NEAREST: |
sampler_state->ss0.min_filter = GEN4_MAPFILTER_NEAREST; |
sampler_state->ss0.mag_filter = GEN4_MAPFILTER_NEAREST; |
break; |
case SAMPLER_FILTER_BILINEAR: |
sampler_state->ss0.min_filter = GEN4_MAPFILTER_LINEAR; |
sampler_state->ss0.mag_filter = GEN4_MAPFILTER_LINEAR; |
break; |
} |
switch (extend) { |
default: |
case SAMPLER_EXTEND_NONE: |
sampler_state->ss1.r_wrap_mode = GEN4_TEXCOORDMODE_CLAMP_BORDER; |
sampler_state->ss1.s_wrap_mode = GEN4_TEXCOORDMODE_CLAMP_BORDER; |
sampler_state->ss1.t_wrap_mode = GEN4_TEXCOORDMODE_CLAMP_BORDER; |
break; |
case SAMPLER_EXTEND_REPEAT: |
sampler_state->ss1.r_wrap_mode = GEN4_TEXCOORDMODE_WRAP; |
sampler_state->ss1.s_wrap_mode = GEN4_TEXCOORDMODE_WRAP; |
sampler_state->ss1.t_wrap_mode = GEN4_TEXCOORDMODE_WRAP; |
break; |
case SAMPLER_EXTEND_PAD: |
sampler_state->ss1.r_wrap_mode = GEN4_TEXCOORDMODE_CLAMP; |
sampler_state->ss1.s_wrap_mode = GEN4_TEXCOORDMODE_CLAMP; |
sampler_state->ss1.t_wrap_mode = GEN4_TEXCOORDMODE_CLAMP; |
break; |
case SAMPLER_EXTEND_REFLECT: |
sampler_state->ss1.r_wrap_mode = GEN4_TEXCOORDMODE_MIRROR; |
sampler_state->ss1.s_wrap_mode = GEN4_TEXCOORDMODE_MIRROR; |
sampler_state->ss1.t_wrap_mode = GEN4_TEXCOORDMODE_MIRROR; |
break; |
} |
} |
static uint32_t |
gen4_tiling_bits(uint32_t tiling) |
{ |
switch (tiling) { |
default: assert(0); |
case I915_TILING_NONE: return 0; |
case I915_TILING_X: return GEN4_SURFACE_TILED; |
case I915_TILING_Y: return GEN4_SURFACE_TILED | GEN4_SURFACE_TILED_Y; |
} |
} |
/** |
* Sets up the common fields for a surface state buffer for the given |
* picture in the given surface state buffer. |
*/ |
static uint32_t |
gen4_bind_bo(struct sna *sna, |
struct kgem_bo *bo, |
uint32_t width, |
uint32_t height, |
uint32_t format, |
bool is_dst) |
{ |
uint32_t domains; |
uint16_t offset; |
uint32_t *ss; |
assert(sna->kgem.gen != 040 || !kgem_bo_is_snoop(bo)); |
/* After the first bind, we manage the cache domains within the batch */ |
offset = kgem_bo_get_binding(bo, format | is_dst << 31); |
if (offset) { |
if (is_dst) |
kgem_bo_mark_dirty(bo); |
return offset * sizeof(uint32_t); |
} |
offset = sna->kgem.surface -= |
sizeof(struct gen4_surface_state_padded) / sizeof(uint32_t); |
ss = sna->kgem.batch + offset; |
ss[0] = (GEN4_SURFACE_2D << GEN4_SURFACE_TYPE_SHIFT | |
GEN4_SURFACE_BLEND_ENABLED | |
format << GEN4_SURFACE_FORMAT_SHIFT); |
if (is_dst) { |
ss[0] |= GEN4_SURFACE_RC_READ_WRITE; |
domains = I915_GEM_DOMAIN_RENDER << 16 | I915_GEM_DOMAIN_RENDER; |
} else |
domains = I915_GEM_DOMAIN_SAMPLER << 16; |
ss[1] = kgem_add_reloc(&sna->kgem, offset + 1, bo, domains, 0); |
ss[2] = ((width - 1) << GEN4_SURFACE_WIDTH_SHIFT | |
(height - 1) << GEN4_SURFACE_HEIGHT_SHIFT); |
ss[3] = (gen4_tiling_bits(bo->tiling) | |
(bo->pitch - 1) << GEN4_SURFACE_PITCH_SHIFT); |
ss[4] = 0; |
ss[5] = 0; |
kgem_bo_set_binding(bo, format | is_dst << 31, offset); |
DBG(("[%x] bind bo(handle=%d, addr=%d), format=%d, width=%d, height=%d, pitch=%d, tiling=%d -> %s\n", |
offset, bo->handle, ss[1], |
format, width, height, bo->pitch, bo->tiling, |
domains & 0xffff ? "render" : "sampler")); |
return offset * sizeof(uint32_t); |
} |
static void gen4_emit_vertex_buffer(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
int id = op->u.gen4.ve_id; |
assert((sna->render.vb_id & (1 << id)) == 0); |
OUT_BATCH(GEN4_3DSTATE_VERTEX_BUFFERS | 3); |
OUT_BATCH((id << VB0_BUFFER_INDEX_SHIFT) | VB0_VERTEXDATA | |
(4*op->floats_per_vertex << VB0_BUFFER_PITCH_SHIFT)); |
assert(sna->render.nvertex_reloc < ARRAY_SIZE(sna->render.vertex_reloc)); |
sna->render.vertex_reloc[sna->render.nvertex_reloc++] = sna->kgem.nbatch; |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
sna->render.vb_id |= 1 << id; |
} |
static void gen4_emit_primitive(struct sna *sna) |
{ |
if (sna->kgem.nbatch == sna->render_state.gen4.last_primitive) { |
sna->render.vertex_offset = sna->kgem.nbatch - 5; |
return; |
} |
OUT_BATCH(GEN4_3DPRIMITIVE | |
GEN4_3DPRIMITIVE_VERTEX_SEQUENTIAL | |
(_3DPRIM_RECTLIST << GEN4_3DPRIMITIVE_TOPOLOGY_SHIFT) | |
(0 << 9) | |
4); |
sna->render.vertex_offset = sna->kgem.nbatch; |
OUT_BATCH(0); /* vertex count, to be filled in later */ |
OUT_BATCH(sna->render.vertex_index); |
OUT_BATCH(1); /* single instance */ |
OUT_BATCH(0); /* start instance location */ |
OUT_BATCH(0); /* index buffer offset, ignored */ |
sna->render.vertex_start = sna->render.vertex_index; |
sna->render_state.gen4.last_primitive = sna->kgem.nbatch; |
} |
static bool gen4_rectangle_begin(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
unsigned int id = 1 << op->u.gen4.ve_id; |
int ndwords; |
if (sna_vertex_wait__locked(&sna->render) && sna->render.vertex_offset) |
return true; |
/* 7xpipelined pointers + 6xprimitive + 1xflush */ |
ndwords = op->need_magic_ca_pass? 20 : 6; |
if ((sna->render.vb_id & id) == 0) |
ndwords += 5; |
ndwords += 2*FORCE_FLUSH; |
if (!kgem_check_batch(&sna->kgem, ndwords)) |
return false; |
if ((sna->render.vb_id & id) == 0) |
gen4_emit_vertex_buffer(sna, op); |
if (sna->render.vertex_offset == 0) |
gen4_emit_primitive(sna); |
return true; |
} |
static int gen4_get_rectangles__flush(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
/* Preventing discarding new vbo after lock contention */ |
if (sna_vertex_wait__locked(&sna->render)) { |
int rem = vertex_space(sna); |
if (rem > op->floats_per_rect) |
return rem; |
} |
if (!kgem_check_batch(&sna->kgem, |
2*FORCE_FLUSH + (op->need_magic_ca_pass ? 25 : 6))) |
return 0; |
if (!kgem_check_reloc_and_exec(&sna->kgem, 2)) |
return 0; |
if (op->need_magic_ca_pass && sna->render.vbo) |
return 0; |
if (sna->render.vertex_offset) { |
gen4_vertex_flush(sna); |
if (gen4_magic_ca_pass(sna, op)) |
gen4_emit_pipelined_pointers(sna, op, op->op, |
op->u.gen4.wm_kernel); |
} |
return gen4_vertex_finish(sna); |
} |
inline static int gen4_get_rectangles(struct sna *sna, |
const struct sna_composite_op *op, |
int want, |
void (*emit_state)(struct sna *sna, const struct sna_composite_op *op)) |
{ |
int rem; |
assert(want); |
#if FORCE_FLUSH |
rem = sna->render.vertex_offset; |
if (sna->kgem.nbatch == sna->render_state.gen4.last_primitive) |
rem = sna->kgem.nbatch - 5; |
if (rem) { |
rem = MAX_FLUSH_VERTICES - (sna->render.vertex_index - sna->render.vertex_start) / 3; |
if (rem <= 0) { |
if (sna->render.vertex_offset) { |
gen4_vertex_flush(sna); |
if (gen4_magic_ca_pass(sna, op)) |
gen4_emit_pipelined_pointers(sna, op, op->op, |
op->u.gen4.wm_kernel); |
} |
OUT_BATCH(MI_FLUSH | MI_INHIBIT_RENDER_CACHE_FLUSH); |
rem = MAX_FLUSH_VERTICES; |
} |
} else |
rem = MAX_FLUSH_VERTICES; |
if (want > rem) |
want = rem; |
#endif |
start: |
rem = vertex_space(sna); |
if (unlikely(rem < op->floats_per_rect)) { |
DBG(("flushing vbo for %s: %d < %d\n", |
__FUNCTION__, rem, op->floats_per_rect)); |
rem = gen4_get_rectangles__flush(sna, op); |
if (unlikely(rem == 0)) |
goto flush; |
} |
if (unlikely(sna->render.vertex_offset == 0)) { |
if (!gen4_rectangle_begin(sna, op)) |
goto flush; |
else |
goto start; |
} |
assert(rem <= vertex_space(sna)); |
assert(op->floats_per_rect <= rem); |
if (want > 1 && want * op->floats_per_rect > rem) |
want = rem / op->floats_per_rect; |
sna->render.vertex_index += 3*want; |
return want; |
flush: |
if (sna->render.vertex_offset) { |
gen4_vertex_flush(sna); |
gen4_magic_ca_pass(sna, op); |
} |
sna_vertex_wait__locked(&sna->render); |
_kgem_submit(&sna->kgem); |
emit_state(sna, op); |
goto start; |
} |
static uint32_t * |
gen4_composite_get_binding_table(struct sna *sna, uint16_t *offset) |
{ |
sna->kgem.surface -= |
sizeof(struct gen4_surface_state_padded) / sizeof(uint32_t); |
DBG(("%s(%x)\n", __FUNCTION__, 4*sna->kgem.surface)); |
/* Clear all surplus entries to zero in case of prefetch */ |
*offset = sna->kgem.surface; |
return memset(sna->kgem.batch + sna->kgem.surface, |
0, sizeof(struct gen4_surface_state_padded)); |
} |
static void |
gen4_emit_urb(struct sna *sna) |
{ |
int urb_vs_start, urb_vs_size; |
int urb_gs_start, urb_gs_size; |
int urb_clip_start, urb_clip_size; |
int urb_sf_start, urb_sf_size; |
int urb_cs_start, urb_cs_size; |
if (!sna->render_state.gen4.needs_urb) |
return; |
urb_vs_start = 0; |
urb_vs_size = URB_VS_ENTRIES * URB_VS_ENTRY_SIZE; |
urb_gs_start = urb_vs_start + urb_vs_size; |
urb_gs_size = URB_GS_ENTRIES * URB_GS_ENTRY_SIZE; |
urb_clip_start = urb_gs_start + urb_gs_size; |
urb_clip_size = URB_CLIP_ENTRIES * URB_CLIP_ENTRY_SIZE; |
urb_sf_start = urb_clip_start + urb_clip_size; |
urb_sf_size = URB_SF_ENTRIES * URB_SF_ENTRY_SIZE; |
urb_cs_start = urb_sf_start + urb_sf_size; |
urb_cs_size = URB_CS_ENTRIES * URB_CS_ENTRY_SIZE; |
while ((sna->kgem.nbatch & 15) > 12) |
OUT_BATCH(MI_NOOP); |
OUT_BATCH(GEN4_URB_FENCE | |
UF0_CS_REALLOC | |
UF0_SF_REALLOC | |
UF0_CLIP_REALLOC | |
UF0_GS_REALLOC | |
UF0_VS_REALLOC | |
1); |
OUT_BATCH(((urb_clip_start + urb_clip_size) << UF1_CLIP_FENCE_SHIFT) | |
((urb_gs_start + urb_gs_size) << UF1_GS_FENCE_SHIFT) | |
((urb_vs_start + urb_vs_size) << UF1_VS_FENCE_SHIFT)); |
OUT_BATCH(((urb_cs_start + urb_cs_size) << UF2_CS_FENCE_SHIFT) | |
((urb_sf_start + urb_sf_size) << UF2_SF_FENCE_SHIFT)); |
/* Constant buffer state */ |
OUT_BATCH(GEN4_CS_URB_STATE | 0); |
OUT_BATCH((URB_CS_ENTRY_SIZE - 1) << 4 | URB_CS_ENTRIES << 0); |
sna->render_state.gen4.needs_urb = false; |
} |
static void |
gen4_emit_state_base_address(struct sna *sna) |
{ |
assert(sna->render_state.gen4.general_bo->proxy == NULL); |
OUT_BATCH(GEN4_STATE_BASE_ADDRESS | 4); |
OUT_BATCH(kgem_add_reloc(&sna->kgem, /* general */ |
sna->kgem.nbatch, |
sna->render_state.gen4.general_bo, |
I915_GEM_DOMAIN_INSTRUCTION << 16, |
BASE_ADDRESS_MODIFY)); |
OUT_BATCH(kgem_add_reloc(&sna->kgem, /* surface */ |
sna->kgem.nbatch, |
NULL, |
I915_GEM_DOMAIN_INSTRUCTION << 16, |
BASE_ADDRESS_MODIFY)); |
OUT_BATCH(0); /* media */ |
/* upper bounds, all disabled */ |
OUT_BATCH(BASE_ADDRESS_MODIFY); |
OUT_BATCH(0); |
} |
static void |
gen4_emit_invariant(struct sna *sna) |
{ |
assert(sna->kgem.surface == sna->kgem.batch_size); |
if (sna->kgem.gen >= 045) |
OUT_BATCH(NEW_PIPELINE_SELECT | PIPELINE_SELECT_3D); |
else |
OUT_BATCH(GEN4_PIPELINE_SELECT | PIPELINE_SELECT_3D); |
gen4_emit_state_base_address(sna); |
sna->render_state.gen4.needs_invariant = false; |
} |
static void |
gen4_get_batch(struct sna *sna, const struct sna_composite_op *op) |
{ |
kgem_set_mode(&sna->kgem, KGEM_RENDER, op->dst.bo); |
if (!kgem_check_batch_with_surfaces(&sna->kgem, 150 + 50*FORCE_FLUSH, 4)) { |
DBG(("%s: flushing batch: %d < %d+%d\n", |
__FUNCTION__, sna->kgem.surface - sna->kgem.nbatch, |
150, 4*8)); |
kgem_submit(&sna->kgem); |
_kgem_set_mode(&sna->kgem, KGEM_RENDER); |
} |
if (sna->render_state.gen4.needs_invariant) |
gen4_emit_invariant(sna); |
} |
static void |
gen4_align_vertex(struct sna *sna, const struct sna_composite_op *op) |
{ |
assert(op->floats_per_rect == 3*op->floats_per_vertex); |
if (op->floats_per_vertex != sna->render_state.gen4.floats_per_vertex) { |
if (sna->render.vertex_size - sna->render.vertex_used < 2*op->floats_per_rect) |
gen4_vertex_finish(sna); |
DBG(("aligning vertex: was %d, now %d floats per vertex, %d->%d\n", |
sna->render_state.gen4.floats_per_vertex, |
op->floats_per_vertex, |
sna->render.vertex_index, |
(sna->render.vertex_used + op->floats_per_vertex - 1) / op->floats_per_vertex)); |
sna->render.vertex_index = (sna->render.vertex_used + op->floats_per_vertex - 1) / op->floats_per_vertex; |
sna->render.vertex_used = sna->render.vertex_index * op->floats_per_vertex; |
sna->render_state.gen4.floats_per_vertex = op->floats_per_vertex; |
} |
} |
static void |
gen4_emit_binding_table(struct sna *sna, uint16_t offset) |
{ |
if (sna->render_state.gen4.surface_table == offset) |
return; |
sna->render_state.gen4.surface_table = offset; |
/* Binding table pointers */ |
OUT_BATCH(GEN4_3DSTATE_BINDING_TABLE_POINTERS | 4); |
OUT_BATCH(0); /* vs */ |
OUT_BATCH(0); /* gs */ |
OUT_BATCH(0); /* clip */ |
OUT_BATCH(0); /* sf */ |
/* Only the PS uses the binding table */ |
OUT_BATCH(offset*4); |
} |
static void |
gen4_emit_pipelined_pointers(struct sna *sna, |
const struct sna_composite_op *op, |
int blend, int kernel) |
{ |
uint16_t sp, bp; |
uint32_t key; |
DBG(("%s: has_mask=%d, src=(%d, %d), mask=(%d, %d),kernel=%d, blend=%d, ca=%d, format=%x\n", |
__FUNCTION__, op->u.gen4.ve_id & 2, |
op->src.filter, op->src.repeat, |
op->mask.filter, op->mask.repeat, |
kernel, blend, op->has_component_alpha, (int)op->dst.format)); |
sp = SAMPLER_OFFSET(op->src.filter, op->src.repeat, |
op->mask.filter, op->mask.repeat, |
kernel); |
bp = gen4_get_blend(blend, op->has_component_alpha, op->dst.format); |
DBG(("%s: sp=%d, bp=%d\n", __FUNCTION__, sp, bp)); |
key = sp | (uint32_t)bp << 16; |
if (key == sna->render_state.gen4.last_pipelined_pointers) |
return; |
OUT_BATCH(GEN4_3DSTATE_PIPELINED_POINTERS | 5); |
OUT_BATCH(sna->render_state.gen4.vs); |
OUT_BATCH(GEN4_GS_DISABLE); /* passthrough */ |
OUT_BATCH(GEN4_CLIP_DISABLE); /* passthrough */ |
OUT_BATCH(sna->render_state.gen4.sf); |
OUT_BATCH(sna->render_state.gen4.wm + sp); |
OUT_BATCH(sna->render_state.gen4.cc + bp); |
sna->render_state.gen4.last_pipelined_pointers = key; |
gen4_emit_urb(sna); |
} |
static bool |
gen4_emit_drawing_rectangle(struct sna *sna, const struct sna_composite_op *op) |
{ |
uint32_t limit = (op->dst.height - 1) << 16 | (op->dst.width - 1); |
uint32_t offset = (uint16_t)op->dst.y << 16 | (uint16_t)op->dst.x; |
assert(!too_large(op->dst.x, op->dst.y)); |
assert(!too_large(op->dst.width, op->dst.height)); |
if (sna->render_state.gen4.drawrect_limit == limit && |
sna->render_state.gen4.drawrect_offset == offset) |
return true; |
sna->render_state.gen4.drawrect_offset = offset; |
sna->render_state.gen4.drawrect_limit = limit; |
OUT_BATCH(GEN4_3DSTATE_DRAWING_RECTANGLE | (4 - 2)); |
OUT_BATCH(0); |
OUT_BATCH(limit); |
OUT_BATCH(offset); |
return false; |
} |
static void |
gen4_emit_vertex_elements(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
/* |
* vertex data in vertex buffer |
* position: (x, y) |
* texture coordinate 0: (u0, v0) if (is_affine is true) else (u0, v0, w0) |
* texture coordinate 1 if (has_mask is true): same as above |
*/ |
struct gen4_render_state *render = &sna->render_state.gen4; |
uint32_t src_format, dw; |
int id = op->u.gen4.ve_id; |
if (render->ve_id == id) |
return; |
render->ve_id = id; |
/* The VUE layout |
* dword 0-3: position (x, y, 1.0, 1.0), |
* dword 4-7: texture coordinate 0 (u0, v0, w0, 1.0) |
* [optional] dword 8-11: texture coordinate 1 (u1, v1, w1, 1.0) |
*/ |
OUT_BATCH(GEN4_3DSTATE_VERTEX_ELEMENTS | (2 * (1 + 2) - 1)); |
/* x,y */ |
OUT_BATCH(id << VE0_VERTEX_BUFFER_INDEX_SHIFT | VE0_VALID | |
GEN4_SURFACEFORMAT_R16G16_SSCALED << VE0_FORMAT_SHIFT | |
0 << VE0_OFFSET_SHIFT); |
OUT_BATCH(VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_0_SHIFT | |
VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_1_SHIFT | |
VFCOMPONENT_STORE_1_FLT << VE1_VFCOMPONENT_2_SHIFT | |
VFCOMPONENT_STORE_1_FLT << VE1_VFCOMPONENT_3_SHIFT | |
(1*4) << VE1_DESTINATION_ELEMENT_OFFSET_SHIFT); |
/* u0, v0, w0 */ |
/* u0, v0, w0 */ |
DBG(("%s: first channel %d floats, offset=4b\n", __FUNCTION__, id & 3)); |
dw = VFCOMPONENT_STORE_1_FLT << VE1_VFCOMPONENT_3_SHIFT; |
switch (id & 3) { |
default: |
assert(0); |
case 0: |
src_format = GEN4_SURFACEFORMAT_R16G16_SSCALED; |
dw |= VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_0_SHIFT; |
dw |= VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_1_SHIFT; |
dw |= VFCOMPONENT_STORE_1_FLT << VE1_VFCOMPONENT_2_SHIFT; |
break; |
case 1: |
src_format = GEN4_SURFACEFORMAT_R32_FLOAT; |
dw |= VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_0_SHIFT; |
dw |= VFCOMPONENT_STORE_0 << VE1_VFCOMPONENT_1_SHIFT; |
dw |= VFCOMPONENT_STORE_1_FLT << VE1_VFCOMPONENT_2_SHIFT; |
break; |
case 2: |
src_format = GEN4_SURFACEFORMAT_R32G32_FLOAT; |
dw |= VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_0_SHIFT; |
dw |= VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_1_SHIFT; |
dw |= VFCOMPONENT_STORE_1_FLT << VE1_VFCOMPONENT_2_SHIFT; |
break; |
case 3: |
src_format = GEN4_SURFACEFORMAT_R32G32B32_FLOAT; |
dw |= VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_0_SHIFT; |
dw |= VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_1_SHIFT; |
dw |= VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_2_SHIFT; |
break; |
} |
OUT_BATCH(id << VE0_VERTEX_BUFFER_INDEX_SHIFT | VE0_VALID | |
src_format << VE0_FORMAT_SHIFT | |
4 << VE0_OFFSET_SHIFT); |
OUT_BATCH(dw | 8 << VE1_DESTINATION_ELEMENT_OFFSET_SHIFT); |
/* u1, v1, w1 */ |
if (id >> 2) { |
unsigned src_offset = 4 + ((id & 3) ?: 1) * sizeof(float); |
DBG(("%s: second channel %d floats, offset=%db\n", __FUNCTION__, |
id >> 2, src_offset)); |
dw = VFCOMPONENT_STORE_1_FLT << VE1_VFCOMPONENT_3_SHIFT; |
switch (id >> 2) { |
case 1: |
src_format = GEN4_SURFACEFORMAT_R32_FLOAT; |
dw |= VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_0_SHIFT; |
dw |= VFCOMPONENT_STORE_0 << VE1_VFCOMPONENT_1_SHIFT; |
dw |= VFCOMPONENT_STORE_1_FLT << VE1_VFCOMPONENT_2_SHIFT; |
break; |
default: |
assert(0); |
case 2: |
src_format = GEN4_SURFACEFORMAT_R32G32_FLOAT; |
dw |= VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_0_SHIFT; |
dw |= VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_1_SHIFT; |
dw |= VFCOMPONENT_STORE_1_FLT << VE1_VFCOMPONENT_2_SHIFT; |
break; |
case 3: |
src_format = GEN4_SURFACEFORMAT_R32G32B32_FLOAT; |
dw |= VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_0_SHIFT; |
dw |= VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_1_SHIFT; |
dw |= VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_2_SHIFT; |
break; |
} |
OUT_BATCH(id << VE0_VERTEX_BUFFER_INDEX_SHIFT | VE0_VALID | |
src_format << VE0_FORMAT_SHIFT | |
src_offset << VE0_OFFSET_SHIFT); |
OUT_BATCH(dw | 12 << VE1_DESTINATION_ELEMENT_OFFSET_SHIFT); |
} else { |
OUT_BATCH(id << VE0_VERTEX_BUFFER_INDEX_SHIFT | VE0_VALID | |
GEN4_SURFACEFORMAT_R16G16_SSCALED << VE0_FORMAT_SHIFT | |
0 << VE0_OFFSET_SHIFT); |
OUT_BATCH(VFCOMPONENT_STORE_0 << VE1_VFCOMPONENT_0_SHIFT | |
VFCOMPONENT_STORE_0 << VE1_VFCOMPONENT_1_SHIFT | |
VFCOMPONENT_STORE_0 << VE1_VFCOMPONENT_2_SHIFT | |
VFCOMPONENT_STORE_1_FLT << VE1_VFCOMPONENT_3_SHIFT | |
12 << VE1_DESTINATION_ELEMENT_OFFSET_SHIFT); |
} |
} |
static void |
gen4_emit_state(struct sna *sna, |
const struct sna_composite_op *op, |
uint16_t wm_binding_table) |
{ |
bool flush; |
assert(op->dst.bo->exec); |
flush = wm_binding_table & 1; |
if (kgem_bo_is_dirty(op->src.bo) || kgem_bo_is_dirty(op->mask.bo)) { |
DBG(("%s: flushing dirty (%d, %d), forced? %d\n", __FUNCTION__, |
kgem_bo_is_dirty(op->src.bo), |
kgem_bo_is_dirty(op->mask.bo), |
flush)); |
OUT_BATCH(MI_FLUSH); |
kgem_clear_dirty(&sna->kgem); |
kgem_bo_mark_dirty(op->dst.bo); |
flush = false; |
} |
flush &= gen4_emit_drawing_rectangle(sna, op); |
if (flush && op->op > PictOpSrc) |
OUT_BATCH(MI_FLUSH | MI_INHIBIT_RENDER_CACHE_FLUSH); |
gen4_emit_binding_table(sna, wm_binding_table & ~1); |
gen4_emit_pipelined_pointers(sna, op, op->op, op->u.gen4.wm_kernel); |
gen4_emit_vertex_elements(sna, op); |
} |
static void |
gen4_bind_surfaces(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
bool dirty = kgem_bo_is_dirty(op->dst.bo); |
uint32_t *binding_table; |
uint16_t offset; |
gen4_get_batch(sna, op); |
binding_table = gen4_composite_get_binding_table(sna, &offset); |
binding_table[0] = |
gen4_bind_bo(sna, |
op->dst.bo, op->dst.width, op->dst.height, |
gen4_get_dest_format(op->dst.format), |
true); |
binding_table[1] = |
gen4_bind_bo(sna, |
op->src.bo, op->src.width, op->src.height, |
op->src.card_format, |
false); |
if (op->mask.bo) { |
assert(op->u.gen4.ve_id >> 2); |
binding_table[2] = |
gen4_bind_bo(sna, |
op->mask.bo, |
op->mask.width, |
op->mask.height, |
op->mask.card_format, |
false); |
} |
if (sna->kgem.surface == offset && |
*(uint64_t *)(sna->kgem.batch + sna->render_state.gen4.surface_table) == *(uint64_t*)binding_table && |
(op->mask.bo == NULL || |
sna->kgem.batch[sna->render_state.gen4.surface_table+2] == binding_table[2])) { |
sna->kgem.surface += sizeof(struct gen4_surface_state_padded) / sizeof(uint32_t); |
offset = sna->render_state.gen4.surface_table; |
} |
gen4_emit_state(sna, op, offset | dirty); |
} |
fastcall static void |
gen4_render_composite_blt(struct sna *sna, |
const struct sna_composite_op *op, |
const struct sna_composite_rectangles *r) |
{ |
DBG(("%s: src=(%d, %d)+(%d, %d), mask=(%d, %d)+(%d, %d), dst=(%d, %d)+(%d, %d), size=(%d, %d)\n", |
__FUNCTION__, |
r->src.x, r->src.y, op->src.offset[0], op->src.offset[1], |
r->mask.x, r->mask.y, op->mask.offset[0], op->mask.offset[1], |
r->dst.x, r->dst.y, op->dst.x, op->dst.y, |
r->width, r->height)); |
gen4_get_rectangles(sna, op, 1, gen4_bind_surfaces); |
op->prim_emit(sna, op, r); |
} |
#if 0 |
fastcall static void |
gen4_render_composite_box(struct sna *sna, |
const struct sna_composite_op *op, |
const BoxRec *box) |
{ |
struct sna_composite_rectangles r; |
DBG((" %s: (%d, %d), (%d, %d)\n", |
__FUNCTION__, |
box->x1, box->y1, box->x2, box->y2)); |
gen4_get_rectangles(sna, op, 1, gen4_bind_surfaces); |
r.dst.x = box->x1; |
r.dst.y = box->y1; |
r.width = box->x2 - box->x1; |
r.height = box->y2 - box->y1; |
r.mask = r.src = r.dst; |
op->prim_emit(sna, op, &r); |
} |
static void |
gen4_render_composite_boxes__blt(struct sna *sna, |
const struct sna_composite_op *op, |
const BoxRec *box, int nbox) |
{ |
DBG(("%s(%d) delta=(%d, %d), src=(%d, %d)/(%d, %d), mask=(%d, %d)/(%d, %d)\n", |
__FUNCTION__, nbox, op->dst.x, op->dst.y, |
op->src.offset[0], op->src.offset[1], |
op->src.width, op->src.height, |
op->mask.offset[0], op->mask.offset[1], |
op->mask.width, op->mask.height)); |
do { |
int nbox_this_time; |
nbox_this_time = gen4_get_rectangles(sna, op, nbox, |
gen4_bind_surfaces); |
nbox -= nbox_this_time; |
do { |
struct sna_composite_rectangles r; |
DBG((" %s: (%d, %d), (%d, %d)\n", |
__FUNCTION__, |
box->x1, box->y1, box->x2, box->y2)); |
r.dst.x = box->x1; |
r.dst.y = box->y1; |
r.width = box->x2 - box->x1; |
r.height = box->y2 - box->y1; |
r.mask = r.src = r.dst; |
op->prim_emit(sna, op, &r); |
box++; |
} while (--nbox_this_time); |
} while (nbox); |
} |
static void |
gen4_render_composite_boxes(struct sna *sna, |
const struct sna_composite_op *op, |
const BoxRec *box, int nbox) |
{ |
DBG(("%s: nbox=%d\n", __FUNCTION__, nbox)); |
do { |
int nbox_this_time; |
float *v; |
nbox_this_time = gen4_get_rectangles(sna, op, nbox, |
gen4_bind_surfaces); |
assert(nbox_this_time); |
nbox -= nbox_this_time; |
v = sna->render.vertices + sna->render.vertex_used; |
sna->render.vertex_used += nbox_this_time * op->floats_per_rect; |
op->emit_boxes(op, box, nbox_this_time, v); |
box += nbox_this_time; |
} while (nbox); |
} |
#if !FORCE_FLUSH |
static void |
gen4_render_composite_boxes__thread(struct sna *sna, |
const struct sna_composite_op *op, |
const BoxRec *box, int nbox) |
{ |
DBG(("%s: nbox=%d\n", __FUNCTION__, nbox)); |
sna_vertex_lock(&sna->render); |
do { |
int nbox_this_time; |
float *v; |
nbox_this_time = gen4_get_rectangles(sna, op, nbox, |
gen4_bind_surfaces); |
assert(nbox_this_time); |
nbox -= nbox_this_time; |
v = sna->render.vertices + sna->render.vertex_used; |
sna->render.vertex_used += nbox_this_time * op->floats_per_rect; |
sna_vertex_acquire__locked(&sna->render); |
sna_vertex_unlock(&sna->render); |
op->emit_boxes(op, box, nbox_this_time, v); |
box += nbox_this_time; |
sna_vertex_lock(&sna->render); |
sna_vertex_release__locked(&sna->render); |
} while (nbox); |
sna_vertex_unlock(&sna->render); |
} |
#endif |
#ifndef MAX |
#define MAX(a,b) ((a) > (b) ? (a) : (b)) |
#endif |
static uint32_t gen4_bind_video_source(struct sna *sna, |
struct kgem_bo *src_bo, |
uint32_t src_offset, |
int src_width, |
int src_height, |
int src_pitch, |
uint32_t src_surf_format) |
{ |
struct gen4_surface_state *ss; |
sna->kgem.surface -= sizeof(struct gen4_surface_state_padded) / sizeof(uint32_t); |
ss = memset(sna->kgem.batch + sna->kgem.surface, 0, sizeof(*ss)); |
ss->ss0.surface_type = GEN4_SURFACE_2D; |
ss->ss0.surface_format = src_surf_format; |
ss->ss0.color_blend = 1; |
ss->ss1.base_addr = |
kgem_add_reloc(&sna->kgem, |
sna->kgem.surface + 1, |
src_bo, |
I915_GEM_DOMAIN_SAMPLER << 16, |
src_offset); |
ss->ss2.width = src_width - 1; |
ss->ss2.height = src_height - 1; |
ss->ss3.pitch = src_pitch - 1; |
return sna->kgem.surface * sizeof(uint32_t); |
} |
static void gen4_video_bind_surfaces(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
bool dirty = kgem_bo_is_dirty(op->dst.bo); |
struct sna_video_frame *frame = op->priv; |
uint32_t src_surf_format; |
uint32_t src_surf_base[6]; |
int src_width[6]; |
int src_height[6]; |
int src_pitch[6]; |
uint32_t *binding_table; |
uint16_t offset; |
int n_src, n; |
src_surf_base[0] = 0; |
src_surf_base[1] = 0; |
src_surf_base[2] = frame->VBufOffset; |
src_surf_base[3] = frame->VBufOffset; |
src_surf_base[4] = frame->UBufOffset; |
src_surf_base[5] = frame->UBufOffset; |
if (is_planar_fourcc(frame->id)) { |
src_surf_format = GEN4_SURFACEFORMAT_R8_UNORM; |
src_width[1] = src_width[0] = frame->width; |
src_height[1] = src_height[0] = frame->height; |
src_pitch[1] = src_pitch[0] = frame->pitch[1]; |
src_width[4] = src_width[5] = src_width[2] = src_width[3] = |
frame->width / 2; |
src_height[4] = src_height[5] = src_height[2] = src_height[3] = |
frame->height / 2; |
src_pitch[4] = src_pitch[5] = src_pitch[2] = src_pitch[3] = |
frame->pitch[0]; |
n_src = 6; |
} else { |
if (frame->id == FOURCC_UYVY) |
src_surf_format = GEN4_SURFACEFORMAT_YCRCB_SWAPY; |
else |
src_surf_format = GEN4_SURFACEFORMAT_YCRCB_NORMAL; |
src_width[0] = frame->width; |
src_height[0] = frame->height; |
src_pitch[0] = frame->pitch[0]; |
n_src = 1; |
} |
gen4_get_batch(sna, op); |
binding_table = gen4_composite_get_binding_table(sna, &offset); |
binding_table[0] = |
gen4_bind_bo(sna, |
op->dst.bo, op->dst.width, op->dst.height, |
gen4_get_dest_format(op->dst.format), |
true); |
for (n = 0; n < n_src; n++) { |
binding_table[1+n] = |
gen4_bind_video_source(sna, |
frame->bo, |
src_surf_base[n], |
src_width[n], |
src_height[n], |
src_pitch[n], |
src_surf_format); |
} |
gen4_emit_state(sna, op, offset | dirty); |
} |
static bool |
gen4_render_video(struct sna *sna, |
struct sna_video *video, |
struct sna_video_frame *frame, |
RegionPtr dstRegion, |
PixmapPtr pixmap) |
{ |
struct sna_composite_op tmp; |
int dst_width = dstRegion->extents.x2 - dstRegion->extents.x1; |
int dst_height = dstRegion->extents.y2 - dstRegion->extents.y1; |
int src_width = frame->src.x2 - frame->src.x1; |
int src_height = frame->src.y2 - frame->src.y1; |
float src_offset_x, src_offset_y; |
float src_scale_x, src_scale_y; |
int nbox, pix_xoff, pix_yoff; |
struct sna_pixmap *priv; |
BoxPtr box; |
DBG(("%s: %dx%d -> %dx%d\n", __FUNCTION__, |
src_width, src_height, dst_width, dst_height)); |
priv = sna_pixmap_force_to_gpu(pixmap, MOVE_READ | MOVE_WRITE); |
if (priv == NULL) |
return false; |
memset(&tmp, 0, sizeof(tmp)); |
tmp.op = PictOpSrc; |
tmp.dst.pixmap = pixmap; |
tmp.dst.width = pixmap->drawable.width; |
tmp.dst.height = pixmap->drawable.height; |
tmp.dst.format = sna_format_for_depth(pixmap->drawable.depth); |
tmp.dst.bo = priv->gpu_bo; |
if (src_width == dst_width && src_height == dst_height) |
tmp.src.filter = SAMPLER_FILTER_NEAREST; |
else |
tmp.src.filter = SAMPLER_FILTER_BILINEAR; |
tmp.src.repeat = SAMPLER_EXTEND_PAD; |
tmp.src.bo = frame->bo; |
tmp.mask.bo = NULL; |
tmp.u.gen4.wm_kernel = |
is_planar_fourcc(frame->id) ? WM_KERNEL_VIDEO_PLANAR : WM_KERNEL_VIDEO_PACKED; |
tmp.u.gen4.ve_id = 2; |
tmp.is_affine = true; |
tmp.floats_per_vertex = 3; |
tmp.floats_per_rect = 9; |
tmp.priv = frame; |
if (!kgem_check_bo(&sna->kgem, tmp.dst.bo, frame->bo, NULL)) { |
kgem_submit(&sna->kgem); |
assert(kgem_check_bo(&sna->kgem, tmp.dst.bo, frame->bo, NULL)); |
} |
gen4_video_bind_surfaces(sna, &tmp); |
gen4_align_vertex(sna, &tmp); |
/* Set up the offset for translating from the given region (in screen |
* coordinates) to the backing pixmap. |
*/ |
#ifdef COMPOSITE |
pix_xoff = -pixmap->screen_x + pixmap->drawable.x; |
pix_yoff = -pixmap->screen_y + pixmap->drawable.y; |
#else |
pix_xoff = 0; |
pix_yoff = 0; |
#endif |
src_scale_x = (float)src_width / dst_width / frame->width; |
src_offset_x = (float)frame->src.x1 / frame->width - dstRegion->extents.x1 * src_scale_x; |
src_scale_y = (float)src_height / dst_height / frame->height; |
src_offset_y = (float)frame->src.y1 / frame->height - dstRegion->extents.y1 * src_scale_y; |
box = REGION_RECTS(dstRegion); |
nbox = REGION_NUM_RECTS(dstRegion); |
do { |
int n; |
n = gen4_get_rectangles(sna, &tmp, nbox, |
gen4_video_bind_surfaces); |
assert(n); |
nbox -= n; |
do { |
BoxRec r; |
r.x1 = box->x1 + pix_xoff; |
r.x2 = box->x2 + pix_xoff; |
r.y1 = box->y1 + pix_yoff; |
r.y2 = box->y2 + pix_yoff; |
OUT_VERTEX(r.x2, r.y2); |
OUT_VERTEX_F(box->x2 * src_scale_x + src_offset_x); |
OUT_VERTEX_F(box->y2 * src_scale_y + src_offset_y); |
OUT_VERTEX(r.x1, r.y2); |
OUT_VERTEX_F(box->x1 * src_scale_x + src_offset_x); |
OUT_VERTEX_F(box->y2 * src_scale_y + src_offset_y); |
OUT_VERTEX(r.x1, r.y1); |
OUT_VERTEX_F(box->x1 * src_scale_x + src_offset_x); |
OUT_VERTEX_F(box->y1 * src_scale_y + src_offset_y); |
if (!DAMAGE_IS_ALL(priv->gpu_damage)) { |
sna_damage_add_box(&priv->gpu_damage, &r); |
sna_damage_subtract_box(&priv->cpu_damage, &r); |
} |
box++; |
} while (--n); |
} while (nbox); |
gen4_vertex_flush(sna); |
return true; |
} |
static int |
gen4_composite_picture(struct sna *sna, |
PicturePtr picture, |
struct sna_composite_channel *channel, |
int x, int y, |
int w, int h, |
int dst_x, int dst_y, |
bool precise) |
{ |
PixmapPtr pixmap; |
uint32_t color; |
int16_t dx, dy; |
DBG(("%s: (%d, %d)x(%d, %d), dst=(%d, %d)\n", |
__FUNCTION__, x, y, w, h, dst_x, dst_y)); |
channel->is_solid = false; |
channel->card_format = -1; |
if (sna_picture_is_solid(picture, &color)) |
return gen4_channel_init_solid(sna, channel, color); |
if (picture->pDrawable == NULL) { |
int ret; |
if (picture->pSourcePict->type == SourcePictTypeLinear) |
return gen4_channel_init_linear(sna, picture, channel, |
x, y, |
w, h, |
dst_x, dst_y); |
DBG(("%s -- fixup, gradient\n", __FUNCTION__)); |
ret = -1; |
if (!precise) |
ret = sna_render_picture_approximate_gradient(sna, picture, channel, |
x, y, w, h, dst_x, dst_y); |
if (ret == -1) |
ret = sna_render_picture_fixup(sna, picture, channel, |
x, y, w, h, dst_x, dst_y); |
return ret; |
} |
if (picture->alphaMap) { |
DBG(("%s -- fallback, alphamap\n", __FUNCTION__)); |
return sna_render_picture_fixup(sna, picture, channel, |
x, y, w, h, dst_x, dst_y); |
} |
if (!gen4_check_repeat(picture)) { |
DBG(("%s: unknown repeat mode fixup\n", __FUNCTION__)); |
return sna_render_picture_fixup(sna, picture, channel, |
x, y, w, h, dst_x, dst_y); |
} |
if (!gen4_check_filter(picture)) { |
DBG(("%s: unhandled filter fixup\n", __FUNCTION__)); |
return sna_render_picture_fixup(sna, picture, channel, |
x, y, w, h, dst_x, dst_y); |
} |
channel->repeat = picture->repeat ? picture->repeatType : RepeatNone; |
channel->filter = picture->filter; |
pixmap = get_drawable_pixmap(picture->pDrawable); |
get_drawable_deltas(picture->pDrawable, pixmap, &dx, &dy); |
x += dx + picture->pDrawable->x; |
y += dy + picture->pDrawable->y; |
channel->is_affine = sna_transform_is_affine(picture->transform); |
if (sna_transform_is_integer_translation(picture->transform, &dx, &dy)) { |
DBG(("%s: integer translation (%d, %d), removing\n", |
__FUNCTION__, dx, dy)); |
x += dx; |
y += dy; |
channel->transform = NULL; |
channel->filter = PictFilterNearest; |
} else |
channel->transform = picture->transform; |
channel->pict_format = picture->format; |
channel->card_format = gen4_get_card_format(picture->format); |
if (channel->card_format == -1) |
return sna_render_picture_convert(sna, picture, channel, pixmap, |
x, y, w, h, dst_x, dst_y, |
false); |
if (too_large(pixmap->drawable.width, pixmap->drawable.height)) |
return sna_render_picture_extract(sna, picture, channel, |
x, y, w, h, dst_x, dst_y); |
return sna_render_pixmap_bo(sna, channel, pixmap, |
x, y, w, h, dst_x, dst_y); |
} |
static void gen4_composite_channel_convert(struct sna_composite_channel *channel) |
{ |
DBG(("%s: repeat %d -> %d, filter %d -> %d\n", |
__FUNCTION__, |
channel->repeat, gen4_repeat(channel->repeat), |
channel->filter, gen4_repeat(channel->filter))); |
channel->repeat = gen4_repeat(channel->repeat); |
channel->filter = gen4_filter(channel->filter); |
if (channel->card_format == (unsigned)-1) |
channel->card_format = gen4_get_card_format(channel->pict_format); |
} |
#endif |
static void |
gen4_render_composite_done(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
DBG(("%s()\n", __FUNCTION__)); |
if (sna->render.vertex_offset) { |
gen4_vertex_flush(sna); |
gen4_magic_ca_pass(sna, op); |
} |
} |
#if 0 |
static bool |
gen4_composite_set_target(struct sna *sna, |
struct sna_composite_op *op, |
PicturePtr dst, |
int x, int y, int w, int h, |
bool partial) |
{ |
BoxRec box; |
op->dst.pixmap = get_drawable_pixmap(dst->pDrawable); |
op->dst.width = op->dst.pixmap->drawable.width; |
op->dst.height = op->dst.pixmap->drawable.height; |
op->dst.format = dst->format; |
if (w && h) { |
box.x1 = x; |
box.y1 = y; |
box.x2 = x + w; |
box.y2 = y + h; |
} else |
sna_render_picture_extents(dst, &box); |
op->dst.bo = sna_drawable_use_bo (dst->pDrawable, |
PREFER_GPU | FORCE_GPU | RENDER_GPU, |
&box, &op->damage); |
if (op->dst.bo == NULL) |
return false; |
get_drawable_deltas(dst->pDrawable, op->dst.pixmap, |
&op->dst.x, &op->dst.y); |
DBG(("%s: pixmap=%p, format=%08x, size=%dx%d, pitch=%d, delta=(%d,%d),damage=%p\n", |
__FUNCTION__, |
op->dst.pixmap, (int)op->dst.format, |
op->dst.width, op->dst.height, |
op->dst.bo->pitch, |
op->dst.x, op->dst.y, |
op->damage ? *op->damage : (void *)-1)); |
assert(op->dst.bo->proxy == NULL); |
if (too_large(op->dst.width, op->dst.height) && |
!sna_render_composite_redirect(sna, op, x, y, w, h, partial)) |
return false; |
return true; |
} |
static bool |
try_blt(struct sna *sna, |
PicturePtr dst, PicturePtr src, |
int width, int height) |
{ |
if (sna->kgem.mode != KGEM_RENDER) { |
DBG(("%s: already performing BLT\n", __FUNCTION__)); |
return true; |
} |
if (too_large(width, height)) { |
DBG(("%s: operation too large for 3D pipe (%d, %d)\n", |
__FUNCTION__, width, height)); |
return true; |
} |
if (too_large(dst->pDrawable->width, dst->pDrawable->height)) |
return true; |
/* The blitter is much faster for solids */ |
if (sna_picture_is_solid(src, NULL)) |
return true; |
/* is the source picture only in cpu memory e.g. a shm pixmap? */ |
return picture_is_cpu(sna, src); |
} |
static bool |
check_gradient(PicturePtr picture, bool precise) |
{ |
switch (picture->pSourcePict->type) { |
case SourcePictTypeSolidFill: |
case SourcePictTypeLinear: |
return false; |
default: |
return precise; |
} |
} |
static bool |
has_alphamap(PicturePtr p) |
{ |
return p->alphaMap != NULL; |
} |
static bool |
need_upload(struct sna *sna, PicturePtr p) |
{ |
return p->pDrawable && untransformed(p) && |
!is_gpu(sna, p->pDrawable, PREFER_GPU_RENDER); |
} |
static bool |
source_is_busy(PixmapPtr pixmap) |
{ |
struct sna_pixmap *priv = sna_pixmap(pixmap); |
if (priv == NULL) |
return false; |
if (priv->clear) |
return false; |
if (priv->gpu_bo && kgem_bo_is_busy(priv->gpu_bo)) |
return true; |
if (priv->cpu_bo && kgem_bo_is_busy(priv->cpu_bo)) |
return true; |
return priv->gpu_damage && !priv->cpu_damage; |
} |
static bool |
source_fallback(struct sna *sna, PicturePtr p, PixmapPtr pixmap, bool precise) |
{ |
if (sna_picture_is_solid(p, NULL)) |
return false; |
if (p->pSourcePict) |
return check_gradient(p, precise); |
if (!gen4_check_repeat(p) || !gen4_check_format(p->format)) |
return true; |
/* soft errors: perfer to upload/compute rather than readback */ |
if (pixmap && source_is_busy(pixmap)) |
return false; |
return has_alphamap(p) || !gen4_check_filter(p) || need_upload(sna, p); |
} |
static bool |
gen4_composite_fallback(struct sna *sna, |
PicturePtr src, |
PicturePtr mask, |
PicturePtr dst) |
{ |
PixmapPtr src_pixmap; |
PixmapPtr mask_pixmap; |
PixmapPtr dst_pixmap; |
bool src_fallback, mask_fallback; |
if (!gen4_check_dst_format(dst->format)) { |
DBG(("%s: unknown destination format: %d\n", |
__FUNCTION__, dst->format)); |
return true; |
} |
dst_pixmap = get_drawable_pixmap(dst->pDrawable); |
src_pixmap = src->pDrawable ? get_drawable_pixmap(src->pDrawable) : NULL; |
src_fallback = source_fallback(sna, src, src_pixmap, |
dst->polyMode == PolyModePrecise); |
if (mask) { |
mask_pixmap = mask->pDrawable ? get_drawable_pixmap(mask->pDrawable) : NULL; |
mask_fallback = source_fallback(sna, mask, mask_pixmap, |
dst->polyMode == PolyModePrecise); |
} else { |
mask_pixmap = NULL; |
mask_fallback = false; |
} |
/* If we are using the destination as a source and need to |
* readback in order to upload the source, do it all |
* on the cpu. |
*/ |
if (src_pixmap == dst_pixmap && src_fallback) { |
DBG(("%s: src is dst and will fallback\n",__FUNCTION__)); |
return true; |
} |
if (mask_pixmap == dst_pixmap && mask_fallback) { |
DBG(("%s: mask is dst and will fallback\n",__FUNCTION__)); |
return true; |
} |
/* If anything is on the GPU, push everything out to the GPU */ |
if (dst_use_gpu(dst_pixmap)) { |
DBG(("%s: dst is already on the GPU, try to use GPU\n", |
__FUNCTION__)); |
return false; |
} |
if (src_pixmap && !src_fallback) { |
DBG(("%s: src is already on the GPU, try to use GPU\n", |
__FUNCTION__)); |
return false; |
} |
if (mask_pixmap && !mask_fallback) { |
DBG(("%s: mask is already on the GPU, try to use GPU\n", |
__FUNCTION__)); |
return false; |
} |
/* However if the dst is not on the GPU and we need to |
* render one of the sources using the CPU, we may |
* as well do the entire operation in place onthe CPU. |
*/ |
if (src_fallback) { |
DBG(("%s: dst is on the CPU and src will fallback\n", |
__FUNCTION__)); |
return true; |
} |
if (mask_fallback) { |
DBG(("%s: dst is on the CPU and mask will fallback\n", |
__FUNCTION__)); |
return true; |
} |
if (too_large(dst_pixmap->drawable.width, |
dst_pixmap->drawable.height) && |
dst_is_cpu(dst_pixmap)) { |
DBG(("%s: dst is on the CPU and too large\n", __FUNCTION__)); |
return true; |
} |
DBG(("%s: dst is not on the GPU and the operation should not fallback\n", |
__FUNCTION__)); |
return dst_use_cpu(dst_pixmap); |
} |
static int |
reuse_source(struct sna *sna, |
PicturePtr src, struct sna_composite_channel *sc, int src_x, int src_y, |
PicturePtr mask, struct sna_composite_channel *mc, int msk_x, int msk_y) |
{ |
uint32_t color; |
if (src_x != msk_x || src_y != msk_y) |
return false; |
if (src == mask) { |
DBG(("%s: mask is source\n", __FUNCTION__)); |
*mc = *sc; |
mc->bo = kgem_bo_reference(mc->bo); |
return true; |
} |
if (sna_picture_is_solid(mask, &color)) |
return gen4_channel_init_solid(sna, mc, color); |
if (sc->is_solid) |
return false; |
if (src->pDrawable == NULL || mask->pDrawable != src->pDrawable) |
return false; |
DBG(("%s: mask reuses source drawable\n", __FUNCTION__)); |
if (!sna_transform_equal(src->transform, mask->transform)) |
return false; |
if (!sna_picture_alphamap_equal(src, mask)) |
return false; |
if (!gen4_check_repeat(mask)) |
return false; |
if (!gen4_check_filter(mask)) |
return false; |
if (!gen4_check_format(mask->format)) |
return false; |
DBG(("%s: reusing source channel for mask with a twist\n", |
__FUNCTION__)); |
*mc = *sc; |
mc->repeat = gen4_repeat(mask->repeat ? mask->repeatType : RepeatNone); |
mc->filter = gen4_filter(mask->filter); |
mc->pict_format = mask->format; |
mc->card_format = gen4_get_card_format(mask->format); |
mc->bo = kgem_bo_reference(mc->bo); |
return true; |
} |
static bool |
gen4_render_composite(struct sna *sna, |
uint8_t op, |
PicturePtr src, |
PicturePtr mask, |
PicturePtr dst, |
int16_t src_x, int16_t src_y, |
int16_t msk_x, int16_t msk_y, |
int16_t dst_x, int16_t dst_y, |
int16_t width, int16_t height, |
struct sna_composite_op *tmp) |
{ |
DBG(("%s: %dx%d, current mode=%d\n", __FUNCTION__, |
width, height, sna->kgem.mode)); |
if (op >= ARRAY_SIZE(gen4_blend_op)) |
return false; |
if (mask == NULL && |
try_blt(sna, dst, src, width, height) && |
sna_blt_composite(sna, op, |
src, dst, |
src_x, src_y, |
dst_x, dst_y, |
width, height, |
tmp, false)) |
return true; |
if (gen4_composite_fallback(sna, src, mask, dst)) |
return false; |
if (need_tiling(sna, width, height)) |
return sna_tiling_composite(op, src, mask, dst, |
src_x, src_y, |
msk_x, msk_y, |
dst_x, dst_y, |
width, height, |
tmp); |
if (!gen4_composite_set_target(sna, tmp, dst, |
dst_x, dst_y, width, height, |
op > PictOpSrc || dst->pCompositeClip->data)) { |
DBG(("%s: failed to set composite target\n", __FUNCTION__)); |
return false; |
} |
tmp->op = op; |
switch (gen4_composite_picture(sna, src, &tmp->src, |
src_x, src_y, |
width, height, |
dst_x, dst_y, |
dst->polyMode == PolyModePrecise)) { |
case -1: |
DBG(("%s: failed to prepare source\n", __FUNCTION__)); |
goto cleanup_dst; |
case 0: |
if (!gen4_channel_init_solid(sna, &tmp->src, 0)) |
goto cleanup_dst; |
/* fall through to fixup */ |
case 1: |
if (mask == NULL && |
sna_blt_composite__convert(sna, |
dst_x, dst_y, width, height, |
tmp)) |
return true; |
gen4_composite_channel_convert(&tmp->src); |
break; |
} |
tmp->is_affine = tmp->src.is_affine; |
tmp->has_component_alpha = false; |
tmp->need_magic_ca_pass = false; |
if (mask) { |
if (mask->componentAlpha && PICT_FORMAT_RGB(mask->format)) { |
tmp->has_component_alpha = true; |
/* Check if it's component alpha that relies on a source alpha and on |
* the source value. We can only get one of those into the single |
* source value that we get to blend with. |
*/ |
if (gen4_blend_op[op].src_alpha && |
(gen4_blend_op[op].src_blend != GEN4_BLENDFACTOR_ZERO)) { |
if (op != PictOpOver) { |
DBG(("%s -- fallback: unhandled component alpha blend\n", |
__FUNCTION__)); |
goto cleanup_src; |
} |
tmp->need_magic_ca_pass = true; |
tmp->op = PictOpOutReverse; |
} |
} |
if (!reuse_source(sna, |
src, &tmp->src, src_x, src_y, |
mask, &tmp->mask, msk_x, msk_y)) { |
switch (gen4_composite_picture(sna, mask, &tmp->mask, |
msk_x, msk_y, |
width, height, |
dst_x, dst_y, |
dst->polyMode == PolyModePrecise)) { |
case -1: |
DBG(("%s: failed to prepare mask\n", __FUNCTION__)); |
goto cleanup_src; |
case 0: |
if (!gen4_channel_init_solid(sna, &tmp->mask, 0)) |
goto cleanup_src; |
/* fall through to fixup */ |
case 1: |
gen4_composite_channel_convert(&tmp->mask); |
break; |
} |
} |
tmp->is_affine &= tmp->mask.is_affine; |
} |
tmp->u.gen4.wm_kernel = |
gen4_choose_composite_kernel(tmp->op, |
tmp->mask.bo != NULL, |
tmp->has_component_alpha, |
tmp->is_affine); |
tmp->u.gen4.ve_id = gen4_choose_composite_emitter(sna, tmp); |
tmp->blt = gen4_render_composite_blt; |
tmp->box = gen4_render_composite_box; |
tmp->boxes = gen4_render_composite_boxes__blt; |
if (tmp->emit_boxes) { |
tmp->boxes = gen4_render_composite_boxes; |
#if !FORCE_FLUSH |
tmp->thread_boxes = gen4_render_composite_boxes__thread; |
#endif |
} |
tmp->done = gen4_render_composite_done; |
if (!kgem_check_bo(&sna->kgem, |
tmp->dst.bo, tmp->src.bo, tmp->mask.bo, |
NULL)) { |
kgem_submit(&sna->kgem); |
if (!kgem_check_bo(&sna->kgem, |
tmp->dst.bo, tmp->src.bo, tmp->mask.bo, |
NULL)) |
goto cleanup_mask; |
} |
gen4_bind_surfaces(sna, tmp); |
gen4_align_vertex(sna, tmp); |
return true; |
cleanup_mask: |
if (tmp->mask.bo) |
kgem_bo_destroy(&sna->kgem, tmp->mask.bo); |
cleanup_src: |
if (tmp->src.bo) |
kgem_bo_destroy(&sna->kgem, tmp->src.bo); |
cleanup_dst: |
if (tmp->redirect.real_bo) |
kgem_bo_destroy(&sna->kgem, tmp->dst.bo); |
return false; |
} |
#endif |
static void |
gen4_render_flush(struct sna *sna) |
{ |
gen4_vertex_close(sna); |
assert(sna->render.vb_id == 0); |
assert(sna->render.vertex_offset == 0); |
} |
static void |
discard_vbo(struct sna *sna) |
{ |
kgem_bo_destroy(&sna->kgem, sna->render.vbo); |
sna->render.vbo = NULL; |
sna->render.vertices = sna->render.vertex_data; |
sna->render.vertex_size = ARRAY_SIZE(sna->render.vertex_data); |
sna->render.vertex_used = 0; |
sna->render.vertex_index = 0; |
} |
static void |
gen4_render_retire(struct kgem *kgem) |
{ |
struct sna *sna; |
sna = container_of(kgem, struct sna, kgem); |
if (kgem->nbatch == 0 && sna->render.vbo && !kgem_bo_is_busy(sna->render.vbo)) { |
DBG(("%s: resetting idle vbo\n", __FUNCTION__)); |
sna->render.vertex_used = 0; |
sna->render.vertex_index = 0; |
} |
} |
static void |
gen4_render_expire(struct kgem *kgem) |
{ |
struct sna *sna; |
sna = container_of(kgem, struct sna, kgem); |
if (sna->render.vbo && !sna->render.vertex_used) { |
DBG(("%s: discarding vbo\n", __FUNCTION__)); |
discard_vbo(sna); |
} |
} |
static void gen4_render_reset(struct sna *sna) |
{ |
sna->render_state.gen4.needs_invariant = true; |
sna->render_state.gen4.needs_urb = true; |
sna->render_state.gen4.ve_id = -1; |
sna->render_state.gen4.last_primitive = -1; |
sna->render_state.gen4.last_pipelined_pointers = -1; |
sna->render_state.gen4.drawrect_offset = -1; |
sna->render_state.gen4.drawrect_limit = -1; |
sna->render_state.gen4.surface_table = -1; |
if (sna->render.vbo && |
!kgem_bo_is_mappable(&sna->kgem, sna->render.vbo)) { |
DBG(("%s: discarding unmappable vbo\n", __FUNCTION__)); |
discard_vbo(sna); |
} |
sna->render.vertex_offset = 0; |
sna->render.nvertex_reloc = 0; |
sna->render.vb_id = 0; |
} |
static void gen4_render_fini(struct sna *sna) |
{ |
kgem_bo_destroy(&sna->kgem, sna->render_state.gen4.general_bo); |
} |
static uint32_t gen4_create_vs_unit_state(struct sna_static_stream *stream) |
{ |
struct gen4_vs_unit_state *vs = sna_static_stream_map(stream, sizeof(*vs), 32); |
/* Set up the vertex shader to be disabled (passthrough) */ |
vs->thread4.nr_urb_entries = URB_VS_ENTRIES; |
vs->thread4.urb_entry_allocation_size = URB_VS_ENTRY_SIZE - 1; |
vs->vs6.vs_enable = 0; |
vs->vs6.vert_cache_disable = 1; |
return sna_static_stream_offsetof(stream, vs); |
} |
static uint32_t gen4_create_sf_state(struct sna_static_stream *stream, |
uint32_t kernel) |
{ |
struct gen4_sf_unit_state *sf; |
sf = sna_static_stream_map(stream, sizeof(*sf), 32); |
sf->thread0.grf_reg_count = GEN4_GRF_BLOCKS(SF_KERNEL_NUM_GRF); |
sf->thread0.kernel_start_pointer = kernel >> 6; |
sf->thread3.const_urb_entry_read_length = 0; /* no const URBs */ |
sf->thread3.const_urb_entry_read_offset = 0; /* no const URBs */ |
sf->thread3.urb_entry_read_length = 1; /* 1 URB per vertex */ |
/* don't smash vertex header, read start from dw8 */ |
sf->thread3.urb_entry_read_offset = 1; |
sf->thread3.dispatch_grf_start_reg = 3; |
sf->thread4.max_threads = GEN4_MAX_SF_THREADS - 1; |
sf->thread4.urb_entry_allocation_size = URB_SF_ENTRY_SIZE - 1; |
sf->thread4.nr_urb_entries = URB_SF_ENTRIES; |
sf->sf5.viewport_transform = false; /* skip viewport */ |
sf->sf6.cull_mode = GEN4_CULLMODE_NONE; |
sf->sf6.scissor = 0; |
sf->sf7.trifan_pv = 2; |
sf->sf6.dest_org_vbias = 0x8; |
sf->sf6.dest_org_hbias = 0x8; |
return sna_static_stream_offsetof(stream, sf); |
} |
static uint32_t gen4_create_sampler_state(struct sna_static_stream *stream, |
sampler_filter_t src_filter, |
sampler_extend_t src_extend, |
sampler_filter_t mask_filter, |
sampler_extend_t mask_extend) |
{ |
struct gen4_sampler_state *sampler_state; |
sampler_state = sna_static_stream_map(stream, |
sizeof(struct gen4_sampler_state) * 2, |
32); |
sampler_state_init(&sampler_state[0], src_filter, src_extend); |
sampler_state_init(&sampler_state[1], mask_filter, mask_extend); |
return sna_static_stream_offsetof(stream, sampler_state); |
} |
static void gen4_init_wm_state(struct gen4_wm_unit_state *wm, |
int gen, |
bool has_mask, |
uint32_t kernel, |
uint32_t sampler) |
{ |
assert((kernel & 63) == 0); |
wm->thread0.kernel_start_pointer = kernel >> 6; |
wm->thread0.grf_reg_count = GEN4_GRF_BLOCKS(PS_KERNEL_NUM_GRF); |
wm->thread1.single_program_flow = 0; |
wm->thread3.const_urb_entry_read_length = 0; |
wm->thread3.const_urb_entry_read_offset = 0; |
wm->thread3.urb_entry_read_offset = 0; |
wm->thread3.dispatch_grf_start_reg = 3; |
assert((sampler & 31) == 0); |
wm->wm4.sampler_state_pointer = sampler >> 5; |
wm->wm4.sampler_count = 1; |
wm->wm5.max_threads = gen >= 045 ? G4X_MAX_WM_THREADS - 1 : GEN4_MAX_WM_THREADS - 1; |
wm->wm5.transposed_urb_read = 0; |
wm->wm5.thread_dispatch_enable = 1; |
/* just use 16-pixel dispatch (4 subspans), don't need to change kernel |
* start point |
*/ |
wm->wm5.enable_16_pix = 1; |
wm->wm5.enable_8_pix = 0; |
wm->wm5.early_depth_test = 1; |
/* Each pair of attributes (src/mask coords) is two URB entries */ |
if (has_mask) { |
wm->thread1.binding_table_entry_count = 3; |
wm->thread3.urb_entry_read_length = 4; |
} else { |
wm->thread1.binding_table_entry_count = 2; |
wm->thread3.urb_entry_read_length = 2; |
} |
} |
static uint32_t gen4_create_cc_unit_state(struct sna_static_stream *stream) |
{ |
uint8_t *ptr, *base; |
int i, j; |
base = ptr = |
sna_static_stream_map(stream, |
GEN4_BLENDFACTOR_COUNT*GEN4_BLENDFACTOR_COUNT*64, |
64); |
for (i = 0; i < GEN4_BLENDFACTOR_COUNT; i++) { |
for (j = 0; j < GEN4_BLENDFACTOR_COUNT; j++) { |
struct gen4_cc_unit_state *state = |
(struct gen4_cc_unit_state *)ptr; |
state->cc3.blend_enable = |
!(j == GEN4_BLENDFACTOR_ZERO && i == GEN4_BLENDFACTOR_ONE); |
state->cc5.logicop_func = 0xc; /* COPY */ |
state->cc5.ia_blend_function = GEN4_BLENDFUNCTION_ADD; |
/* Fill in alpha blend factors same as color, for the future. */ |
state->cc5.ia_src_blend_factor = i; |
state->cc5.ia_dest_blend_factor = j; |
state->cc6.blend_function = GEN4_BLENDFUNCTION_ADD; |
state->cc6.clamp_post_alpha_blend = 1; |
state->cc6.clamp_pre_alpha_blend = 1; |
state->cc6.src_blend_factor = i; |
state->cc6.dest_blend_factor = j; |
ptr += 64; |
} |
} |
return sna_static_stream_offsetof(stream, base); |
} |
static bool gen4_render_setup(struct sna *sna) |
{ |
struct gen4_render_state *state = &sna->render_state.gen4; |
struct sna_static_stream general; |
struct gen4_wm_unit_state_padded *wm_state; |
uint32_t sf, wm[KERNEL_COUNT]; |
int i, j, k, l, m; |
sna_static_stream_init(&general); |
/* Zero pad the start. If you see an offset of 0x0 in the batchbuffer |
* dumps, you know it points to zero. |
*/ |
null_create(&general); |
sf = sna_static_stream_compile_sf(sna, &general, brw_sf_kernel__mask); |
for (m = 0; m < KERNEL_COUNT; m++) { |
if (wm_kernels[m].size) { |
wm[m] = sna_static_stream_add(&general, |
wm_kernels[m].data, |
wm_kernels[m].size, |
64); |
} else { |
wm[m] = sna_static_stream_compile_wm(sna, &general, |
wm_kernels[m].data, |
16); |
} |
} |
state->vs = gen4_create_vs_unit_state(&general); |
state->sf = gen4_create_sf_state(&general, sf); |
wm_state = sna_static_stream_map(&general, |
sizeof(*wm_state) * KERNEL_COUNT * |
FILTER_COUNT * EXTEND_COUNT * |
FILTER_COUNT * EXTEND_COUNT, |
64); |
state->wm = sna_static_stream_offsetof(&general, wm_state); |
for (i = 0; i < FILTER_COUNT; i++) { |
for (j = 0; j < EXTEND_COUNT; j++) { |
for (k = 0; k < FILTER_COUNT; k++) { |
for (l = 0; l < EXTEND_COUNT; l++) { |
uint32_t sampler_state; |
sampler_state = |
gen4_create_sampler_state(&general, |
i, j, |
k, l); |
for (m = 0; m < KERNEL_COUNT; m++) { |
gen4_init_wm_state(&wm_state->state, |
sna->kgem.gen, |
wm_kernels[m].has_mask, |
wm[m], sampler_state); |
wm_state++; |
} |
} |
} |
} |
} |
state->cc = gen4_create_cc_unit_state(&general); |
state->general_bo = sna_static_stream_fini(sna, &general); |
return state->general_bo != NULL; |
} |
const char *gen4_render_init(struct sna *sna, const char *backend) |
{ |
if (!gen4_render_setup(sna)) |
return backend; |
sna->kgem.retire = gen4_render_retire; |
sna->kgem.expire = gen4_render_expire; |
#if 0 |
#if !NO_COMPOSITE |
sna->render.composite = gen4_render_composite; |
sna->render.prefer_gpu |= PREFER_GPU_RENDER; |
#endif |
#if !NO_COMPOSITE_SPANS |
sna->render.check_composite_spans = gen4_check_composite_spans; |
sna->render.composite_spans = gen4_render_composite_spans; |
if (0) |
sna->render.prefer_gpu |= PREFER_GPU_SPANS; |
#endif |
#if !NO_VIDEO |
sna->render.video = gen4_render_video; |
#endif |
#if !NO_COPY_BOXES |
sna->render.copy_boxes = gen4_render_copy_boxes; |
#endif |
#if !NO_COPY |
sna->render.copy = gen4_render_copy; |
#endif |
#if !NO_FILL_BOXES |
sna->render.fill_boxes = gen4_render_fill_boxes; |
#endif |
#if !NO_FILL |
sna->render.fill = gen4_render_fill; |
#endif |
#if !NO_FILL_ONE |
sna->render.fill_one = gen4_render_fill_one; |
#endif |
#endif |
sna->render.blit_tex = gen4_blit_tex; |
sna->render.caps = HW_BIT_BLIT | HW_TEX_BLIT; |
sna->render.flush = gen4_render_flush; |
sna->render.reset = gen4_render_reset; |
sna->render.fini = gen4_render_fini; |
sna->render.max_3d_size = GEN4_MAX_3D_SIZE; |
sna->render.max_3d_pitch = 1 << 18; |
return sna->kgem.gen >= 045 ? "Eaglelake (gen4.5)" : "Broadwater (gen4)"; |
} |
static bool |
gen4_blit_tex(struct sna *sna, |
uint8_t op, bool scale, |
PixmapPtr src, struct kgem_bo *src_bo, |
PixmapPtr mask,struct kgem_bo *mask_bo, |
PixmapPtr dst, struct kgem_bo *dst_bo, |
int32_t src_x, int32_t src_y, |
int32_t msk_x, int32_t msk_y, |
int32_t dst_x, int32_t dst_y, |
int32_t width, int32_t height, |
struct sna_composite_op *tmp) |
{ |
DBG(("%s: %dx%d, current mode=%d\n", __FUNCTION__, |
width, height, sna->kgem.ring)); |
tmp->op = PictOpSrc; |
tmp->dst.pixmap = dst; |
tmp->dst.bo = dst_bo; |
tmp->dst.width = dst->drawable.width; |
tmp->dst.height = dst->drawable.height; |
tmp->dst.format = PICT_a8r8g8b8; |
tmp->src.repeat = RepeatNone; |
tmp->src.filter = PictFilterNearest; |
tmp->src.is_affine = true; |
tmp->src.bo = src_bo; |
tmp->src.pict_format = PICT_x8r8g8b8; |
tmp->src.card_format = gen4_get_card_format(tmp->src.pict_format); |
tmp->src.width = src->drawable.width; |
tmp->src.height = src->drawable.height; |
tmp->is_affine = tmp->src.is_affine; |
tmp->has_component_alpha = false; |
tmp->need_magic_ca_pass = false; |
tmp->mask.repeat = SAMPLER_EXTEND_NONE; |
tmp->mask.filter = SAMPLER_FILTER_NEAREST; |
tmp->mask.is_affine = true; |
tmp->mask.bo = mask_bo; |
tmp->mask.pict_format = PIXMAN_a8; |
tmp->mask.card_format = gen4_get_card_format(tmp->mask.pict_format); |
tmp->mask.width = mask->drawable.width; |
tmp->mask.height = mask->drawable.height; |
if( scale ) |
{ |
tmp->src.scale[0] = 1.f/width; |
tmp->src.scale[1] = 1.f/height; |
} |
else |
{ |
tmp->src.scale[0] = 1.f/src->drawable.width; |
tmp->src.scale[1] = 1.f/src->drawable.height; |
} |
// tmp->src.offset[0] = -dst_x; |
// tmp->src.offset[1] = -dst_y; |
tmp->mask.scale[0] = 1.f/mask->drawable.width; |
tmp->mask.scale[1] = 1.f/mask->drawable.height; |
// tmp->mask.offset[0] = -dst_x; |
// tmp->mask.offset[1] = -dst_y; |
tmp->u.gen4.wm_kernel = WM_KERNEL_MASK; |
// gen4_choose_composite_kernel(tmp->op, |
// tmp->mask.bo != NULL, |
// tmp->has_component_alpha, |
// tmp->is_affine); |
tmp->u.gen4.ve_id = gen4_choose_composite_emitter(sna, tmp); |
tmp->blt = gen4_render_composite_blt; |
tmp->done = gen4_render_composite_done; |
if (!kgem_check_bo(&sna->kgem, |
tmp->dst.bo, tmp->src.bo, tmp->mask.bo, |
NULL)) { |
kgem_submit(&sna->kgem); |
} |
gen4_bind_surfaces(sna, tmp); |
gen4_align_vertex(sna, tmp); |
return true; |
} |
/drivers/video/Intel-2D/sna/gen4_render.h |
---|
0,0 → 1,2693 |
/************************************************************************** |
* |
* Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. |
* All Rights Reserved. |
* |
* Permission is hereby granted, free of charge, to any person obtaining a |
* copy of this software and associated documentation files (the |
* "Software"), to deal in the Software without restriction, including |
* without limitation the rights to use, copy, modify, merge, publish, |
* distribute, sub license, and/or sell copies of the Software, and to |
* permit persons to whom the Software is furnished to do so, subject to |
* the following conditions: |
* |
* The above copyright notice and this permission notice (including the |
* next paragraph) shall be included in all copies or substantial portions |
* of the Software. |
* |
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS |
* OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF |
* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. |
* IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR |
* ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, |
* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE |
* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. |
* |
**************************************************************************/ |
#ifndef GEN4_RENDER_H |
#define GEN4_RENDER_H |
#define GEN4_3D(Pipeline,Opcode,Subopcode) ((3 << 29) | \ |
((Pipeline) << 27) | \ |
((Opcode) << 24) | \ |
((Subopcode) << 16)) |
#define GEN4_URB_FENCE GEN4_3D(0, 0, 0) |
#define GEN4_CS_URB_STATE GEN4_3D(0, 0, 1) |
#define GEN4_CONSTANT_BUFFER GEN4_3D(0, 0, 2) |
#define GEN4_STATE_PREFETCH GEN4_3D(0, 0, 3) |
#define GEN4_STATE_BASE_ADDRESS GEN4_3D(0, 1, 1) |
#define GEN4_STATE_SIP GEN4_3D(0, 1, 2) |
#define GEN4_PIPELINE_SELECT GEN4_3D(0, 1, 4) |
#define NEW_PIPELINE_SELECT GEN4_3D(1, 1, 4) |
#define GEN4_MEDIA_STATE_POINTERS GEN4_3D(2, 0, 0) |
#define GEN4_MEDIA_OBJECT GEN4_3D(2, 1, 0) |
#define GEN4_3DSTATE_PIPELINED_POINTERS GEN4_3D(3, 0, 0) |
#define GEN4_3DSTATE_BINDING_TABLE_POINTERS GEN4_3D(3, 0, 1) |
#define GEN4_3DSTATE_VERTEX_BUFFERS GEN4_3D(3, 0, 8) |
#define GEN4_3DSTATE_VERTEX_ELEMENTS GEN4_3D(3, 0, 9) |
#define GEN4_3DSTATE_INDEX_BUFFER GEN4_3D(3, 0, 0xa) |
#define GEN4_3DSTATE_VF_STATISTICS GEN4_3D(3, 0, 0xb) |
#define GEN4_3DSTATE_DRAWING_RECTANGLE GEN4_3D(3, 1, 0) |
#define GEN4_3DSTATE_CONSTANT_COLOR GEN4_3D(3, 1, 1) |
#define GEN4_3DSTATE_SAMPLER_PALETTE_LOAD GEN4_3D(3, 1, 2) |
#define GEN4_3DSTATE_CHROMA_KEY GEN4_3D(3, 1, 4) |
#define GEN4_3DSTATE_DEPTH_BUFFER GEN4_3D(3, 1, 5) |
# define GEN4_3DSTATE_DEPTH_BUFFER_TYPE_SHIFT 29 |
# define GEN4_3DSTATE_DEPTH_BUFFER_FORMAT_SHIFT 18 |
#define GEN4_3DSTATE_POLY_STIPPLE_OFFSET GEN4_3D(3, 1, 6) |
#define GEN4_3DSTATE_POLY_STIPPLE_PATTERN GEN4_3D(3, 1, 7) |
#define GEN4_3DSTATE_LINE_STIPPLE GEN4_3D(3, 1, 8) |
#define GEN4_3DSTATE_GLOBAL_DEPTH_OFFSET_CLAMP GEN4_3D(3, 1, 9) |
/* These two are BLC and CTG only, not BW or CL */ |
#define GEN4_3DSTATE_AA_LINE_PARAMS GEN4_3D(3, 1, 0xa) |
#define GEN4_3DSTATE_GS_SVB_INDEX GEN4_3D(3, 1, 0xb) |
#define GEN4_PIPE_CONTROL GEN4_3D(3, 2, 0) |
#define GEN4_3DPRIMITIVE GEN4_3D(3, 3, 0) |
#define GEN4_3DSTATE_CLEAR_PARAMS GEN4_3D(3, 1, 0x10) |
/* DW1 */ |
# define GEN4_3DSTATE_DEPTH_CLEAR_VALID (1 << 15) |
#define PIPELINE_SELECT_3D 0 |
#define PIPELINE_SELECT_MEDIA 1 |
#define UF0_CS_REALLOC (1 << 13) |
#define UF0_VFE_REALLOC (1 << 12) |
#define UF0_SF_REALLOC (1 << 11) |
#define UF0_CLIP_REALLOC (1 << 10) |
#define UF0_GS_REALLOC (1 << 9) |
#define UF0_VS_REALLOC (1 << 8) |
#define UF1_CLIP_FENCE_SHIFT 20 |
#define UF1_GS_FENCE_SHIFT 10 |
#define UF1_VS_FENCE_SHIFT 0 |
#define UF2_CS_FENCE_SHIFT 20 |
#define UF2_VFE_FENCE_SHIFT 10 |
#define UF2_SF_FENCE_SHIFT 0 |
/* for GEN4_STATE_BASE_ADDRESS */ |
#define BASE_ADDRESS_MODIFY (1 << 0) |
/* for GEN4_3DSTATE_PIPELINED_POINTERS */ |
#define GEN4_GS_DISABLE 0 |
#define GEN4_GS_ENABLE 1 |
#define GEN4_CLIP_DISABLE 0 |
#define GEN4_CLIP_ENABLE 1 |
/* for GEN4_PIPE_CONTROL */ |
#define GEN4_PIPE_CONTROL_NOWRITE (0 << 14) |
#define GEN4_PIPE_CONTROL_WRITE_QWORD (1 << 14) |
#define GEN4_PIPE_CONTROL_WRITE_DEPTH (2 << 14) |
#define GEN4_PIPE_CONTROL_WRITE_TIME (3 << 14) |
#define GEN4_PIPE_CONTROL_DEPTH_STALL (1 << 13) |
#define GEN4_PIPE_CONTROL_WC_FLUSH (1 << 12) |
#define GEN4_PIPE_CONTROL_IS_FLUSH (1 << 11) |
#define GEN4_PIPE_CONTROL_TC_FLUSH (1 << 10) |
#define GEN4_PIPE_CONTROL_NOTIFY_ENABLE (1 << 8) |
#define GEN4_PIPE_CONTROL_GLOBAL_GTT (1 << 2) |
#define GEN4_PIPE_CONTROL_LOCAL_PGTT (0 << 2) |
#define GEN4_PIPE_CONTROL_DEPTH_CACHE_FLUSH (1 << 0) |
/* VERTEX_BUFFER_STATE Structure */ |
#define VB0_BUFFER_INDEX_SHIFT 27 |
#define VB0_VERTEXDATA (0 << 26) |
#define VB0_INSTANCEDATA (1 << 26) |
#define VB0_BUFFER_PITCH_SHIFT 0 |
/* VERTEX_ELEMENT_STATE Structure */ |
#define VE0_VERTEX_BUFFER_INDEX_SHIFT 27 |
#define VE0_VALID (1 << 26) |
#define VE0_FORMAT_SHIFT 16 |
#define VE0_OFFSET_SHIFT 0 |
#define VE1_VFCOMPONENT_0_SHIFT 28 |
#define VE1_VFCOMPONENT_1_SHIFT 24 |
#define VE1_VFCOMPONENT_2_SHIFT 20 |
#define VE1_VFCOMPONENT_3_SHIFT 16 |
#define VE1_DESTINATION_ELEMENT_OFFSET_SHIFT 0 |
/* 3DPRIMITIVE bits */ |
#define GEN4_3DPRIMITIVE_VERTEX_SEQUENTIAL (0 << 15) |
#define GEN4_3DPRIMITIVE_VERTEX_RANDOM (1 << 15) |
/* Primitive types are in gen4_defines.h */ |
#define GEN4_3DPRIMITIVE_TOPOLOGY_SHIFT 10 |
#define GEN4_SVG_CTL 0x7400 |
#define GEN4_SVG_CTL_GS_BA (0 << 8) |
#define GEN4_SVG_CTL_SS_BA (1 << 8) |
#define GEN4_SVG_CTL_IO_BA (2 << 8) |
#define GEN4_SVG_CTL_GS_AUB (3 << 8) |
#define GEN4_SVG_CTL_IO_AUB (4 << 8) |
#define GEN4_SVG_CTL_SIP (5 << 8) |
#define GEN4_SVG_RDATA 0x7404 |
#define GEN4_SVG_WORK_CTL 0x7408 |
#define GEN4_VF_CTL 0x7500 |
#define GEN4_VF_CTL_SNAPSHOT_COMPLETE (1 << 31) |
#define GEN4_VF_CTL_SNAPSHOT_MUX_SELECT_THREADID (0 << 8) |
#define GEN4_VF_CTL_SNAPSHOT_MUX_SELECT_VF_DEBUG (1 << 8) |
#define GEN4_VF_CTL_SNAPSHOT_TYPE_VERTEX_SEQUENCE (0 << 4) |
#define GEN4_VF_CTL_SNAPSHOT_TYPE_VERTEX_INDEX (1 << 4) |
#define GEN4_VF_CTL_SKIP_INITIAL_PRIMITIVES (1 << 3) |
#define GEN4_VF_CTL_MAX_PRIMITIVES_LIMIT_ENABLE (1 << 2) |
#define GEN4_VF_CTL_VERTEX_RANGE_LIMIT_ENABLE (1 << 1) |
#define GEN4_VF_CTL_SNAPSHOT_ENABLE (1 << 0) |
#define GEN4_VF_STRG_VAL 0x7504 |
#define GEN4_VF_STR_VL_OVR 0x7508 |
#define GEN4_VF_VC_OVR 0x750c |
#define GEN4_VF_STR_PSKIP 0x7510 |
#define GEN4_VF_MAX_PRIM 0x7514 |
#define GEN4_VF_RDATA 0x7518 |
#define GEN4_VS_CTL 0x7600 |
#define GEN4_VS_CTL_SNAPSHOT_COMPLETE (1 << 31) |
#define GEN4_VS_CTL_SNAPSHOT_MUX_VERTEX_0 (0 << 8) |
#define GEN4_VS_CTL_SNAPSHOT_MUX_VERTEX_1 (1 << 8) |
#define GEN4_VS_CTL_SNAPSHOT_MUX_VALID_COUNT (2 << 8) |
#define GEN4_VS_CTL_SNAPSHOT_MUX_VS_KERNEL_POINTER (3 << 8) |
#define GEN4_VS_CTL_SNAPSHOT_ALL_THREADS (1 << 2) |
#define GEN4_VS_CTL_THREAD_SNAPSHOT_ENABLE (1 << 1) |
#define GEN4_VS_CTL_SNAPSHOT_ENABLE (1 << 0) |
#define GEN4_VS_STRG_VAL 0x7604 |
#define GEN4_VS_RDATA 0x7608 |
#define GEN4_SF_CTL 0x7b00 |
#define GEN4_SF_CTL_SNAPSHOT_COMPLETE (1 << 31) |
#define GEN4_SF_CTL_SNAPSHOT_MUX_VERTEX_0_FF_ID (0 << 8) |
#define GEN4_SF_CTL_SNAPSHOT_MUX_VERTEX_0_REL_COUNT (1 << 8) |
#define GEN4_SF_CTL_SNAPSHOT_MUX_VERTEX_1_FF_ID (2 << 8) |
#define GEN4_SF_CTL_SNAPSHOT_MUX_VERTEX_1_REL_COUNT (3 << 8) |
#define GEN4_SF_CTL_SNAPSHOT_MUX_VERTEX_2_FF_ID (4 << 8) |
#define GEN4_SF_CTL_SNAPSHOT_MUX_VERTEX_2_REL_COUNT (5 << 8) |
#define GEN4_SF_CTL_SNAPSHOT_MUX_VERTEX_COUNT (6 << 8) |
#define GEN4_SF_CTL_SNAPSHOT_MUX_SF_KERNEL_POINTER (7 << 8) |
#define GEN4_SF_CTL_MIN_MAX_PRIMITIVE_RANGE_ENABLE (1 << 4) |
#define GEN4_SF_CTL_DEBUG_CLIP_RECTANGLE_ENABLE (1 << 3) |
#define GEN4_SF_CTL_SNAPSHOT_ALL_THREADS (1 << 2) |
#define GEN4_SF_CTL_THREAD_SNAPSHOT_ENABLE (1 << 1) |
#define GEN4_SF_CTL_SNAPSHOT_ENABLE (1 << 0) |
#define GEN4_SF_STRG_VAL 0x7b04 |
#define GEN4_SF_RDATA 0x7b18 |
#define GEN4_WIZ_CTL 0x7c00 |
#define GEN4_WIZ_CTL_SNAPSHOT_COMPLETE (1 << 31) |
#define GEN4_WIZ_CTL_SUBSPAN_INSTANCE_SHIFT 16 |
#define GEN4_WIZ_CTL_SNAPSHOT_MUX_WIZ_KERNEL_POINTER (0 << 8) |
#define GEN4_WIZ_CTL_SNAPSHOT_MUX_SUBSPAN_INSTANCE (1 << 8) |
#define GEN4_WIZ_CTL_SNAPSHOT_MUX_PRIMITIVE_SEQUENCE (2 << 8) |
#define GEN4_WIZ_CTL_SINGLE_SUBSPAN_DISPATCH (1 << 6) |
#define GEN4_WIZ_CTL_IGNORE_COLOR_SCOREBOARD_STALLS (1 << 5) |
#define GEN4_WIZ_CTL_ENABLE_SUBSPAN_INSTANCE_COMPARE (1 << 4) |
#define GEN4_WIZ_CTL_USE_UPSTREAM_SNAPSHOT_FLAG (1 << 3) |
#define GEN4_WIZ_CTL_SNAPSHOT_ALL_THREADS (1 << 2) |
#define GEN4_WIZ_CTL_THREAD_SNAPSHOT_ENABLE (1 << 1) |
#define GEN4_WIZ_CTL_SNAPSHOT_ENABLE (1 << 0) |
#define GEN4_WIZ_STRG_VAL 0x7c04 |
#define GEN4_WIZ_RDATA 0x7c18 |
#define GEN4_TS_CTL 0x7e00 |
#define GEN4_TS_CTL_SNAPSHOT_COMPLETE (1 << 31) |
#define GEN4_TS_CTL_SNAPSHOT_MESSAGE_ERROR (0 << 8) |
#define GEN4_TS_CTL_SNAPSHOT_INTERFACE_DESCRIPTOR (3 << 8) |
#define GEN4_TS_CTL_SNAPSHOT_ALL_CHILD_THREADS (1 << 2) |
#define GEN4_TS_CTL_SNAPSHOT_ALL_ROOT_THREADS (1 << 1) |
#define GEN4_TS_CTL_SNAPSHOT_ENABLE (1 << 0) |
#define GEN4_TS_STRG_VAL 0x7e04 |
#define GEN4_TS_RDATA 0x7e08 |
#define GEN4_TD_CTL 0x8000 |
#define GEN4_TD_CTL_MUX_SHIFT 8 |
#define GEN4_TD_CTL_EXTERNAL_HALT_R0_DEBUG_MATCH (1 << 7) |
#define GEN4_TD_CTL_FORCE_EXTERNAL_HALT (1 << 6) |
#define GEN4_TD_CTL_EXCEPTION_MASK_OVERRIDE (1 << 5) |
#define GEN4_TD_CTL_FORCE_THREAD_BREAKPOINT_ENABLE (1 << 4) |
#define GEN4_TD_CTL_BREAKPOINT_ENABLE (1 << 2) |
#define GEN4_TD_CTL2 0x8004 |
#define GEN4_TD_CTL2_ILLEGAL_OPCODE_EXCEPTION_OVERRIDE (1 << 28) |
#define GEN4_TD_CTL2_MASKSTACK_EXCEPTION_OVERRIDE (1 << 26) |
#define GEN4_TD_CTL2_SOFTWARE_EXCEPTION_OVERRIDE (1 << 25) |
#define GEN4_TD_CTL2_ACTIVE_THREAD_LIMIT_SHIFT 16 |
#define GEN4_TD_CTL2_ACTIVE_THREAD_LIMIT_ENABLE (1 << 8) |
#define GEN4_TD_CTL2_THREAD_SPAWNER_EXECUTION_MASK_ENABLE (1 << 7) |
#define GEN4_TD_CTL2_WIZ_EXECUTION_MASK_ENABLE (1 << 6) |
#define GEN4_TD_CTL2_SF_EXECUTION_MASK_ENABLE (1 << 5) |
#define GEN4_TD_CTL2_CLIPPER_EXECUTION_MASK_ENABLE (1 << 4) |
#define GEN4_TD_CTL2_GS_EXECUTION_MASK_ENABLE (1 << 3) |
#define GEN4_TD_CTL2_VS_EXECUTION_MASK_ENABLE (1 << 0) |
#define GEN4_TD_VF_VS_EMSK 0x8008 |
#define GEN4_TD_GS_EMSK 0x800c |
#define GEN4_TD_CLIP_EMSK 0x8010 |
#define GEN4_TD_SF_EMSK 0x8014 |
#define GEN4_TD_WIZ_EMSK 0x8018 |
#define GEN4_TD_0_6_EHTRG_VAL 0x801c |
#define GEN4_TD_0_7_EHTRG_VAL 0x8020 |
#define GEN4_TD_0_6_EHTRG_MSK 0x8024 |
#define GEN4_TD_0_7_EHTRG_MSK 0x8028 |
#define GEN4_TD_RDATA 0x802c |
#define GEN4_TD_TS_EMSK 0x8030 |
#define GEN4_EU_CTL 0x8800 |
#define GEN4_EU_CTL_SELECT_SHIFT 16 |
#define GEN4_EU_CTL_DATA_MUX_SHIFT 8 |
#define GEN4_EU_ATT_0 0x8810 |
#define GEN4_EU_ATT_1 0x8814 |
#define GEN4_EU_ATT_DATA_0 0x8820 |
#define GEN4_EU_ATT_DATA_1 0x8824 |
#define GEN4_EU_ATT_CLR_0 0x8830 |
#define GEN4_EU_ATT_CLR_1 0x8834 |
#define GEN4_EU_RDATA 0x8840 |
/* 3D state: |
*/ |
#define _3DOP_3DSTATE_PIPELINED 0x0 |
#define _3DOP_3DSTATE_NONPIPELINED 0x1 |
#define _3DOP_3DCONTROL 0x2 |
#define _3DOP_3DPRIMITIVE 0x3 |
#define _3DSTATE_PIPELINED_POINTERS 0x00 |
#define _3DSTATE_BINDING_TABLE_POINTERS 0x01 |
#define _3DSTATE_VERTEX_BUFFERS 0x08 |
#define _3DSTATE_VERTEX_ELEMENTS 0x09 |
#define _3DSTATE_INDEX_BUFFER 0x0A |
#define _3DSTATE_VF_STATISTICS 0x0B |
#define _3DSTATE_DRAWING_RECTANGLE 0x00 |
#define _3DSTATE_CONSTANT_COLOR 0x01 |
#define _3DSTATE_SAMPLER_PALETTE_LOAD 0x02 |
#define _3DSTATE_CHROMA_KEY 0x04 |
#define _3DSTATE_DEPTH_BUFFER 0x05 |
#define _3DSTATE_POLY_STIPPLE_OFFSET 0x06 |
#define _3DSTATE_POLY_STIPPLE_PATTERN 0x07 |
#define _3DSTATE_LINE_STIPPLE 0x08 |
#define _3DSTATE_GLOBAL_DEPTH_OFFSET_CLAMP 0x09 |
#define _3DCONTROL 0x00 |
#define _3DPRIMITIVE 0x00 |
#define _3DPRIM_POINTLIST 0x01 |
#define _3DPRIM_LINELIST 0x02 |
#define _3DPRIM_LINESTRIP 0x03 |
#define _3DPRIM_TRILIST 0x04 |
#define _3DPRIM_TRISTRIP 0x05 |
#define _3DPRIM_TRIFAN 0x06 |
#define _3DPRIM_QUADLIST 0x07 |
#define _3DPRIM_QUADSTRIP 0x08 |
#define _3DPRIM_LINELIST_ADJ 0x09 |
#define _3DPRIM_LINESTRIP_ADJ 0x0A |
#define _3DPRIM_TRILIST_ADJ 0x0B |
#define _3DPRIM_TRISTRIP_ADJ 0x0C |
#define _3DPRIM_TRISTRIP_REVERSE 0x0D |
#define _3DPRIM_POLYGON 0x0E |
#define _3DPRIM_RECTLIST 0x0F |
#define _3DPRIM_LINELOOP 0x10 |
#define _3DPRIM_POINTLIST_BF 0x11 |
#define _3DPRIM_LINESTRIP_CONT 0x12 |
#define _3DPRIM_LINESTRIP_BF 0x13 |
#define _3DPRIM_LINESTRIP_CONT_BF 0x14 |
#define _3DPRIM_TRIFAN_NOSTIPPLE 0x15 |
#define _3DPRIM_VERTEXBUFFER_ACCESS_SEQUENTIAL 0 |
#define _3DPRIM_VERTEXBUFFER_ACCESS_RANDOM 1 |
#define GEN4_ANISORATIO_2 0 |
#define GEN4_ANISORATIO_4 1 |
#define GEN4_ANISORATIO_6 2 |
#define GEN4_ANISORATIO_8 3 |
#define GEN4_ANISORATIO_10 4 |
#define GEN4_ANISORATIO_12 5 |
#define GEN4_ANISORATIO_14 6 |
#define GEN4_ANISORATIO_16 7 |
#define GEN4_BLENDFACTOR_ONE 0x1 |
#define GEN4_BLENDFACTOR_SRC_COLOR 0x2 |
#define GEN4_BLENDFACTOR_SRC_ALPHA 0x3 |
#define GEN4_BLENDFACTOR_DST_ALPHA 0x4 |
#define GEN4_BLENDFACTOR_DST_COLOR 0x5 |
#define GEN4_BLENDFACTOR_SRC_ALPHA_SATURATE 0x6 |
#define GEN4_BLENDFACTOR_CONST_COLOR 0x7 |
#define GEN4_BLENDFACTOR_CONST_ALPHA 0x8 |
#define GEN4_BLENDFACTOR_SRC1_COLOR 0x9 |
#define GEN4_BLENDFACTOR_SRC1_ALPHA 0x0A |
#define GEN4_BLENDFACTOR_ZERO 0x11 |
#define GEN4_BLENDFACTOR_INV_SRC_COLOR 0x12 |
#define GEN4_BLENDFACTOR_INV_SRC_ALPHA 0x13 |
#define GEN4_BLENDFACTOR_INV_DST_ALPHA 0x14 |
#define GEN4_BLENDFACTOR_INV_DST_COLOR 0x15 |
#define GEN4_BLENDFACTOR_INV_CONST_COLOR 0x17 |
#define GEN4_BLENDFACTOR_INV_CONST_ALPHA 0x18 |
#define GEN4_BLENDFACTOR_INV_SRC1_COLOR 0x19 |
#define GEN4_BLENDFACTOR_INV_SRC1_ALPHA 0x1A |
#define GEN4_BLENDFUNCTION_ADD 0 |
#define GEN4_BLENDFUNCTION_SUBTRACT 1 |
#define GEN4_BLENDFUNCTION_REVERSE_SUBTRACT 2 |
#define GEN4_BLENDFUNCTION_MIN 3 |
#define GEN4_BLENDFUNCTION_MAX 4 |
#define GEN4_ALPHATEST_FORMAT_UNORM8 0 |
#define GEN4_ALPHATEST_FORMAT_FLOAT32 1 |
#define GEN4_CHROMAKEY_KILL_ON_ANY_MATCH 0 |
#define GEN4_CHROMAKEY_REPLACE_BLACK 1 |
#define GEN4_CLIP_API_OGL 0 |
#define GEN4_CLIP_API_DX 1 |
#define GEN4_CLIPMODE_NORMAL 0 |
#define GEN4_CLIPMODE_CLIP_ALL 1 |
#define GEN4_CLIPMODE_CLIP_NON_REJECTED 2 |
#define GEN4_CLIPMODE_REJECT_ALL 3 |
#define GEN4_CLIPMODE_ACCEPT_ALL 4 |
#define GEN4_CLIP_NDCSPACE 0 |
#define GEN4_CLIP_SCREENSPACE 1 |
#define GEN4_COMPAREFUNCTION_ALWAYS 0 |
#define GEN4_COMPAREFUNCTION_NEVER 1 |
#define GEN4_COMPAREFUNCTION_LESS 2 |
#define GEN4_COMPAREFUNCTION_EQUAL 3 |
#define GEN4_COMPAREFUNCTION_LEQUAL 4 |
#define GEN4_COMPAREFUNCTION_GREATER 5 |
#define GEN4_COMPAREFUNCTION_NOTEQUAL 6 |
#define GEN4_COMPAREFUNCTION_GEQUAL 7 |
#define GEN4_COVERAGE_PIXELS_HALF 0 |
#define GEN4_COVERAGE_PIXELS_1 1 |
#define GEN4_COVERAGE_PIXELS_2 2 |
#define GEN4_COVERAGE_PIXELS_4 3 |
#define GEN4_CULLMODE_BOTH 0 |
#define GEN4_CULLMODE_NONE 1 |
#define GEN4_CULLMODE_FRONT 2 |
#define GEN4_CULLMODE_BACK 3 |
#define GEN4_DEFAULTCOLOR_R8G8B8A8_UNORM 0 |
#define GEN4_DEFAULTCOLOR_R32G32B32A32_FLOAT 1 |
#define GEN4_DEPTHFORMAT_D32_FLOAT_S8X24_UINT 0 |
#define GEN4_DEPTHFORMAT_D32_FLOAT 1 |
#define GEN4_DEPTHFORMAT_D24_UNORM_S8_UINT 2 |
#define GEN4_DEPTHFORMAT_D16_UNORM 5 |
#define GEN4_FLOATING_POINT_IEEE_754 0 |
#define GEN4_FLOATING_POINT_NON_IEEE_754 1 |
#define GEN4_FRONTWINDING_CW 0 |
#define GEN4_FRONTWINDING_CCW 1 |
#define GEN4_INDEX_BYTE 0 |
#define GEN4_INDEX_WORD 1 |
#define GEN4_INDEX_DWORD 2 |
#define GEN4_LOGICOPFUNCTION_CLEAR 0 |
#define GEN4_LOGICOPFUNCTION_NOR 1 |
#define GEN4_LOGICOPFUNCTION_AND_INVERTED 2 |
#define GEN4_LOGICOPFUNCTION_COPY_INVERTED 3 |
#define GEN4_LOGICOPFUNCTION_AND_REVERSE 4 |
#define GEN4_LOGICOPFUNCTION_INVERT 5 |
#define GEN4_LOGICOPFUNCTION_XOR 6 |
#define GEN4_LOGICOPFUNCTION_NAND 7 |
#define GEN4_LOGICOPFUNCTION_AND 8 |
#define GEN4_LOGICOPFUNCTION_EQUIV 9 |
#define GEN4_LOGICOPFUNCTION_NOOP 10 |
#define GEN4_LOGICOPFUNCTION_OR_INVERTED 11 |
#define GEN4_LOGICOPFUNCTION_COPY 12 |
#define GEN4_LOGICOPFUNCTION_OR_REVERSE 13 |
#define GEN4_LOGICOPFUNCTION_OR 14 |
#define GEN4_LOGICOPFUNCTION_SET 15 |
#define GEN4_MAPFILTER_NEAREST 0x0 |
#define GEN4_MAPFILTER_LINEAR 0x1 |
#define GEN4_MAPFILTER_ANISOTROPIC 0x2 |
#define GEN4_MIPFILTER_NONE 0 |
#define GEN4_MIPFILTER_NEAREST 1 |
#define GEN4_MIPFILTER_LINEAR 3 |
#define GEN4_POLYGON_FRONT_FACING 0 |
#define GEN4_POLYGON_BACK_FACING 1 |
#define GEN4_PREFILTER_ALWAYS 0x0 |
#define GEN4_PREFILTER_NEVER 0x1 |
#define GEN4_PREFILTER_LESS 0x2 |
#define GEN4_PREFILTER_EQUAL 0x3 |
#define GEN4_PREFILTER_LEQUAL 0x4 |
#define GEN4_PREFILTER_GREATER 0x5 |
#define GEN4_PREFILTER_NOTEQUAL 0x6 |
#define GEN4_PREFILTER_GEQUAL 0x7 |
#define GEN4_PROVOKING_VERTEX_0 0 |
#define GEN4_PROVOKING_VERTEX_1 1 |
#define GEN4_PROVOKING_VERTEX_2 2 |
#define GEN4_RASTRULE_UPPER_LEFT 0 |
#define GEN4_RASTRULE_UPPER_RIGHT 1 |
#define GEN4_RENDERTARGET_CLAMPRANGE_UNORM 0 |
#define GEN4_RENDERTARGET_CLAMPRANGE_SNORM 1 |
#define GEN4_RENDERTARGET_CLAMPRANGE_FORMAT 2 |
#define GEN4_STENCILOP_KEEP 0 |
#define GEN4_STENCILOP_ZERO 1 |
#define GEN4_STENCILOP_REPLACE 2 |
#define GEN4_STENCILOP_INCRSAT 3 |
#define GEN4_STENCILOP_DECRSAT 4 |
#define GEN4_STENCILOP_INCR 5 |
#define GEN4_STENCILOP_DECR 6 |
#define GEN4_STENCILOP_INVERT 7 |
#define GEN4_SURFACE_MIPMAPLAYOUT_BELOW 0 |
#define GEN4_SURFACE_MIPMAPLAYOUT_RIGHT 1 |
#define GEN4_SURFACEFORMAT_R32G32B32A32_FLOAT 0x000 |
#define GEN4_SURFACEFORMAT_R32G32B32A32_SINT 0x001 |
#define GEN4_SURFACEFORMAT_R32G32B32A32_UINT 0x002 |
#define GEN4_SURFACEFORMAT_R32G32B32A32_UNORM 0x003 |
#define GEN4_SURFACEFORMAT_R32G32B32A32_SNORM 0x004 |
#define GEN4_SURFACEFORMAT_R64G64_FLOAT 0x005 |
#define GEN4_SURFACEFORMAT_R32G32B32X32_FLOAT 0x006 |
#define GEN4_SURFACEFORMAT_R32G32B32A32_SSCALED 0x007 |
#define GEN4_SURFACEFORMAT_R32G32B32A32_USCALED 0x008 |
#define GEN4_SURFACEFORMAT_R32G32B32_FLOAT 0x040 |
#define GEN4_SURFACEFORMAT_R32G32B32_SINT 0x041 |
#define GEN4_SURFACEFORMAT_R32G32B32_UINT 0x042 |
#define GEN4_SURFACEFORMAT_R32G32B32_UNORM 0x043 |
#define GEN4_SURFACEFORMAT_R32G32B32_SNORM 0x044 |
#define GEN4_SURFACEFORMAT_R32G32B32_SSCALED 0x045 |
#define GEN4_SURFACEFORMAT_R32G32B32_USCALED 0x046 |
#define GEN4_SURFACEFORMAT_R16G16B16A16_UNORM 0x080 |
#define GEN4_SURFACEFORMAT_R16G16B16A16_SNORM 0x081 |
#define GEN4_SURFACEFORMAT_R16G16B16A16_SINT 0x082 |
#define GEN4_SURFACEFORMAT_R16G16B16A16_UINT 0x083 |
#define GEN4_SURFACEFORMAT_R16G16B16A16_FLOAT 0x084 |
#define GEN4_SURFACEFORMAT_R32G32_FLOAT 0x085 |
#define GEN4_SURFACEFORMAT_R32G32_SINT 0x086 |
#define GEN4_SURFACEFORMAT_R32G32_UINT 0x087 |
#define GEN4_SURFACEFORMAT_R32_FLOAT_X8X24_TYPELESS 0x088 |
#define GEN4_SURFACEFORMAT_X32_TYPELESS_G8X24_UINT 0x089 |
#define GEN4_SURFACEFORMAT_L32A32_FLOAT 0x08A |
#define GEN4_SURFACEFORMAT_R32G32_UNORM 0x08B |
#define GEN4_SURFACEFORMAT_R32G32_SNORM 0x08C |
#define GEN4_SURFACEFORMAT_R64_FLOAT 0x08D |
#define GEN4_SURFACEFORMAT_R16G16B16X16_UNORM 0x08E |
#define GEN4_SURFACEFORMAT_R16G16B16X16_FLOAT 0x08F |
#define GEN4_SURFACEFORMAT_A32X32_FLOAT 0x090 |
#define GEN4_SURFACEFORMAT_L32X32_FLOAT 0x091 |
#define GEN4_SURFACEFORMAT_I32X32_FLOAT 0x092 |
#define GEN4_SURFACEFORMAT_R16G16B16A16_SSCALED 0x093 |
#define GEN4_SURFACEFORMAT_R16G16B16A16_USCALED 0x094 |
#define GEN4_SURFACEFORMAT_R32G32_SSCALED 0x095 |
#define GEN4_SURFACEFORMAT_R32G32_USCALED 0x096 |
#define GEN4_SURFACEFORMAT_B8G8R8A8_UNORM 0x0C0 |
#define GEN4_SURFACEFORMAT_B8G8R8A8_UNORM_SRGB 0x0C1 |
#define GEN4_SURFACEFORMAT_R10G10B10A2_UNORM 0x0C2 |
#define GEN4_SURFACEFORMAT_R10G10B10A2_UNORM_SRGB 0x0C3 |
#define GEN4_SURFACEFORMAT_R10G10B10A2_UINT 0x0C4 |
#define GEN4_SURFACEFORMAT_R10G10B10_SNORM_A2_UNORM 0x0C5 |
#define GEN4_SURFACEFORMAT_R8G8B8A8_UNORM 0x0C7 |
#define GEN4_SURFACEFORMAT_R8G8B8A8_UNORM_SRGB 0x0C8 |
#define GEN4_SURFACEFORMAT_R8G8B8A8_SNORM 0x0C9 |
#define GEN4_SURFACEFORMAT_R8G8B8A8_SINT 0x0CA |
#define GEN4_SURFACEFORMAT_R8G8B8A8_UINT 0x0CB |
#define GEN4_SURFACEFORMAT_R16G16_UNORM 0x0CC |
#define GEN4_SURFACEFORMAT_R16G16_SNORM 0x0CD |
#define GEN4_SURFACEFORMAT_R16G16_SINT 0x0CE |
#define GEN4_SURFACEFORMAT_R16G16_UINT 0x0CF |
#define GEN4_SURFACEFORMAT_R16G16_FLOAT 0x0D0 |
#define GEN4_SURFACEFORMAT_B10G10R10A2_UNORM 0x0D1 |
#define GEN4_SURFACEFORMAT_B10G10R10A2_UNORM_SRGB 0x0D2 |
#define GEN4_SURFACEFORMAT_R11G11B10_FLOAT 0x0D3 |
#define GEN4_SURFACEFORMAT_R32_SINT 0x0D6 |
#define GEN4_SURFACEFORMAT_R32_UINT 0x0D7 |
#define GEN4_SURFACEFORMAT_R32_FLOAT 0x0D8 |
#define GEN4_SURFACEFORMAT_R24_UNORM_X8_TYPELESS 0x0D9 |
#define GEN4_SURFACEFORMAT_X24_TYPELESS_G8_UINT 0x0DA |
#define GEN4_SURFACEFORMAT_L16A16_UNORM 0x0DF |
#define GEN4_SURFACEFORMAT_I24X8_UNORM 0x0E0 |
#define GEN4_SURFACEFORMAT_L24X8_UNORM 0x0E1 |
#define GEN4_SURFACEFORMAT_A24X8_UNORM 0x0E2 |
#define GEN4_SURFACEFORMAT_I32_FLOAT 0x0E3 |
#define GEN4_SURFACEFORMAT_L32_FLOAT 0x0E4 |
#define GEN4_SURFACEFORMAT_A32_FLOAT 0x0E5 |
#define GEN4_SURFACEFORMAT_B8G8R8X8_UNORM 0x0E9 |
#define GEN4_SURFACEFORMAT_B8G8R8X8_UNORM_SRGB 0x0EA |
#define GEN4_SURFACEFORMAT_R8G8B8X8_UNORM 0x0EB |
#define GEN4_SURFACEFORMAT_R8G8B8X8_UNORM_SRGB 0x0EC |
#define GEN4_SURFACEFORMAT_R9G9B9E5_SHAREDEXP 0x0ED |
#define GEN4_SURFACEFORMAT_B10G10R10X2_UNORM 0x0EE |
#define GEN4_SURFACEFORMAT_L16A16_FLOAT 0x0F0 |
#define GEN4_SURFACEFORMAT_R32_UNORM 0x0F1 |
#define GEN4_SURFACEFORMAT_R32_SNORM 0x0F2 |
#define GEN4_SURFACEFORMAT_R10G10B10X2_USCALED 0x0F3 |
#define GEN4_SURFACEFORMAT_R8G8B8A8_SSCALED 0x0F4 |
#define GEN4_SURFACEFORMAT_R8G8B8A8_USCALED 0x0F5 |
#define GEN4_SURFACEFORMAT_R16G16_SSCALED 0x0F6 |
#define GEN4_SURFACEFORMAT_R16G16_USCALED 0x0F7 |
#define GEN4_SURFACEFORMAT_R32_SSCALED 0x0F8 |
#define GEN4_SURFACEFORMAT_R32_USCALED 0x0F9 |
#define GEN4_SURFACEFORMAT_B5G6R5_UNORM 0x100 |
#define GEN4_SURFACEFORMAT_B5G6R5_UNORM_SRGB 0x101 |
#define GEN4_SURFACEFORMAT_B5G5R5A1_UNORM 0x102 |
#define GEN4_SURFACEFORMAT_B5G5R5A1_UNORM_SRGB 0x103 |
#define GEN4_SURFACEFORMAT_B4G4R4A4_UNORM 0x104 |
#define GEN4_SURFACEFORMAT_B4G4R4A4_UNORM_SRGB 0x105 |
#define GEN4_SURFACEFORMAT_R8G8_UNORM 0x106 |
#define GEN4_SURFACEFORMAT_R8G8_SNORM 0x107 |
#define GEN4_SURFACEFORMAT_R8G8_SINT 0x108 |
#define GEN4_SURFACEFORMAT_R8G8_UINT 0x109 |
#define GEN4_SURFACEFORMAT_R16_UNORM 0x10A |
#define GEN4_SURFACEFORMAT_R16_SNORM 0x10B |
#define GEN4_SURFACEFORMAT_R16_SINT 0x10C |
#define GEN4_SURFACEFORMAT_R16_UINT 0x10D |
#define GEN4_SURFACEFORMAT_R16_FLOAT 0x10E |
#define GEN4_SURFACEFORMAT_I16_UNORM 0x111 |
#define GEN4_SURFACEFORMAT_L16_UNORM 0x112 |
#define GEN4_SURFACEFORMAT_A16_UNORM 0x113 |
#define GEN4_SURFACEFORMAT_L8A8_UNORM 0x114 |
#define GEN4_SURFACEFORMAT_I16_FLOAT 0x115 |
#define GEN4_SURFACEFORMAT_L16_FLOAT 0x116 |
#define GEN4_SURFACEFORMAT_A16_FLOAT 0x117 |
#define GEN4_SURFACEFORMAT_R5G5_SNORM_B6_UNORM 0x119 |
#define GEN4_SURFACEFORMAT_B5G5R5X1_UNORM 0x11A |
#define GEN4_SURFACEFORMAT_B5G5R5X1_UNORM_SRGB 0x11B |
#define GEN4_SURFACEFORMAT_R8G8_SSCALED 0x11C |
#define GEN4_SURFACEFORMAT_R8G8_USCALED 0x11D |
#define GEN4_SURFACEFORMAT_R16_SSCALED 0x11E |
#define GEN4_SURFACEFORMAT_R16_USCALED 0x11F |
#define GEN4_SURFACEFORMAT_R8_UNORM 0x140 |
#define GEN4_SURFACEFORMAT_R8_SNORM 0x141 |
#define GEN4_SURFACEFORMAT_R8_SINT 0x142 |
#define GEN4_SURFACEFORMAT_R8_UINT 0x143 |
#define GEN4_SURFACEFORMAT_A8_UNORM 0x144 |
#define GEN4_SURFACEFORMAT_I8_UNORM 0x145 |
#define GEN4_SURFACEFORMAT_L8_UNORM 0x146 |
#define GEN4_SURFACEFORMAT_P4A4_UNORM 0x147 |
#define GEN4_SURFACEFORMAT_A4P4_UNORM 0x148 |
#define GEN4_SURFACEFORMAT_R8_SSCALED 0x149 |
#define GEN4_SURFACEFORMAT_R8_USCALED 0x14A |
#define GEN4_SURFACEFORMAT_R1_UINT 0x181 |
#define GEN4_SURFACEFORMAT_YCRCB_NORMAL 0x182 |
#define GEN4_SURFACEFORMAT_YCRCB_SWAPUVY 0x183 |
#define GEN4_SURFACEFORMAT_BC1_UNORM 0x186 |
#define GEN4_SURFACEFORMAT_BC2_UNORM 0x187 |
#define GEN4_SURFACEFORMAT_BC3_UNORM 0x188 |
#define GEN4_SURFACEFORMAT_BC4_UNORM 0x189 |
#define GEN4_SURFACEFORMAT_BC5_UNORM 0x18A |
#define GEN4_SURFACEFORMAT_BC1_UNORM_SRGB 0x18B |
#define GEN4_SURFACEFORMAT_BC2_UNORM_SRGB 0x18C |
#define GEN4_SURFACEFORMAT_BC3_UNORM_SRGB 0x18D |
#define GEN4_SURFACEFORMAT_MONO8 0x18E |
#define GEN4_SURFACEFORMAT_YCRCB_SWAPUV 0x18F |
#define GEN4_SURFACEFORMAT_YCRCB_SWAPY 0x190 |
#define GEN4_SURFACEFORMAT_DXT1_RGB 0x191 |
#define GEN4_SURFACEFORMAT_FXT1 0x192 |
#define GEN4_SURFACEFORMAT_R8G8B8_UNORM 0x193 |
#define GEN4_SURFACEFORMAT_R8G8B8_SNORM 0x194 |
#define GEN4_SURFACEFORMAT_R8G8B8_SSCALED 0x195 |
#define GEN4_SURFACEFORMAT_R8G8B8_USCALED 0x196 |
#define GEN4_SURFACEFORMAT_R64G64B64A64_FLOAT 0x197 |
#define GEN4_SURFACEFORMAT_R64G64B64_FLOAT 0x198 |
#define GEN4_SURFACEFORMAT_BC4_SNORM 0x199 |
#define GEN4_SURFACEFORMAT_BC5_SNORM 0x19A |
#define GEN4_SURFACEFORMAT_R16G16B16_UNORM 0x19C |
#define GEN4_SURFACEFORMAT_R16G16B16_SNORM 0x19D |
#define GEN4_SURFACEFORMAT_R16G16B16_SSCALED 0x19E |
#define GEN4_SURFACEFORMAT_R16G16B16_USCALED 0x19F |
#define GEN4_SURFACERETURNFORMAT_FLOAT32 0 |
#define GEN4_SURFACERETURNFORMAT_S1 1 |
#define GEN4_SURFACE_1D 0 |
#define GEN4_SURFACE_2D 1 |
#define GEN4_SURFACE_3D 2 |
#define GEN4_SURFACE_CUBE 3 |
#define GEN4_SURFACE_BUFFER 4 |
#define GEN4_SURFACE_NULL 7 |
#define GEN4_BORDER_COLOR_MODE_DEFAULT 0 |
#define GEN4_BORDER_COLOR_MODE_LEGACY 1 |
#define GEN4_TEXCOORDMODE_WRAP 0 |
#define GEN4_TEXCOORDMODE_MIRROR 1 |
#define GEN4_TEXCOORDMODE_CLAMP 2 |
#define GEN4_TEXCOORDMODE_CUBE 3 |
#define GEN4_TEXCOORDMODE_CLAMP_BORDER 4 |
#define GEN4_TEXCOORDMODE_MIRROR_ONCE 5 |
#define GEN4_THREAD_PRIORITY_NORMAL 0 |
#define GEN4_THREAD_PRIORITY_HIGH 1 |
#define GEN4_TILEWALK_XMAJOR 0 |
#define GEN4_TILEWALK_YMAJOR 1 |
#define GEN4_VERTEX_SUBPIXEL_PRECISION_8BITS 0 |
#define GEN4_VERTEX_SUBPIXEL_PRECISION_4BITS 1 |
#define GEN4_VERTEXBUFFER_ACCESS_VERTEXDATA 0 |
#define GEN4_VERTEXBUFFER_ACCESS_INSTANCEDATA 1 |
#define VFCOMPONENT_NOSTORE 0 |
#define VFCOMPONENT_STORE_SRC 1 |
#define VFCOMPONENT_STORE_0 2 |
#define VFCOMPONENT_STORE_1_FLT 3 |
#define VFCOMPONENT_STORE_1_INT 4 |
#define VFCOMPONENT_STORE_VID 5 |
#define VFCOMPONENT_STORE_IID 6 |
#define VFCOMPONENT_STORE_PID 7 |
/* Execution Unit (EU) defines |
*/ |
#define GEN4_ALIGN_1 0 |
#define GEN4_ALIGN_16 1 |
#define GEN4_ADDRESS_DIRECT 0 |
#define GEN4_ADDRESS_REGISTER_INDIRECT_REGISTER 1 |
#define GEN4_CHANNEL_X 0 |
#define GEN4_CHANNEL_Y 1 |
#define GEN4_CHANNEL_Z 2 |
#define GEN4_CHANNEL_W 3 |
#define GEN4_COMPRESSION_NONE 0 |
#define GEN4_COMPRESSION_2NDHALF 1 |
#define GEN4_COMPRESSION_COMPRESSED 2 |
#define GEN4_CONDITIONAL_NONE 0 |
#define GEN4_CONDITIONAL_Z 1 |
#define GEN4_CONDITIONAL_NZ 2 |
#define GEN4_CONDITIONAL_EQ 1 /* Z */ |
#define GEN4_CONDITIONAL_NEQ 2 /* NZ */ |
#define GEN4_CONDITIONAL_G 3 |
#define GEN4_CONDITIONAL_GE 4 |
#define GEN4_CONDITIONAL_L 5 |
#define GEN4_CONDITIONAL_LE 6 |
#define GEN4_CONDITIONAL_C 7 |
#define GEN4_CONDITIONAL_O 8 |
#define GEN4_DEBUG_NONE 0 |
#define GEN4_DEBUG_BREAKPOINT 1 |
#define GEN4_DEPENDENCY_NORMAL 0 |
#define GEN4_DEPENDENCY_NOTCLEARED 1 |
#define GEN4_DEPENDENCY_NOTCHECKED 2 |
#define GEN4_DEPENDENCY_DISABLE 3 |
#define GEN4_EXECUTE_1 0 |
#define GEN4_EXECUTE_2 1 |
#define GEN4_EXECUTE_4 2 |
#define GEN4_EXECUTE_8 3 |
#define GEN4_EXECUTE_16 4 |
#define GEN4_EXECUTE_32 5 |
#define GEN4_HORIZONTAL_STRIDE_0 0 |
#define GEN4_HORIZONTAL_STRIDE_1 1 |
#define GEN4_HORIZONTAL_STRIDE_2 2 |
#define GEN4_HORIZONTAL_STRIDE_4 3 |
#define GEN4_INSTRUCTION_NORMAL 0 |
#define GEN4_INSTRUCTION_SATURATE 1 |
#define _MASK_ENABLE 0 |
#define _MASK_DISABLE 1 |
#define GEN4_OPCODE_MOV 1 |
#define GEN4_OPCODE_SEL 2 |
#define GEN4_OPCODE_NOT 4 |
#define GEN4_OPCODE_AND 5 |
#define GEN4_OPCODE_OR 6 |
#define GEN4_OPCODE_XOR 7 |
#define GEN4_OPCODE_SHR 8 |
#define GEN4_OPCODE_SHL 9 |
#define GEN4_OPCODE_RSR 10 |
#define GEN4_OPCODE_RSL 11 |
#define GEN4_OPCODE_ASR 12 |
#define GEN4_OPCODE_CMP 16 |
#define GEN4_OPCODE_JMPI 32 |
#define GEN4_OPCODE_IF 34 |
#define GEN4_OPCODE_IFF 35 |
#define GEN4_OPCODE_ELSE 36 |
#define GEN4_OPCODE_ENDIF 37 |
#define GEN4_OPCODE_DO 38 |
#define GEN4_OPCODE_WHILE 39 |
#define GEN4_OPCODE_BREAK 40 |
#define GEN4_OPCODE_CONTINUE 41 |
#define GEN4_OPCODE_HALT 42 |
#define GEN4_OPCODE_MSAVE 44 |
#define GEN4_OPCODE_MRESTORE 45 |
#define GEN4_OPCODE_PUSH 46 |
#define GEN4_OPCODE_POP 47 |
#define GEN4_OPCODE_WAIT 48 |
#define GEN4_OPCODE_SEND 49 |
#define GEN4_OPCODE_ADD 64 |
#define GEN4_OPCODE_MUL 65 |
#define GEN4_OPCODE_AVG 66 |
#define GEN4_OPCODE_FRC 67 |
#define GEN4_OPCODE_RNDU 68 |
#define GEN4_OPCODE_RNDD 69 |
#define GEN4_OPCODE_RNDE 70 |
#define GEN4_OPCODE_RNDZ 71 |
#define GEN4_OPCODE_MAC 72 |
#define GEN4_OPCODE_MACH 73 |
#define GEN4_OPCODE_LZD 74 |
#define GEN4_OPCODE_SAD2 80 |
#define GEN4_OPCODE_SADA2 81 |
#define GEN4_OPCODE_DP4 84 |
#define GEN4_OPCODE_DPH 85 |
#define GEN4_OPCODE_DP3 86 |
#define GEN4_OPCODE_DP2 87 |
#define GEN4_OPCODE_DPA2 88 |
#define GEN4_OPCODE_LINE 89 |
#define GEN4_OPCODE_NOP 126 |
#define GEN4_PREDICATE_NONE 0 |
#define GEN4_PREDICATE_NORMAL 1 |
#define GEN4_PREDICATE_ALIGN1_ANYV 2 |
#define GEN4_PREDICATE_ALIGN1_ALLV 3 |
#define GEN4_PREDICATE_ALIGN1_ANY2H 4 |
#define GEN4_PREDICATE_ALIGN1_ALL2H 5 |
#define GEN4_PREDICATE_ALIGN1_ANY4H 6 |
#define GEN4_PREDICATE_ALIGN1_ALL4H 7 |
#define GEN4_PREDICATE_ALIGN1_ANY8H 8 |
#define GEN4_PREDICATE_ALIGN1_ALL8H 9 |
#define GEN4_PREDICATE_ALIGN1_ANY16H 10 |
#define GEN4_PREDICATE_ALIGN1_ALL16H 11 |
#define GEN4_PREDICATE_ALIGN16_REPLICATE_X 2 |
#define GEN4_PREDICATE_ALIGN16_REPLICATE_Y 3 |
#define GEN4_PREDICATE_ALIGN16_REPLICATE_Z 4 |
#define GEN4_PREDICATE_ALIGN16_REPLICATE_W 5 |
#define GEN4_PREDICATE_ALIGN16_ANY4H 6 |
#define GEN4_PREDICATE_ALIGN16_ALL4H 7 |
#define GEN4_ARCHITECTURE_REGISTER_FILE 0 |
#define GEN4_GENERAL_REGISTER_FILE 1 |
#define GEN4_MESSAGE_REGISTER_FILE 2 |
#define GEN4_IMMEDIATE_VALUE 3 |
#define GEN4_REGISTER_TYPE_UD 0 |
#define GEN4_REGISTER_TYPE_D 1 |
#define GEN4_REGISTER_TYPE_UW 2 |
#define GEN4_REGISTER_TYPE_W 3 |
#define GEN4_REGISTER_TYPE_UB 4 |
#define GEN4_REGISTER_TYPE_B 5 |
#define GEN4_REGISTER_TYPE_VF 5 /* packed float vector, immediates only? */ |
#define GEN4_REGISTER_TYPE_HF 6 |
#define GEN4_REGISTER_TYPE_V 6 /* packed int vector, immediates only, uword dest only */ |
#define GEN4_REGISTER_TYPE_F 7 |
#define GEN4_ARF_NULL 0x00 |
#define GEN4_ARF_ADDRESS 0x10 |
#define GEN4_ARF_ACCUMULATOR 0x20 |
#define GEN4_ARF_FLAG 0x30 |
#define GEN4_ARF_MASK 0x40 |
#define GEN4_ARF_MASK_STACK 0x50 |
#define GEN4_ARF_MASK_STACK_DEPTH 0x60 |
#define GEN4_ARF_STATE 0x70 |
#define GEN4_ARF_CONTROL 0x80 |
#define GEN4_ARF_NOTIFICATION_COUNT 0x90 |
#define GEN4_ARF_IP 0xA0 |
#define GEN4_AMASK 0 |
#define GEN4_IMASK 1 |
#define GEN4_LMASK 2 |
#define GEN4_CMASK 3 |
#define GEN4_THREAD_NORMAL 0 |
#define GEN4_THREAD_ATOMIC 1 |
#define GEN4_THREAD_SWITCH 2 |
#define GEN4_VERTICAL_STRIDE_0 0 |
#define GEN4_VERTICAL_STRIDE_1 1 |
#define GEN4_VERTICAL_STRIDE_2 2 |
#define GEN4_VERTICAL_STRIDE_4 3 |
#define GEN4_VERTICAL_STRIDE_8 4 |
#define GEN4_VERTICAL_STRIDE_16 5 |
#define GEN4_VERTICAL_STRIDE_32 6 |
#define GEN4_VERTICAL_STRIDE_64 7 |
#define GEN4_VERTICAL_STRIDE_128 8 |
#define GEN4_VERTICAL_STRIDE_256 9 |
#define GEN4_VERTICAL_STRIDE_ONE_DIMENSIONAL 0xF |
#define GEN4_WIDTH_1 0 |
#define GEN4_WIDTH_2 1 |
#define GEN4_WIDTH_4 2 |
#define GEN4_WIDTH_8 3 |
#define GEN4_WIDTH_16 4 |
#define GEN4_STATELESS_BUFFER_BOUNDARY_1K 0 |
#define GEN4_STATELESS_BUFFER_BOUNDARY_2K 1 |
#define GEN4_STATELESS_BUFFER_BOUNDARY_4K 2 |
#define GEN4_STATELESS_BUFFER_BOUNDARY_8K 3 |
#define GEN4_STATELESS_BUFFER_BOUNDARY_16K 4 |
#define GEN4_STATELESS_BUFFER_BOUNDARY_32K 5 |
#define GEN4_STATELESS_BUFFER_BOUNDARY_64K 6 |
#define GEN4_STATELESS_BUFFER_BOUNDARY_128K 7 |
#define GEN4_STATELESS_BUFFER_BOUNDARY_256K 8 |
#define GEN4_STATELESS_BUFFER_BOUNDARY_512K 9 |
#define GEN4_STATELESS_BUFFER_BOUNDARY_1M 10 |
#define GEN4_STATELESS_BUFFER_BOUNDARY_2M 11 |
#define GEN4_POLYGON_FACING_FRONT 0 |
#define GEN4_POLYGON_FACING_BACK 1 |
#define GEN4_MESSAGE_TARGET_NULL 0 |
#define GEN4_MESSAGE_TARGET_MATH 1 |
#define GEN4_MESSAGE_TARGET_SAMPLER 2 |
#define GEN4_MESSAGE_TARGET_GATEWAY 3 |
#define GEN4_MESSAGE_TARGET_DATAPORT_READ 4 |
#define GEN4_MESSAGE_TARGET_DATAPORT_WRITE 5 |
#define GEN4_MESSAGE_TARGET_URB 6 |
#define GEN4_MESSAGE_TARGET_THREAD_SPAWNER 7 |
#define GEN4_SAMPLER_RETURN_FORMAT_FLOAT32 0 |
#define GEN4_SAMPLER_RETURN_FORMAT_UINT32 2 |
#define GEN4_SAMPLER_RETURN_FORMAT_SINT32 3 |
#define GEN4_SAMPLER_MESSAGE_SIMD8_SAMPLE 0 |
#define GEN4_SAMPLER_MESSAGE_SIMD16_SAMPLE 0 |
#define GEN4_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS 0 |
#define GEN4_SAMPLER_MESSAGE_SIMD8_KILLPIX 1 |
#define GEN4_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_LOD 1 |
#define GEN4_SAMPLER_MESSAGE_SIMD16_SAMPLE_LOD 1 |
#define GEN4_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_GRADIENTS 2 |
#define GEN4_SAMPLER_MESSAGE_SIMD8_SAMPLE_GRADIENTS 2 |
#define GEN4_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_COMPARE 0 |
#define GEN4_SAMPLER_MESSAGE_SIMD16_SAMPLE_COMPARE 2 |
#define GEN4_SAMPLER_MESSAGE_SIMD4X2_RESINFO 2 |
#define GEN4_SAMPLER_MESSAGE_SIMD8_RESINFO 2 |
#define GEN4_SAMPLER_MESSAGE_SIMD16_RESINFO 2 |
#define GEN4_SAMPLER_MESSAGE_SIMD4X2_LD 3 |
#define GEN4_SAMPLER_MESSAGE_SIMD8_LD 3 |
#define GEN4_SAMPLER_MESSAGE_SIMD16_LD 3 |
#define GEN4_DATAPORT_OWORD_BLOCK_1_OWORDLOW 0 |
#define GEN4_DATAPORT_OWORD_BLOCK_1_OWORDHIGH 1 |
#define GEN4_DATAPORT_OWORD_BLOCK_2_OWORDS 2 |
#define GEN4_DATAPORT_OWORD_BLOCK_4_OWORDS 3 |
#define GEN4_DATAPORT_OWORD_BLOCK_8_OWORDS 4 |
#define GEN4_DATAPORT_OWORD_DUAL_BLOCK_1OWORD 0 |
#define GEN4_DATAPORT_OWORD_DUAL_BLOCK_4OWORDS 2 |
#define GEN4_DATAPORT_DWORD_SCATTERED_BLOCK_8DWORDS 2 |
#define GEN4_DATAPORT_DWORD_SCATTERED_BLOCK_16DWORDS 3 |
#define GEN4_DATAPORT_READ_MESSAGE_OWORD_BLOCK_READ 0 |
#define GEN4_DATAPORT_READ_MESSAGE_OWORD_DUAL_BLOCK_READ 1 |
#define GEN4_DATAPORT_READ_MESSAGE_DWORD_BLOCK_READ 2 |
#define GEN4_DATAPORT_READ_MESSAGE_DWORD_SCATTERED_READ 3 |
#define GEN4_DATAPORT_READ_TARGET_DATA_CACHE 0 |
#define GEN4_DATAPORT_READ_TARGET_RENDER_CACHE 1 |
#define GEN4_DATAPORT_READ_TARGET_SAMPLER_CACHE 2 |
#define GEN4_DATAPORT_RENDER_TARGET_WRITE_SIMD16_SINGLE_SOURCE 0 |
#define GEN4_DATAPORT_RENDER_TARGET_WRITE_SIMD16_SINGLE_SOURCE_REPLICATED 1 |
#define GEN4_DATAPORT_RENDER_TARGET_WRITE_SIMD8_DUAL_SOURCE_SUBSPAN01 2 |
#define GEN4_DATAPORT_RENDER_TARGET_WRITE_SIMD8_DUAL_SOURCE_SUBSPAN23 3 |
#define GEN4_DATAPORT_RENDER_TARGET_WRITE_SIMD8_SINGLE_SOURCE_SUBSPAN01 4 |
#define GEN4_DATAPORT_WRITE_MESSAGE_OWORD_BLOCK_WRITE 0 |
#define GEN4_DATAPORT_WRITE_MESSAGE_OWORD_DUAL_BLOCK_WRITE 1 |
#define GEN4_DATAPORT_WRITE_MESSAGE_DWORD_BLOCK_WRITE 2 |
#define GEN4_DATAPORT_WRITE_MESSAGE_DWORD_SCATTERED_WRITE 3 |
#define GEN4_DATAPORT_WRITE_MESSAGE_RENDER_TARGET_WRITE 4 |
#define GEN4_DATAPORT_WRITE_MESSAGE_STREAMED_VERTEX_BUFFER_WRITE 5 |
#define GEN4_DATAPORT_WRITE_MESSAGE_FLUSH_RENDER_CACHE 7 |
#define GEN4_MATH_FUNCTION_INV 1 |
#define GEN4_MATH_FUNCTION_LOG 2 |
#define GEN4_MATH_FUNCTION_EXP 3 |
#define GEN4_MATH_FUNCTION_SQRT 4 |
#define GEN4_MATH_FUNCTION_RSQ 5 |
#define GEN4_MATH_FUNCTION_SIN 6 /* was 7 */ |
#define GEN4_MATH_FUNCTION_COS 7 /* was 8 */ |
#define GEN4_MATH_FUNCTION_SINCOS 8 /* was 6 */ |
#define GEN4_MATH_FUNCTION_TAN 9 |
#define GEN4_MATH_FUNCTION_POW 10 |
#define GEN4_MATH_FUNCTION_INT_DIV_QUOTIENT_AND_REMAINDER 11 |
#define GEN4_MATH_FUNCTION_INT_DIV_QUOTIENT 12 |
#define GEN4_MATH_FUNCTION_INT_DIV_REMAINDER 13 |
#define GEN4_MATH_INTEGER_UNSIGNED 0 |
#define GEN4_MATH_INTEGER_SIGNED 1 |
#define GEN4_MATH_PRECISION_FULL 0 |
#define GEN4_MATH_PRECISION_PARTIAL 1 |
#define GEN4_MATH_SATURATE_NONE 0 |
#define GEN4_MATH_SATURATE_SATURATE 1 |
#define GEN4_MATH_DATA_VECTOR 0 |
#define GEN4_MATH_DATA_SCALAR 1 |
#define GEN4_URB_OPCODE_WRITE 0 |
#define GEN4_URB_SWIZZLE_NONE 0 |
#define GEN4_URB_SWIZZLE_INTERLEAVE 1 |
#define GEN4_URB_SWIZZLE_TRANSPOSE 2 |
#define GEN4_SCRATCH_SPACE_SIZE_1K 0 |
#define GEN4_SCRATCH_SPACE_SIZE_2K 1 |
#define GEN4_SCRATCH_SPACE_SIZE_4K 2 |
#define GEN4_SCRATCH_SPACE_SIZE_8K 3 |
#define GEN4_SCRATCH_SPACE_SIZE_16K 4 |
#define GEN4_SCRATCH_SPACE_SIZE_32K 5 |
#define GEN4_SCRATCH_SPACE_SIZE_64K 6 |
#define GEN4_SCRATCH_SPACE_SIZE_128K 7 |
#define GEN4_SCRATCH_SPACE_SIZE_256K 8 |
#define GEN4_SCRATCH_SPACE_SIZE_512K 9 |
#define GEN4_SCRATCH_SPACE_SIZE_1M 10 |
#define GEN4_SCRATCH_SPACE_SIZE_2M 11 |
#define CMD_URB_FENCE 0x6000 |
#define CMD_CONST_BUFFER_STATE 0x6001 |
#define CMD_CONST_BUFFER 0x6002 |
#define CMD_STATE_BASE_ADDRESS 0x6101 |
#define CMD_STATE_INSN_POINTER 0x6102 |
#define CMD_PIPELINE_SELECT 0x6104 |
#define CMD_PIPELINED_STATE_POINTERS 0x7800 |
#define CMD_BINDING_TABLE_PTRS 0x7801 |
#define CMD_VERTEX_BUFFER 0x7808 |
#define CMD_VERTEX_ELEMENT 0x7809 |
#define CMD_INDEX_BUFFER 0x780a |
#define CMD_VF_STATISTICS 0x780b |
#define CMD_DRAW_RECT 0x7900 |
#define CMD_BLEND_CONSTANT_COLOR 0x7901 |
#define CMD_CHROMA_KEY 0x7904 |
#define CMD_DEPTH_BUFFER 0x7905 |
#define CMD_POLY_STIPPLE_OFFSET 0x7906 |
#define CMD_POLY_STIPPLE_PATTERN 0x7907 |
#define CMD_LINE_STIPPLE_PATTERN 0x7908 |
#define CMD_GLOBAL_DEPTH_OFFSET_CLAMP 0x7908 |
#define CMD_PIPE_CONTROL 0x7a00 |
#define CMD_3D_PRIM 0x7b00 |
#define CMD_MI_FLUSH 0x0200 |
/* Various values from the R0 vertex header: |
*/ |
#define R02_PRIM_END 0x1 |
#define R02_PRIM_START 0x2 |
/* media pipeline */ |
#define GEN4_VFE_MODE_GENERIC 0x0 |
#define GEN4_VFE_MODE_VLD_MPEG2 0x1 |
#define GEN4_VFE_MODE_IS 0x2 |
#define GEN4_VFE_MODE_AVC_MC 0x4 |
#define GEN4_VFE_MODE_AVC_IT 0x7 |
#define GEN4_VFE_MODE_VC1_IT 0xB |
#define GEN4_VFE_DEBUG_COUNTER_FREE 0 |
#define GEN4_VFE_DEBUG_COUNTER_FROZEN 1 |
#define GEN4_VFE_DEBUG_COUNTER_ONCE 2 |
#define GEN4_VFE_DEBUG_COUNTER_ALWAYS 3 |
/* VLD_STATE */ |
#define GEN4_MPEG_TOP_FIELD 1 |
#define GEN4_MPEG_BOTTOM_FIELD 2 |
#define GEN4_MPEG_FRAME 3 |
#define GEN4_MPEG_QSCALE_LINEAR 0 |
#define GEN4_MPEG_QSCALE_NONLINEAR 1 |
#define GEN4_MPEG_ZIGZAG_SCAN 0 |
#define GEN4_MPEG_ALTER_VERTICAL_SCAN 1 |
#define GEN4_MPEG_I_PICTURE 1 |
#define GEN4_MPEG_P_PICTURE 2 |
#define GEN4_MPEG_B_PICTURE 3 |
/* Command packets: |
*/ |
struct header |
{ |
unsigned int length:16; |
unsigned int opcode:16; |
}; |
union header_union |
{ |
struct header bits; |
unsigned int dword; |
}; |
struct gen4_3d_control |
{ |
struct |
{ |
unsigned int length:8; |
unsigned int notify_enable:1; |
unsigned int pad:3; |
unsigned int wc_flush_enable:1; |
unsigned int depth_stall_enable:1; |
unsigned int operation:2; |
unsigned int opcode:16; |
} header; |
struct |
{ |
unsigned int pad:2; |
unsigned int dest_addr_type:1; |
unsigned int dest_addr:29; |
} dest; |
unsigned int dword2; |
unsigned int dword3; |
}; |
struct gen4_3d_primitive |
{ |
struct |
{ |
unsigned int length:8; |
unsigned int pad:2; |
unsigned int topology:5; |
unsigned int indexed:1; |
unsigned int opcode:16; |
} header; |
unsigned int verts_per_instance; |
unsigned int start_vert_location; |
unsigned int instance_count; |
unsigned int start_instance_location; |
unsigned int base_vert_location; |
}; |
/* These seem to be passed around as function args, so it works out |
* better to keep them as #defines: |
*/ |
#define GEN4_FLUSH_READ_CACHE 0x1 |
#define GEN4_FLUSH_STATE_CACHE 0x2 |
#define GEN4_INHIBIT_FLUSH_RENDER_CACHE 0x4 |
#define GEN4_FLUSH_SNAPSHOT_COUNTERS 0x8 |
struct gen4_mi_flush |
{ |
unsigned int flags:4; |
unsigned int pad:12; |
unsigned int opcode:16; |
}; |
struct gen4_vf_statistics |
{ |
unsigned int statistics_enable:1; |
unsigned int pad:15; |
unsigned int opcode:16; |
}; |
struct gen4_binding_table_pointers |
{ |
struct header header; |
unsigned int vs; |
unsigned int gs; |
unsigned int clp; |
unsigned int sf; |
unsigned int wm; |
}; |
struct gen4_blend_constant_color |
{ |
struct header header; |
float blend_constant_color[4]; |
}; |
struct gen4_depthbuffer |
{ |
union header_union header; |
union { |
struct { |
unsigned int pitch:18; |
unsigned int format:3; |
unsigned int pad:4; |
unsigned int depth_offset_disable:1; |
unsigned int tile_walk:1; |
unsigned int tiled_surface:1; |
unsigned int pad2:1; |
unsigned int surface_type:3; |
} bits; |
unsigned int dword; |
} dword1; |
unsigned int dword2_base_addr; |
union { |
struct { |
unsigned int pad:1; |
unsigned int mipmap_layout:1; |
unsigned int lod:4; |
unsigned int width:13; |
unsigned int height:13; |
} bits; |
unsigned int dword; |
} dword3; |
union { |
struct { |
unsigned int pad:12; |
unsigned int min_array_element:9; |
unsigned int depth:11; |
} bits; |
unsigned int dword; |
} dword4; |
}; |
struct gen4_drawrect |
{ |
struct header header; |
unsigned int xmin:16; |
unsigned int ymin:16; |
unsigned int xmax:16; |
unsigned int ymax:16; |
unsigned int xorg:16; |
unsigned int yorg:16; |
}; |
struct gen4_global_depth_offset_clamp |
{ |
struct header header; |
float depth_offset_clamp; |
}; |
struct gen4_indexbuffer |
{ |
union { |
struct |
{ |
unsigned int length:8; |
unsigned int index_format:2; |
unsigned int cut_index_enable:1; |
unsigned int pad:5; |
unsigned int opcode:16; |
} bits; |
unsigned int dword; |
} header; |
unsigned int buffer_start; |
unsigned int buffer_end; |
}; |
struct gen4_line_stipple |
{ |
struct header header; |
struct |
{ |
unsigned int pattern:16; |
unsigned int pad:16; |
} bits0; |
struct |
{ |
unsigned int repeat_count:9; |
unsigned int pad:7; |
unsigned int inverse_repeat_count:16; |
} bits1; |
}; |
struct gen4_pipelined_state_pointers |
{ |
struct header header; |
struct { |
unsigned int pad:5; |
unsigned int offset:27; |
} vs; |
struct |
{ |
unsigned int enable:1; |
unsigned int pad:4; |
unsigned int offset:27; |
} gs; |
struct |
{ |
unsigned int enable:1; |
unsigned int pad:4; |
unsigned int offset:27; |
} clp; |
struct |
{ |
unsigned int pad:5; |
unsigned int offset:27; |
} sf; |
struct |
{ |
unsigned int pad:5; |
unsigned int offset:27; |
} wm; |
struct |
{ |
unsigned int pad:5; |
unsigned int offset:27; /* KW: check me! */ |
} cc; |
}; |
struct gen4_polygon_stipple_offset |
{ |
struct header header; |
struct { |
unsigned int y_offset:5; |
unsigned int pad:3; |
unsigned int x_offset:5; |
unsigned int pad0:19; |
} bits0; |
}; |
struct gen4_polygon_stipple |
{ |
struct header header; |
unsigned int stipple[32]; |
}; |
struct gen4_pipeline_select |
{ |
struct |
{ |
unsigned int pipeline_select:1; |
unsigned int pad:15; |
unsigned int opcode:16; |
} header; |
}; |
struct gen4_pipe_control |
{ |
struct |
{ |
unsigned int length:8; |
unsigned int notify_enable:1; |
unsigned int pad:2; |
unsigned int instruction_state_cache_flush_enable:1; |
unsigned int write_cache_flush_enable:1; |
unsigned int depth_stall_enable:1; |
unsigned int post_sync_operation:2; |
unsigned int opcode:16; |
} header; |
struct |
{ |
unsigned int pad:2; |
unsigned int dest_addr_type:1; |
unsigned int dest_addr:29; |
} bits1; |
unsigned int data0; |
unsigned int data1; |
}; |
struct gen4_urb_fence |
{ |
struct |
{ |
unsigned int length:8; |
unsigned int vs_realloc:1; |
unsigned int gs_realloc:1; |
unsigned int clp_realloc:1; |
unsigned int sf_realloc:1; |
unsigned int vfe_realloc:1; |
unsigned int cs_realloc:1; |
unsigned int pad:2; |
unsigned int opcode:16; |
} header; |
struct |
{ |
unsigned int vs_fence:10; |
unsigned int gs_fence:10; |
unsigned int clp_fence:10; |
unsigned int pad:2; |
} bits0; |
struct |
{ |
unsigned int sf_fence:10; |
unsigned int vf_fence:10; |
unsigned int cs_fence:10; |
unsigned int pad:2; |
} bits1; |
}; |
struct gen4_constant_buffer_state /* previously gen4_command_streamer */ |
{ |
struct header header; |
struct |
{ |
unsigned int nr_urb_entries:3; |
unsigned int pad:1; |
unsigned int urb_entry_size:5; |
unsigned int pad0:23; |
} bits0; |
}; |
struct gen4_constant_buffer |
{ |
struct |
{ |
unsigned int length:8; |
unsigned int valid:1; |
unsigned int pad:7; |
unsigned int opcode:16; |
} header; |
struct |
{ |
unsigned int buffer_length:6; |
unsigned int buffer_address:26; |
} bits0; |
}; |
struct gen4_state_base_address |
{ |
struct header header; |
struct |
{ |
unsigned int modify_enable:1; |
unsigned int pad:4; |
unsigned int general_state_address:27; |
} bits0; |
struct |
{ |
unsigned int modify_enable:1; |
unsigned int pad:4; |
unsigned int surface_state_address:27; |
} bits1; |
struct |
{ |
unsigned int modify_enable:1; |
unsigned int pad:4; |
unsigned int indirect_object_state_address:27; |
} bits2; |
struct |
{ |
unsigned int modify_enable:1; |
unsigned int pad:11; |
unsigned int general_state_upper_bound:20; |
} bits3; |
struct |
{ |
unsigned int modify_enable:1; |
unsigned int pad:11; |
unsigned int indirect_object_state_upper_bound:20; |
} bits4; |
}; |
struct gen4_state_prefetch |
{ |
struct header header; |
struct |
{ |
unsigned int prefetch_count:3; |
unsigned int pad:3; |
unsigned int prefetch_pointer:26; |
} bits0; |
}; |
struct gen4_system_instruction_pointer |
{ |
struct header header; |
struct |
{ |
unsigned int pad:4; |
unsigned int system_instruction_pointer:28; |
} bits0; |
}; |
/* State structs for the various fixed function units: |
*/ |
struct thread0 |
{ |
unsigned int pad0:1; |
unsigned int grf_reg_count:3; |
unsigned int pad1:2; |
unsigned int kernel_start_pointer:26; |
}; |
struct thread1 |
{ |
unsigned int ext_halt_exception_enable:1; |
unsigned int sw_exception_enable:1; |
unsigned int mask_stack_exception_enable:1; |
unsigned int timeout_exception_enable:1; |
unsigned int illegal_op_exception_enable:1; |
unsigned int pad0:3; |
unsigned int depth_coef_urb_read_offset:6; /* WM only */ |
unsigned int pad1:2; |
unsigned int floating_point_mode:1; |
unsigned int thread_priority:1; |
unsigned int binding_table_entry_count:8; |
unsigned int pad3:5; |
unsigned int single_program_flow:1; |
}; |
struct thread2 |
{ |
unsigned int per_thread_scratch_space:4; |
unsigned int pad0:6; |
unsigned int scratch_space_base_pointer:22; |
}; |
struct thread3 |
{ |
unsigned int dispatch_grf_start_reg:4; |
unsigned int urb_entry_read_offset:6; |
unsigned int pad0:1; |
unsigned int urb_entry_read_length:6; |
unsigned int pad1:1; |
unsigned int const_urb_entry_read_offset:6; |
unsigned int pad2:1; |
unsigned int const_urb_entry_read_length:6; |
unsigned int pad3:1; |
}; |
struct gen4_clip_unit_state |
{ |
struct thread0 thread0; |
struct thread1 thread1; |
struct thread2 thread2; |
struct thread3 thread3; |
struct |
{ |
unsigned int pad0:9; |
unsigned int gs_output_stats:1; /* not always */ |
unsigned int stats_enable:1; |
unsigned int nr_urb_entries:7; |
unsigned int pad1:1; |
unsigned int urb_entry_allocation_size:5; |
unsigned int pad2:1; |
unsigned int max_threads:6; /* may be less */ |
unsigned int pad3:1; |
} thread4; |
struct |
{ |
unsigned int pad0:13; |
unsigned int clip_mode:3; |
unsigned int userclip_enable_flags:8; |
unsigned int userclip_must_clip:1; |
unsigned int pad1:1; |
unsigned int guard_band_enable:1; |
unsigned int viewport_z_clip_enable:1; |
unsigned int viewport_xy_clip_enable:1; |
unsigned int vertex_position_space:1; |
unsigned int api_mode:1; |
unsigned int pad2:1; |
} clip5; |
struct |
{ |
unsigned int pad0:5; |
unsigned int clipper_viewport_state_ptr:27; |
} clip6; |
float viewport_xmin; |
float viewport_xmax; |
float viewport_ymin; |
float viewport_ymax; |
}; |
struct gen4_cc_unit_state |
{ |
struct |
{ |
unsigned int pad0:3; |
unsigned int bf_stencil_pass_depth_pass_op:3; |
unsigned int bf_stencil_pass_depth_fail_op:3; |
unsigned int bf_stencil_fail_op:3; |
unsigned int bf_stencil_func:3; |
unsigned int bf_stencil_enable:1; |
unsigned int pad1:2; |
unsigned int stencil_write_enable:1; |
unsigned int stencil_pass_depth_pass_op:3; |
unsigned int stencil_pass_depth_fail_op:3; |
unsigned int stencil_fail_op:3; |
unsigned int stencil_func:3; |
unsigned int stencil_enable:1; |
} cc0; |
struct |
{ |
unsigned int bf_stencil_ref:8; |
unsigned int stencil_write_mask:8; |
unsigned int stencil_test_mask:8; |
unsigned int stencil_ref:8; |
} cc1; |
struct |
{ |
unsigned int logicop_enable:1; |
unsigned int pad0:10; |
unsigned int depth_write_enable:1; |
unsigned int depth_test_function:3; |
unsigned int depth_test:1; |
unsigned int bf_stencil_write_mask:8; |
unsigned int bf_stencil_test_mask:8; |
} cc2; |
struct |
{ |
unsigned int pad0:8; |
unsigned int alpha_test_func:3; |
unsigned int alpha_test:1; |
unsigned int blend_enable:1; |
unsigned int ia_blend_enable:1; |
unsigned int pad1:1; |
unsigned int alpha_test_format:1; |
unsigned int pad2:16; |
} cc3; |
struct |
{ |
unsigned int pad0:5; |
unsigned int cc_viewport_state_offset:27; |
} cc4; |
struct |
{ |
unsigned int pad0:2; |
unsigned int ia_dest_blend_factor:5; |
unsigned int ia_src_blend_factor:5; |
unsigned int ia_blend_function:3; |
unsigned int statistics_enable:1; |
unsigned int logicop_func:4; |
unsigned int pad1:11; |
unsigned int dither_enable:1; |
} cc5; |
struct |
{ |
unsigned int clamp_post_alpha_blend:1; |
unsigned int clamp_pre_alpha_blend:1; |
unsigned int clamp_range:2; |
unsigned int pad0:11; |
unsigned int y_dither_offset:2; |
unsigned int x_dither_offset:2; |
unsigned int dest_blend_factor:5; |
unsigned int src_blend_factor:5; |
unsigned int blend_function:3; |
} cc6; |
struct { |
union { |
float f; |
unsigned char ub[4]; |
} alpha_ref; |
} cc7; |
}; |
struct gen4_sf_unit_state |
{ |
struct thread0 thread0; |
struct { |
unsigned int pad0:7; |
unsigned int sw_exception_enable:1; |
unsigned int pad1:3; |
unsigned int mask_stack_exception_enable:1; |
unsigned int pad2:1; |
unsigned int illegal_op_exception_enable:1; |
unsigned int pad3:2; |
unsigned int floating_point_mode:1; |
unsigned int thread_priority:1; |
unsigned int binding_table_entry_count:8; |
unsigned int pad4:5; |
unsigned int single_program_flow:1; |
} sf1; |
struct thread2 thread2; |
struct thread3 thread3; |
struct |
{ |
unsigned int pad0:10; |
unsigned int stats_enable:1; |
unsigned int nr_urb_entries:7; |
unsigned int pad1:1; |
unsigned int urb_entry_allocation_size:5; |
unsigned int pad2:1; |
unsigned int max_threads:6; |
unsigned int pad3:1; |
} thread4; |
struct |
{ |
unsigned int front_winding:1; |
unsigned int viewport_transform:1; |
unsigned int pad0:3; |
unsigned int sf_viewport_state_offset:27; |
} sf5; |
struct |
{ |
unsigned int pad0:9; |
unsigned int dest_org_vbias:4; |
unsigned int dest_org_hbias:4; |
unsigned int scissor:1; |
unsigned int disable_2x2_trifilter:1; |
unsigned int disable_zero_pix_trifilter:1; |
unsigned int point_rast_rule:2; |
unsigned int line_endcap_aa_region_width:2; |
unsigned int line_width:4; |
unsigned int fast_scissor_disable:1; |
unsigned int cull_mode:2; |
unsigned int aa_enable:1; |
} sf6; |
struct |
{ |
unsigned int point_size:11; |
unsigned int use_point_size_state:1; |
unsigned int subpixel_precision:1; |
unsigned int sprite_point:1; |
unsigned int pad0:11; |
unsigned int trifan_pv:2; |
unsigned int linestrip_pv:2; |
unsigned int tristrip_pv:2; |
unsigned int line_last_pixel_enable:1; |
} sf7; |
}; |
struct gen4_gs_unit_state |
{ |
struct thread0 thread0; |
struct thread1 thread1; |
struct thread2 thread2; |
struct thread3 thread3; |
struct |
{ |
unsigned int pad0:10; |
unsigned int stats_enable:1; |
unsigned int nr_urb_entries:7; |
unsigned int pad1:1; |
unsigned int urb_entry_allocation_size:5; |
unsigned int pad2:1; |
unsigned int max_threads:1; |
unsigned int pad3:6; |
} thread4; |
struct |
{ |
unsigned int sampler_count:3; |
unsigned int pad0:2; |
unsigned int sampler_state_pointer:27; |
} gs5; |
struct |
{ |
unsigned int max_vp_index:4; |
unsigned int pad0:26; |
unsigned int reorder_enable:1; |
unsigned int pad1:1; |
} gs6; |
}; |
struct gen4_vs_unit_state |
{ |
struct thread0 thread0; |
struct thread1 thread1; |
struct thread2 thread2; |
struct thread3 thread3; |
struct |
{ |
unsigned int pad0:10; |
unsigned int stats_enable:1; |
unsigned int nr_urb_entries:7; |
unsigned int pad1:1; |
unsigned int urb_entry_allocation_size:5; |
unsigned int pad2:1; |
unsigned int max_threads:4; |
unsigned int pad3:3; |
} thread4; |
struct |
{ |
unsigned int sampler_count:3; |
unsigned int pad0:2; |
unsigned int sampler_state_pointer:27; |
} vs5; |
struct |
{ |
unsigned int vs_enable:1; |
unsigned int vert_cache_disable:1; |
unsigned int pad0:30; |
} vs6; |
}; |
struct gen4_wm_unit_state |
{ |
struct thread0 thread0; |
struct thread1 thread1; |
struct thread2 thread2; |
struct thread3 thread3; |
struct { |
unsigned int stats_enable:1; |
unsigned int pad0:1; |
unsigned int sampler_count:3; |
unsigned int sampler_state_pointer:27; |
} wm4; |
struct |
{ |
unsigned int enable_8_pix:1; |
unsigned int enable_16_pix:1; |
unsigned int enable_32_pix:1; |
unsigned int pad0:7; |
unsigned int legacy_global_depth_bias:1; |
unsigned int line_stipple:1; |
unsigned int depth_offset:1; |
unsigned int polygon_stipple:1; |
unsigned int line_aa_region_width:2; |
unsigned int line_endcap_aa_region_width:2; |
unsigned int early_depth_test:1; |
unsigned int thread_dispatch_enable:1; |
unsigned int program_uses_depth:1; |
unsigned int program_computes_depth:1; |
unsigned int program_uses_killpixel:1; |
unsigned int legacy_line_rast: 1; |
unsigned int transposed_urb_read:1; |
unsigned int max_threads:7; |
} wm5; |
float global_depth_offset_constant; |
float global_depth_offset_scale; |
struct { |
unsigned int pad0:1; |
unsigned int grf_reg_count_1:3; |
unsigned int pad1:2; |
unsigned int kernel_start_pointer_1:26; |
} wm8; |
struct { |
unsigned int pad0:1; |
unsigned int grf_reg_count_2:3; |
unsigned int pad1:2; |
unsigned int kernel_start_pointer_2:26; |
} wm9; |
struct { |
unsigned int pad0:1; |
unsigned int grf_reg_count_3:3; |
unsigned int pad1:2; |
unsigned int kernel_start_pointer_3:26; |
} wm10; |
}; |
struct gen4_wm_unit_state_padded { |
struct gen4_wm_unit_state state; |
char pad[64 - sizeof(struct gen4_wm_unit_state)]; |
}; |
/* The hardware supports two different modes for border color. The |
* default (OpenGL) mode uses floating-point color channels, while the |
* legacy mode uses 4 bytes. |
* |
* More significantly, the legacy mode respects the components of the |
* border color for channels not present in the source, (whereas the |
* default mode will ignore the border color's alpha channel and use |
* alpha==1 for an RGB source, for example). |
* |
* The legacy mode matches the semantics specified by the Render |
* extension. |
*/ |
struct gen4_sampler_default_border_color { |
float color[4]; |
}; |
struct gen4_sampler_legacy_border_color { |
uint8_t color[4]; |
}; |
struct gen4_sampler_state |
{ |
struct |
{ |
unsigned int shadow_function:3; |
unsigned int lod_bias:11; |
unsigned int min_filter:3; |
unsigned int mag_filter:3; |
unsigned int mip_filter:2; |
unsigned int base_level:5; |
unsigned int pad:1; |
unsigned int lod_preclamp:1; |
unsigned int border_color_mode:1; |
unsigned int pad0:1; |
unsigned int disable:1; |
} ss0; |
struct |
{ |
unsigned int r_wrap_mode:3; |
unsigned int t_wrap_mode:3; |
unsigned int s_wrap_mode:3; |
unsigned int pad:3; |
unsigned int max_lod:10; |
unsigned int min_lod:10; |
} ss1; |
struct |
{ |
unsigned int pad:5; |
unsigned int border_color_pointer:27; |
} ss2; |
struct |
{ |
unsigned int pad:19; |
unsigned int max_aniso:3; |
unsigned int chroma_key_mode:1; |
unsigned int chroma_key_index:2; |
unsigned int chroma_key_enable:1; |
unsigned int monochrome_filter_width:3; |
unsigned int monochrome_filter_height:3; |
} ss3; |
}; |
struct gen4_clipper_viewport |
{ |
float xmin; |
float xmax; |
float ymin; |
float ymax; |
}; |
struct gen4_cc_viewport |
{ |
float min_depth; |
float max_depth; |
}; |
struct gen4_sf_viewport |
{ |
struct { |
float m00; |
float m11; |
float m22; |
float m30; |
float m31; |
float m32; |
} viewport; |
struct { |
short xmin; |
short ymin; |
short xmax; |
short ymax; |
} scissor; |
}; |
/* Documented in the subsystem/shared-functions/sampler chapter... |
*/ |
struct gen4_surface_state |
{ |
struct { |
unsigned int cube_pos_z:1; |
unsigned int cube_neg_z:1; |
unsigned int cube_pos_y:1; |
unsigned int cube_neg_y:1; |
unsigned int cube_pos_x:1; |
unsigned int cube_neg_x:1; |
unsigned int pad:3; |
unsigned int render_cache_read_mode:1; |
unsigned int mipmap_layout_mode:1; |
unsigned int vert_line_stride_ofs:1; |
unsigned int vert_line_stride:1; |
unsigned int color_blend:1; |
unsigned int writedisable_blue:1; |
unsigned int writedisable_green:1; |
unsigned int writedisable_red:1; |
unsigned int writedisable_alpha:1; |
unsigned int surface_format:9; |
unsigned int data_return_format:1; |
unsigned int pad0:1; |
unsigned int surface_type:3; |
} ss0; |
struct { |
unsigned int base_addr; |
} ss1; |
struct { |
unsigned int render_target_rotation:2; |
unsigned int mip_count:4; |
unsigned int width:13; |
unsigned int height:13; |
} ss2; |
struct { |
unsigned int tile_walk:1; |
unsigned int tiled_surface:1; |
unsigned int pad:1; |
unsigned int pitch:18; |
unsigned int depth:11; |
} ss3; |
struct { |
unsigned int pad:19; |
unsigned int min_array_elt:9; |
unsigned int min_lod:4; |
} ss4; |
struct { |
unsigned int pad:20; |
unsigned int y_offset:4; |
unsigned int pad2:1; |
unsigned int x_offset:7; |
} ss5; |
}; |
/* Surface state DW0 */ |
#define GEN4_SURFACE_RC_READ_WRITE (1 << 8) |
#define GEN4_SURFACE_MIPLAYOUT_SHIFT 10 |
#define GEN4_SURFACE_MIPMAPLAYOUT_BELOW 0 |
#define GEN4_SURFACE_MIPMAPLAYOUT_RIGHT 1 |
#define GEN4_SURFACE_CUBEFACE_ENABLES 0x3f |
#define GEN4_SURFACE_BLEND_ENABLED (1 << 13) |
#define GEN4_SURFACE_WRITEDISABLE_B_SHIFT 14 |
#define GEN4_SURFACE_WRITEDISABLE_G_SHIFT 15 |
#define GEN4_SURFACE_WRITEDISABLE_R_SHIFT 16 |
#define GEN4_SURFACE_WRITEDISABLE_A_SHIFT 17 |
#define GEN4_SURFACE_FORMAT_SHIFT 18 |
#define GEN4_SURFACE_FORMAT_MASK _MASK(26, 18) |
#define GEN4_SURFACE_TYPE_SHIFT 29 |
#define GEN4_SURFACE_TYPE_MASK _MASK(31, 29) |
#define GEN4_SURFACE_1D 0 |
#define GEN4_SURFACE_2D 1 |
#define GEN4_SURFACE_3D 2 |
#define GEN4_SURFACE_CUBE 3 |
#define GEN4_SURFACE_BUFFER 4 |
#define GEN4_SURFACE_NULL 7 |
/* Surface state DW2 */ |
#define GEN4_SURFACE_HEIGHT_SHIFT 19 |
#define GEN4_SURFACE_HEIGHT_MASK _MASK(31, 19) |
#define GEN4_SURFACE_WIDTH_SHIFT 6 |
#define GEN4_SURFACE_WIDTH_MASK _MASK(18, 6) |
#define GEN4_SURFACE_LOD_SHIFT 2 |
#define GEN4_SURFACE_LOD_MASK _MASK(5, 2) |
/* Surface state DW3 */ |
#define GEN4_SURFACE_DEPTH_SHIFT 21 |
#define GEN4_SURFACE_DEPTH_MASK _MASK(31, 21) |
#define GEN4_SURFACE_PITCH_SHIFT 3 |
#define GEN4_SURFACE_PITCH_MASK _MASK(19, 3) |
#define GEN4_SURFACE_TILED (1 << 1) |
#define GEN4_SURFACE_TILED_Y (1 << 0) |
/* Surface state DW4 */ |
#define GEN4_SURFACE_MIN_LOD_SHIFT 28 |
#define GEN4_SURFACE_MIN_LOD_MASK _MASK(31, 28) |
/* Surface state DW5 */ |
#define GEN4_SURFACE_X_OFFSET_SHIFT 25 |
#define GEN4_SURFACE_X_OFFSET_MASK _MASK(31, 25) |
#define GEN4_SURFACE_Y_OFFSET_SHIFT 20 |
#define GEN4_SURFACE_Y_OFFSET_MASK _MASK(23, 20) |
struct gen4_vertex_buffer_state |
{ |
struct { |
unsigned int pitch:11; |
unsigned int pad:15; |
unsigned int access_type:1; |
unsigned int vb_index:5; |
} vb0; |
unsigned int start_addr; |
unsigned int max_index; |
#if 1 |
unsigned int instance_data_step_rate; /* not included for sequential/random vertices? */ |
#endif |
}; |
#define GEN4_VBP_MAX 17 |
struct gen4_vb_array_state { |
struct header header; |
struct gen4_vertex_buffer_state vb[GEN4_VBP_MAX]; |
}; |
struct gen4_vertex_element_state |
{ |
struct |
{ |
unsigned int src_offset:11; |
unsigned int pad:5; |
unsigned int src_format:9; |
unsigned int pad0:1; |
unsigned int valid:1; |
unsigned int vertex_buffer_index:5; |
} ve0; |
struct |
{ |
unsigned int dst_offset:8; |
unsigned int pad:8; |
unsigned int vfcomponent3:4; |
unsigned int vfcomponent2:4; |
unsigned int vfcomponent1:4; |
unsigned int vfcomponent0:4; |
} ve1; |
}; |
#define GEN4_VEP_MAX 18 |
struct gen4_vertex_element_packet { |
struct header header; |
struct gen4_vertex_element_state ve[GEN4_VEP_MAX]; /* note: less than _TNL_ATTRIB_MAX */ |
}; |
struct gen4_urb_immediate { |
unsigned int opcode:4; |
unsigned int offset:6; |
unsigned int swizzle_control:2; |
unsigned int pad:1; |
unsigned int allocate:1; |
unsigned int used:1; |
unsigned int complete:1; |
unsigned int response_length:4; |
unsigned int msg_length:4; |
unsigned int msg_target:4; |
unsigned int pad1:3; |
unsigned int end_of_thread:1; |
}; |
/* Instruction format for the execution units: |
*/ |
struct gen4_instruction |
{ |
struct |
{ |
unsigned int opcode:7; |
unsigned int pad:1; |
unsigned int access_mode:1; |
unsigned int mask_control:1; |
unsigned int dependency_control:2; |
unsigned int compression_control:2; |
unsigned int thread_control:2; |
unsigned int predicate_control:4; |
unsigned int predicate_inverse:1; |
unsigned int execution_size:3; |
unsigned int destreg__conditonalmod:4; /* destreg - send, conditionalmod - others */ |
unsigned int pad0:2; |
unsigned int debug_control:1; |
unsigned int saturate:1; |
} header; |
union { |
struct |
{ |
unsigned int dest_reg_file:2; |
unsigned int dest_reg_type:3; |
unsigned int src0_reg_file:2; |
unsigned int src0_reg_type:3; |
unsigned int src1_reg_file:2; |
unsigned int src1_reg_type:3; |
unsigned int pad:1; |
unsigned int dest_subreg_nr:5; |
unsigned int dest_reg_nr:8; |
unsigned int dest_horiz_stride:2; |
unsigned int dest_address_mode:1; |
} da1; |
struct |
{ |
unsigned int dest_reg_file:2; |
unsigned int dest_reg_type:3; |
unsigned int src0_reg_file:2; |
unsigned int src0_reg_type:3; |
unsigned int pad:6; |
int dest_indirect_offset:10; /* offset against the deref'd address reg */ |
unsigned int dest_subreg_nr:3; /* subnr for the address reg a0.x */ |
unsigned int dest_horiz_stride:2; |
unsigned int dest_address_mode:1; |
} ia1; |
struct |
{ |
unsigned int dest_reg_file:2; |
unsigned int dest_reg_type:3; |
unsigned int src0_reg_file:2; |
unsigned int src0_reg_type:3; |
unsigned int src1_reg_file:2; |
unsigned int src1_reg_type:3; |
unsigned int pad0:1; |
unsigned int dest_writemask:4; |
unsigned int dest_subreg_nr:1; |
unsigned int dest_reg_nr:8; |
unsigned int pad1:2; |
unsigned int dest_address_mode:1; |
} da16; |
struct |
{ |
unsigned int dest_reg_file:2; |
unsigned int dest_reg_type:3; |
unsigned int src0_reg_file:2; |
unsigned int src0_reg_type:3; |
unsigned int pad0:6; |
unsigned int dest_writemask:4; |
int dest_indirect_offset:6; |
unsigned int dest_subreg_nr:3; |
unsigned int pad1:2; |
unsigned int dest_address_mode:1; |
} ia16; |
} bits1; |
union { |
struct |
{ |
unsigned int src0_subreg_nr:5; |
unsigned int src0_reg_nr:8; |
unsigned int src0_abs:1; |
unsigned int src0_negate:1; |
unsigned int src0_address_mode:1; |
unsigned int src0_horiz_stride:2; |
unsigned int src0_width:3; |
unsigned int src0_vert_stride:4; |
unsigned int flag_reg_nr:1; |
unsigned int pad:6; |
} da1; |
struct |
{ |
int src0_indirect_offset:10; |
unsigned int src0_subreg_nr:3; |
unsigned int src0_abs:1; |
unsigned int src0_negate:1; |
unsigned int src0_address_mode:1; |
unsigned int src0_horiz_stride:2; |
unsigned int src0_width:3; |
unsigned int src0_vert_stride:4; |
unsigned int flag_reg_nr:1; |
unsigned int pad:6; |
} ia1; |
struct |
{ |
unsigned int src0_swz_x:2; |
unsigned int src0_swz_y:2; |
unsigned int src0_subreg_nr:1; |
unsigned int src0_reg_nr:8; |
unsigned int src0_abs:1; |
unsigned int src0_negate:1; |
unsigned int src0_address_mode:1; |
unsigned int src0_swz_z:2; |
unsigned int src0_swz_w:2; |
unsigned int pad0:1; |
unsigned int src0_vert_stride:4; |
unsigned int flag_reg_nr:1; |
unsigned int pad1:6; |
} da16; |
struct |
{ |
unsigned int src0_swz_x:2; |
unsigned int src0_swz_y:2; |
int src0_indirect_offset:6; |
unsigned int src0_subreg_nr:3; |
unsigned int src0_abs:1; |
unsigned int src0_negate:1; |
unsigned int src0_address_mode:1; |
unsigned int src0_swz_z:2; |
unsigned int src0_swz_w:2; |
unsigned int pad0:1; |
unsigned int src0_vert_stride:4; |
unsigned int flag_reg_nr:1; |
unsigned int pad1:6; |
} ia16; |
} bits2; |
union |
{ |
struct |
{ |
unsigned int src1_subreg_nr:5; |
unsigned int src1_reg_nr:8; |
unsigned int src1_abs:1; |
unsigned int src1_negate:1; |
unsigned int pad:1; |
unsigned int src1_horiz_stride:2; |
unsigned int src1_width:3; |
unsigned int src1_vert_stride:4; |
unsigned int pad0:7; |
} da1; |
struct |
{ |
unsigned int src1_swz_x:2; |
unsigned int src1_swz_y:2; |
unsigned int src1_subreg_nr:1; |
unsigned int src1_reg_nr:8; |
unsigned int src1_abs:1; |
unsigned int src1_negate:1; |
unsigned int pad0:1; |
unsigned int src1_swz_z:2; |
unsigned int src1_swz_w:2; |
unsigned int pad1:1; |
unsigned int src1_vert_stride:4; |
unsigned int pad2:7; |
} da16; |
struct |
{ |
int src1_indirect_offset:10; |
unsigned int src1_subreg_nr:3; |
unsigned int src1_abs:1; |
unsigned int src1_negate:1; |
unsigned int pad0:1; |
unsigned int src1_horiz_stride:2; |
unsigned int src1_width:3; |
unsigned int src1_vert_stride:4; |
unsigned int flag_reg_nr:1; |
unsigned int pad1:6; |
} ia1; |
struct |
{ |
unsigned int src1_swz_x:2; |
unsigned int src1_swz_y:2; |
int src1_indirect_offset:6; |
unsigned int src1_subreg_nr:3; |
unsigned int src1_abs:1; |
unsigned int src1_negate:1; |
unsigned int pad0:1; |
unsigned int src1_swz_z:2; |
unsigned int src1_swz_w:2; |
unsigned int pad1:1; |
unsigned int src1_vert_stride:4; |
unsigned int flag_reg_nr:1; |
unsigned int pad2:6; |
} ia16; |
struct |
{ |
int jump_count:16; /* note: signed */ |
unsigned int pop_count:4; |
unsigned int pad0:12; |
} if_else; |
struct { |
unsigned int function:4; |
unsigned int int_type:1; |
unsigned int precision:1; |
unsigned int saturate:1; |
unsigned int data_type:1; |
unsigned int pad0:8; |
unsigned int response_length:4; |
unsigned int msg_length:4; |
unsigned int msg_target:4; |
unsigned int pad1:3; |
unsigned int end_of_thread:1; |
} math; |
struct { |
unsigned int binding_table_index:8; |
unsigned int sampler:4; |
unsigned int return_format:2; |
unsigned int msg_type:2; |
unsigned int response_length:4; |
unsigned int msg_length:4; |
unsigned int msg_target:4; |
unsigned int pad1:3; |
unsigned int end_of_thread:1; |
} sampler; |
struct gen4_urb_immediate urb; |
struct { |
unsigned int binding_table_index:8; |
unsigned int msg_control:4; |
unsigned int msg_type:2; |
unsigned int target_cache:2; |
unsigned int response_length:4; |
unsigned int msg_length:4; |
unsigned int msg_target:4; |
unsigned int pad1:3; |
unsigned int end_of_thread:1; |
} dp_read; |
struct { |
unsigned int binding_table_index:8; |
unsigned int msg_control:3; |
unsigned int pixel_scoreboard_clear:1; |
unsigned int msg_type:3; |
unsigned int send_commit_msg:1; |
unsigned int response_length:4; |
unsigned int msg_length:4; |
unsigned int msg_target:4; |
unsigned int pad1:3; |
unsigned int end_of_thread:1; |
} dp_write; |
struct { |
unsigned int pad:16; |
unsigned int response_length:4; |
unsigned int msg_length:4; |
unsigned int msg_target:4; |
unsigned int pad1:3; |
unsigned int end_of_thread:1; |
} generic; |
unsigned int ud; |
} bits3; |
}; |
/* media pipeline */ |
struct gen4_vfe_state { |
struct { |
unsigned int per_thread_scratch_space:4; |
unsigned int pad3:3; |
unsigned int extend_vfe_state_present:1; |
unsigned int pad2:2; |
unsigned int scratch_base:22; |
} vfe0; |
struct { |
unsigned int debug_counter_control:2; |
unsigned int children_present:1; |
unsigned int vfe_mode:4; |
unsigned int pad2:2; |
unsigned int num_urb_entries:7; |
unsigned int urb_entry_alloc_size:9; |
unsigned int max_threads:7; |
} vfe1; |
struct { |
unsigned int pad4:4; |
unsigned int interface_descriptor_base:28; |
} vfe2; |
}; |
struct gen4_vld_state { |
struct { |
unsigned int pad6:6; |
unsigned int scan_order:1; |
unsigned int intra_vlc_format:1; |
unsigned int quantizer_scale_type:1; |
unsigned int concealment_motion_vector:1; |
unsigned int frame_predict_frame_dct:1; |
unsigned int top_field_first:1; |
unsigned int picture_structure:2; |
unsigned int intra_dc_precision:2; |
unsigned int f_code_0_0:4; |
unsigned int f_code_0_1:4; |
unsigned int f_code_1_0:4; |
unsigned int f_code_1_1:4; |
} vld0; |
struct { |
unsigned int pad2:9; |
unsigned int picture_coding_type:2; |
unsigned int pad:21; |
} vld1; |
struct { |
unsigned int index_0:4; |
unsigned int index_1:4; |
unsigned int index_2:4; |
unsigned int index_3:4; |
unsigned int index_4:4; |
unsigned int index_5:4; |
unsigned int index_6:4; |
unsigned int index_7:4; |
} desc_remap_table0; |
struct { |
unsigned int index_8:4; |
unsigned int index_9:4; |
unsigned int index_10:4; |
unsigned int index_11:4; |
unsigned int index_12:4; |
unsigned int index_13:4; |
unsigned int index_14:4; |
unsigned int index_15:4; |
} desc_remap_table1; |
}; |
struct gen4_interface_descriptor { |
struct { |
unsigned int grf_reg_blocks:4; |
unsigned int pad:2; |
unsigned int kernel_start_pointer:26; |
} desc0; |
struct { |
unsigned int pad:7; |
unsigned int software_exception:1; |
unsigned int pad2:3; |
unsigned int maskstack_exception:1; |
unsigned int pad3:1; |
unsigned int illegal_opcode_exception:1; |
unsigned int pad4:2; |
unsigned int floating_point_mode:1; |
unsigned int thread_priority:1; |
unsigned int single_program_flow:1; |
unsigned int pad5:1; |
unsigned int const_urb_entry_read_offset:6; |
unsigned int const_urb_entry_read_len:6; |
} desc1; |
struct { |
unsigned int pad:2; |
unsigned int sampler_count:3; |
unsigned int sampler_state_pointer:27; |
} desc2; |
struct { |
unsigned int binding_table_entry_count:5; |
unsigned int binding_table_pointer:27; |
} desc3; |
}; |
struct gen6_blend_state |
{ |
struct { |
unsigned int dest_blend_factor:5; |
unsigned int source_blend_factor:5; |
unsigned int pad3:1; |
unsigned int blend_func:3; |
unsigned int pad2:1; |
unsigned int ia_dest_blend_factor:5; |
unsigned int ia_source_blend_factor:5; |
unsigned int pad1:1; |
unsigned int ia_blend_func:3; |
unsigned int pad0:1; |
unsigned int ia_blend_enable:1; |
unsigned int blend_enable:1; |
} blend0; |
struct { |
unsigned int post_blend_clamp_enable:1; |
unsigned int pre_blend_clamp_enable:1; |
unsigned int clamp_range:2; |
unsigned int pad0:4; |
unsigned int x_dither_offset:2; |
unsigned int y_dither_offset:2; |
unsigned int dither_enable:1; |
unsigned int alpha_test_func:3; |
unsigned int alpha_test_enable:1; |
unsigned int pad1:1; |
unsigned int logic_op_func:4; |
unsigned int logic_op_enable:1; |
unsigned int pad2:1; |
unsigned int write_disable_b:1; |
unsigned int write_disable_g:1; |
unsigned int write_disable_r:1; |
unsigned int write_disable_a:1; |
unsigned int pad3:1; |
unsigned int alpha_to_coverage_dither:1; |
unsigned int alpha_to_one:1; |
unsigned int alpha_to_coverage:1; |
} blend1; |
}; |
struct gen6_color_calc_state |
{ |
struct { |
unsigned int alpha_test_format:1; |
unsigned int pad0:14; |
unsigned int round_disable:1; |
unsigned int bf_stencil_ref:8; |
unsigned int stencil_ref:8; |
} cc0; |
union { |
float alpha_ref_f; |
struct { |
unsigned int ui:8; |
unsigned int pad0:24; |
} alpha_ref_fi; |
} cc1; |
float constant_r; |
float constant_g; |
float constant_b; |
float constant_a; |
}; |
struct gen6_depth_stencil_state |
{ |
struct { |
unsigned int pad0:3; |
unsigned int bf_stencil_pass_depth_pass_op:3; |
unsigned int bf_stencil_pass_depth_fail_op:3; |
unsigned int bf_stencil_fail_op:3; |
unsigned int bf_stencil_func:3; |
unsigned int bf_stencil_enable:1; |
unsigned int pad1:2; |
unsigned int stencil_write_enable:1; |
unsigned int stencil_pass_depth_pass_op:3; |
unsigned int stencil_pass_depth_fail_op:3; |
unsigned int stencil_fail_op:3; |
unsigned int stencil_func:3; |
unsigned int stencil_enable:1; |
} ds0; |
struct { |
unsigned int bf_stencil_write_mask:8; |
unsigned int bf_stencil_test_mask:8; |
unsigned int stencil_write_mask:8; |
unsigned int stencil_test_mask:8; |
} ds1; |
struct { |
unsigned int pad0:26; |
unsigned int depth_write_enable:1; |
unsigned int depth_test_func:3; |
unsigned int pad1:1; |
unsigned int depth_test_enable:1; |
} ds2; |
}; |
typedef enum { |
SAMPLER_FILTER_NEAREST = 0, |
SAMPLER_FILTER_BILINEAR, |
FILTER_COUNT |
} sampler_filter_t; |
typedef enum { |
SAMPLER_EXTEND_NONE = 0, |
SAMPLER_EXTEND_REPEAT, |
SAMPLER_EXTEND_PAD, |
SAMPLER_EXTEND_REFLECT, |
EXTEND_COUNT |
} sampler_extend_t; |
typedef enum { |
WM_KERNEL = 0, |
WM_KERNEL_P, |
WM_KERNEL_MASK, |
WM_KERNEL_MASK_P, |
WM_KERNEL_MASKCA, |
WM_KERNEL_MASKCA_P, |
WM_KERNEL_MASKSA, |
WM_KERNEL_MASKSA_P, |
WM_KERNEL_OPACITY, |
WM_KERNEL_OPACITY_P, |
WM_KERNEL_VIDEO_PLANAR, |
WM_KERNEL_VIDEO_PACKED, |
KERNEL_COUNT |
} wm_kernel_t; |
#endif |
/drivers/video/Intel-2D/sna/gen4_source.h |
---|
0,0 → 1,22 |
#ifndef GEN4_SOURCE_H |
#define GEN4_SOURCE_H |
#include "compiler.h" |
#include "sna.h" |
#include "sna_render.h" |
bool |
gen4_channel_init_solid(struct sna *sna, |
struct sna_composite_channel *channel, |
uint32_t color); |
bool |
gen4_channel_init_linear(struct sna *sna, |
PicturePtr picture, |
struct sna_composite_channel *channel, |
int x, int y, |
int w, int h, |
int dst_x, int dst_y); |
#endif /* GEN4_SOURCE_H */ |
/drivers/video/Intel-2D/sna/gen4_vertex.c |
---|
0,0 → 1,1868 |
/* |
* Copyright © 2012 Intel Corporation |
* |
* Permission is hereby granted, free of charge, to any person obtaining a |
* copy of this software and associated documentation files (the "Software"), |
* to deal in the Software without restriction, including without limitation |
* the rights to use, copy, modify, merge, publish, distribute, sublicense, |
* and/or sell copies of the Software, and to permit persons to whom the |
* Software is furnished to do so, subject to the following conditions: |
* |
* The above copyright notice and this permission notice (including the next |
* paragraph) shall be included in all copies or substantial portions of the |
* Software. |
* |
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL |
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, |
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE |
* SOFTWARE. |
* |
* Authors: |
* Chris Wilson <chris@chris-wilson.co.uk> |
* |
*/ |
#ifdef HAVE_CONFIG_H |
#include "config.h" |
#endif |
#include "sna.h" |
#include "sna_render.h" |
#include "sna_render_inline.h" |
#include "gen4_vertex.h" |
#ifndef sse2 |
#define sse2 |
#endif |
void gen4_vertex_flush(struct sna *sna) |
{ |
DBG(("%s[%x] = %d\n", __FUNCTION__, |
4*sna->render.vertex_offset, |
sna->render.vertex_index - sna->render.vertex_start)); |
assert(sna->render.vertex_offset); |
assert(sna->render.vertex_index > sna->render.vertex_start); |
sna->kgem.batch[sna->render.vertex_offset] = |
sna->render.vertex_index - sna->render.vertex_start; |
sna->render.vertex_offset = 0; |
} |
int gen4_vertex_finish(struct sna *sna) |
{ |
struct kgem_bo *bo; |
unsigned int i; |
unsigned hint, size; |
DBG(("%s: used=%d / %d\n", __FUNCTION__, |
sna->render.vertex_used, sna->render.vertex_size)); |
assert(sna->render.vertex_offset == 0); |
assert(sna->render.vertex_used); |
sna_vertex_wait__locked(&sna->render); |
/* Note: we only need dword alignment (currently) */ |
bo = sna->render.vbo; |
if (bo) { |
for (i = 0; i < sna->render.nvertex_reloc; i++) { |
DBG(("%s: reloc[%d] = %d\n", __FUNCTION__, |
i, sna->render.vertex_reloc[i])); |
sna->kgem.batch[sna->render.vertex_reloc[i]] = |
kgem_add_reloc(&sna->kgem, |
sna->render.vertex_reloc[i], bo, |
I915_GEM_DOMAIN_VERTEX << 16, |
0); |
} |
assert(!sna->render.active); |
sna->render.nvertex_reloc = 0; |
sna->render.vertex_used = 0; |
sna->render.vertex_index = 0; |
sna->render.vbo = NULL; |
sna->render.vb_id = 0; |
kgem_bo_destroy(&sna->kgem, bo); |
} |
hint = CREATE_GTT_MAP; |
if (bo) |
hint |= CREATE_CACHED | CREATE_NO_THROTTLE; |
size = 256*1024; |
assert(!sna->render.active); |
sna->render.vertices = NULL; |
sna->render.vbo = kgem_create_linear(&sna->kgem, size, hint); |
while (sna->render.vbo == NULL && size > 16*1024) { |
size /= 2; |
sna->render.vbo = kgem_create_linear(&sna->kgem, size, hint); |
} |
if (sna->render.vbo == NULL) |
sna->render.vbo = kgem_create_linear(&sna->kgem, |
256*1024, CREATE_GTT_MAP); |
if (sna->render.vbo) |
sna->render.vertices = kgem_bo_map(&sna->kgem, sna->render.vbo); |
if (sna->render.vertices == NULL) { |
if (sna->render.vbo) { |
kgem_bo_destroy(&sna->kgem, sna->render.vbo); |
sna->render.vbo = NULL; |
} |
sna->render.vertices = sna->render.vertex_data; |
sna->render.vertex_size = ARRAY_SIZE(sna->render.vertex_data); |
return 0; |
} |
if (sna->render.vertex_used) { |
DBG(("%s: copying initial buffer x %d to handle=%d\n", |
__FUNCTION__, |
sna->render.vertex_used, |
sna->render.vbo->handle)); |
assert(sizeof(float)*sna->render.vertex_used <= |
__kgem_bo_size(sna->render.vbo)); |
memcpy(sna->render.vertices, |
sna->render.vertex_data, |
sizeof(float)*sna->render.vertex_used); |
} |
size = __kgem_bo_size(sna->render.vbo)/4; |
if (size >= UINT16_MAX) |
size = UINT16_MAX - 1; |
DBG(("%s: create vbo handle=%d, size=%d\n", |
__FUNCTION__, sna->render.vbo->handle, size)); |
sna->render.vertex_size = size; |
return sna->render.vertex_size - sna->render.vertex_used; |
} |
void gen4_vertex_close(struct sna *sna) |
{ |
struct kgem_bo *bo, *free_bo = NULL; |
unsigned int i, delta = 0; |
assert(sna->render.vertex_offset == 0); |
if (!sna->render.vb_id) |
return; |
DBG(("%s: used=%d, vbo active? %d, vb=%x, nreloc=%d\n", |
__FUNCTION__, sna->render.vertex_used, sna->render.vbo ? sna->render.vbo->handle : 0, |
sna->render.vb_id, sna->render.nvertex_reloc)); |
assert(!sna->render.active); |
bo = sna->render.vbo; |
if (bo) { |
if (sna->render.vertex_size - sna->render.vertex_used < 64) { |
DBG(("%s: discarding vbo (full), handle=%d\n", __FUNCTION__, sna->render.vbo->handle)); |
sna->render.vbo = NULL; |
sna->render.vertices = sna->render.vertex_data; |
sna->render.vertex_size = ARRAY_SIZE(sna->render.vertex_data); |
free_bo = bo; |
} else if (IS_CPU_MAP(bo->map) && !sna->kgem.has_llc) { |
DBG(("%s: converting CPU map to GTT\n", __FUNCTION__)); |
sna->render.vertices = |
kgem_bo_map__gtt(&sna->kgem, sna->render.vbo); |
if (sna->render.vertices == NULL) { |
sna->render.vbo = NULL; |
sna->render.vertices = sna->render.vertex_data; |
sna->render.vertex_size = ARRAY_SIZE(sna->render.vertex_data); |
free_bo = bo; |
} |
} |
} else { |
if (sna->kgem.nbatch + sna->render.vertex_used <= sna->kgem.surface) { |
DBG(("%s: copy to batch: %d @ %d\n", __FUNCTION__, |
sna->render.vertex_used, sna->kgem.nbatch)); |
memcpy(sna->kgem.batch + sna->kgem.nbatch, |
sna->render.vertex_data, |
sna->render.vertex_used * 4); |
delta = sna->kgem.nbatch * 4; |
bo = NULL; |
sna->kgem.nbatch += sna->render.vertex_used; |
} else { |
bo = kgem_create_linear(&sna->kgem, |
4*sna->render.vertex_used, |
CREATE_NO_THROTTLE); |
if (bo && !kgem_bo_write(&sna->kgem, bo, |
sna->render.vertex_data, |
4*sna->render.vertex_used)) { |
kgem_bo_destroy(&sna->kgem, bo); |
bo = NULL; |
} |
DBG(("%s: new vbo: %d\n", __FUNCTION__, |
sna->render.vertex_used)); |
free_bo = bo; |
} |
} |
assert(sna->render.nvertex_reloc); |
for (i = 0; i < sna->render.nvertex_reloc; i++) { |
DBG(("%s: reloc[%d] = %d\n", __FUNCTION__, |
i, sna->render.vertex_reloc[i])); |
sna->kgem.batch[sna->render.vertex_reloc[i]] = |
kgem_add_reloc(&sna->kgem, |
sna->render.vertex_reloc[i], bo, |
I915_GEM_DOMAIN_VERTEX << 16, |
delta); |
} |
sna->render.nvertex_reloc = 0; |
sna->render.vb_id = 0; |
if (sna->render.vbo == NULL) { |
assert(!sna->render.active); |
sna->render.vertex_used = 0; |
sna->render.vertex_index = 0; |
assert(sna->render.vertices == sna->render.vertex_data); |
assert(sna->render.vertex_size == ARRAY_SIZE(sna->render.vertex_data)); |
} |
if (free_bo) |
kgem_bo_destroy(&sna->kgem, free_bo); |
} |
/* specialised vertex emission routines */ |
#define OUT_VERTEX(x,y) vertex_emit_2s(sna, x,y) /* XXX assert(!too_large(x, y)); */ |
#define OUT_VERTEX_F(v) vertex_emit(sna, v) |
force_inline static float |
compute_linear(const struct sna_composite_channel *channel, |
int16_t x, int16_t y) |
{ |
return ((x+channel->offset[0]) * channel->u.linear.dx + |
(y+channel->offset[1]) * channel->u.linear.dy + |
channel->u.linear.offset); |
} |
sse2 inline static void |
emit_texcoord(struct sna *sna, |
const struct sna_composite_channel *channel, |
int16_t x, int16_t y) |
{ |
if (channel->is_solid) { |
OUT_VERTEX_F(x); |
return; |
} |
x += channel->offset[0]; |
y += channel->offset[1]; |
if (channel->is_affine) { |
float s, t; |
sna_get_transformed_coordinates(x, y, |
channel->transform, |
&s, &t); |
OUT_VERTEX_F(s * channel->scale[0]); |
OUT_VERTEX_F(t * channel->scale[1]); |
} else { |
float s, t, w; |
sna_get_transformed_coordinates_3d(x, y, |
channel->transform, |
&s, &t, &w); |
OUT_VERTEX_F(s * channel->scale[0]); |
OUT_VERTEX_F(t * channel->scale[1]); |
OUT_VERTEX_F(w); |
} |
} |
sse2 force_inline static void |
emit_vertex(struct sna *sna, |
const struct sna_composite_op *op, |
int16_t srcX, int16_t srcY, |
int16_t mskX, int16_t mskY, |
int16_t dstX, int16_t dstY) |
{ |
OUT_VERTEX(dstX, dstY); |
emit_texcoord(sna, &op->src, srcX, srcY); |
} |
sse2 fastcall static void |
emit_primitive(struct sna *sna, |
const struct sna_composite_op *op, |
const struct sna_composite_rectangles *r) |
{ |
emit_vertex(sna, op, |
r->src.x + r->width, r->src.y + r->height, |
r->mask.x + r->width, r->mask.y + r->height, |
r->dst.x + r->width, r->dst.y + r->height); |
emit_vertex(sna, op, |
r->src.x, r->src.y + r->height, |
r->mask.x, r->mask.y + r->height, |
r->dst.x, r->dst.y + r->height); |
emit_vertex(sna, op, |
r->src.x, r->src.y, |
r->mask.x, r->mask.y, |
r->dst.x, r->dst.y); |
} |
sse2 force_inline static void |
emit_vertex_mask(struct sna *sna, |
const struct sna_composite_op *op, |
int16_t srcX, int16_t srcY, |
int16_t mskX, int16_t mskY, |
int16_t dstX, int16_t dstY) |
{ |
OUT_VERTEX(dstX, dstY); |
emit_texcoord(sna, &op->src, srcX, srcY); |
emit_texcoord(sna, &op->mask, mskX, mskY); |
} |
sse2 fastcall static void |
emit_primitive_mask(struct sna *sna, |
const struct sna_composite_op *op, |
const struct sna_composite_rectangles *r) |
{ |
emit_vertex_mask(sna, op, |
r->src.x + r->width, r->src.y + r->height, |
r->mask.x + r->width, r->mask.y + r->height, |
r->dst.x + r->width, r->dst.y + r->height); |
emit_vertex_mask(sna, op, |
r->src.x, r->src.y + r->height, |
r->mask.x, r->mask.y + r->height, |
r->dst.x, r->dst.y + r->height); |
emit_vertex_mask(sna, op, |
r->src.x, r->src.y, |
r->mask.x, r->mask.y, |
r->dst.x, r->dst.y); |
} |
sse2 fastcall static void |
emit_primitive_solid(struct sna *sna, |
const struct sna_composite_op *op, |
const struct sna_composite_rectangles *r) |
{ |
float *v; |
union { |
struct sna_coordinate p; |
float f; |
} dst; |
assert(op->floats_per_rect == 6); |
assert((sna->render.vertex_used % 2) == 0); |
v = sna->render.vertices + sna->render.vertex_used; |
sna->render.vertex_used += 6; |
assert(sna->render.vertex_used <= sna->render.vertex_size); |
dst.p.x = r->dst.x + r->width; |
dst.p.y = r->dst.y + r->height; |
v[0] = dst.f; |
dst.p.x = r->dst.x; |
v[2] = dst.f; |
dst.p.y = r->dst.y; |
v[4] = dst.f; |
v[5] = v[3] = v[1] = .5; |
} |
sse2 fastcall static void |
emit_boxes_solid(const struct sna_composite_op *op, |
const BoxRec *box, int nbox, |
float *v) |
{ |
do { |
union { |
struct sna_coordinate p; |
float f; |
} dst; |
dst.p.x = box->x2; |
dst.p.y = box->y2; |
v[0] = dst.f; |
dst.p.x = box->x1; |
v[2] = dst.f; |
dst.p.y = box->y1; |
v[4] = dst.f; |
v[5] = v[3] = v[1] = .5; |
box++; |
v += 6; |
} while (--nbox); |
} |
sse2 fastcall static void |
emit_primitive_linear(struct sna *sna, |
const struct sna_composite_op *op, |
const struct sna_composite_rectangles *r) |
{ |
float *v; |
union { |
struct sna_coordinate p; |
float f; |
} dst; |
assert(op->floats_per_rect == 6); |
assert((sna->render.vertex_used % 2) == 0); |
v = sna->render.vertices + sna->render.vertex_used; |
sna->render.vertex_used += 6; |
assert(sna->render.vertex_used <= sna->render.vertex_size); |
dst.p.x = r->dst.x + r->width; |
dst.p.y = r->dst.y + r->height; |
v[0] = dst.f; |
dst.p.x = r->dst.x; |
v[2] = dst.f; |
dst.p.y = r->dst.y; |
v[4] = dst.f; |
v[1] = compute_linear(&op->src, r->src.x+r->width, r->src.y+r->height); |
v[3] = compute_linear(&op->src, r->src.x, r->src.y+r->height); |
v[5] = compute_linear(&op->src, r->src.x, r->src.y); |
} |
sse2 fastcall static void |
emit_boxes_linear(const struct sna_composite_op *op, |
const BoxRec *box, int nbox, |
float *v) |
{ |
union { |
struct sna_coordinate p; |
float f; |
} dst; |
do { |
dst.p.x = box->x2; |
dst.p.y = box->y2; |
v[0] = dst.f; |
dst.p.x = box->x1; |
v[2] = dst.f; |
dst.p.y = box->y1; |
v[4] = dst.f; |
v[1] = compute_linear(&op->src, box->x2, box->y2); |
v[3] = compute_linear(&op->src, box->x1, box->y2); |
v[5] = compute_linear(&op->src, box->x1, box->y1); |
v += 6; |
box++; |
} while (--nbox); |
} |
sse2 fastcall static void |
emit_primitive_identity_source(struct sna *sna, |
const struct sna_composite_op *op, |
const struct sna_composite_rectangles *r) |
{ |
union { |
struct sna_coordinate p; |
float f; |
} dst; |
float *v; |
assert(op->floats_per_rect == 9); |
assert((sna->render.vertex_used % 3) == 0); |
v = sna->render.vertices + sna->render.vertex_used; |
sna->render.vertex_used += 9; |
dst.p.x = r->dst.x + r->width; |
dst.p.y = r->dst.y + r->height; |
v[0] = dst.f; |
dst.p.x = r->dst.x; |
v[3] = dst.f; |
dst.p.y = r->dst.y; |
v[6] = dst.f; |
v[7] = v[4] = (r->src.x + op->src.offset[0]) * op->src.scale[0]; |
v[1] = v[4] + r->width * op->src.scale[0]; |
v[8] = (r->src.y + op->src.offset[1]) * op->src.scale[1]; |
v[5] = v[2] = v[8] + r->height * op->src.scale[1]; |
} |
sse2 fastcall static void |
emit_boxes_identity_source(const struct sna_composite_op *op, |
const BoxRec *box, int nbox, |
float *v) |
{ |
do { |
union { |
struct sna_coordinate p; |
float f; |
} dst; |
dst.p.x = box->x2; |
dst.p.y = box->y2; |
v[0] = dst.f; |
dst.p.x = box->x1; |
v[3] = dst.f; |
dst.p.y = box->y1; |
v[6] = dst.f; |
v[7] = v[4] = (box->x1 + op->src.offset[0]) * op->src.scale[0]; |
v[1] = (box->x2 + op->src.offset[0]) * op->src.scale[0]; |
v[8] = (box->y1 + op->src.offset[1]) * op->src.scale[1]; |
v[2] = v[5] = (box->y2 + op->src.offset[1]) * op->src.scale[1]; |
v += 9; |
box++; |
} while (--nbox); |
} |
sse2 fastcall static void |
emit_primitive_simple_source(struct sna *sna, |
const struct sna_composite_op *op, |
const struct sna_composite_rectangles *r) |
{ |
float *v; |
union { |
struct sna_coordinate p; |
float f; |
} dst; |
float xx = op->src.transform->matrix[0][0]; |
float x0 = op->src.transform->matrix[0][2]; |
float yy = op->src.transform->matrix[1][1]; |
float y0 = op->src.transform->matrix[1][2]; |
float sx = op->src.scale[0]; |
float sy = op->src.scale[1]; |
int16_t tx = op->src.offset[0]; |
int16_t ty = op->src.offset[1]; |
assert(op->floats_per_rect == 9); |
assert((sna->render.vertex_used % 3) == 0); |
v = sna->render.vertices + sna->render.vertex_used; |
sna->render.vertex_used += 3*3; |
dst.p.x = r->dst.x + r->width; |
dst.p.y = r->dst.y + r->height; |
v[0] = dst.f; |
v[1] = ((r->src.x + r->width + tx) * xx + x0) * sx; |
v[5] = v[2] = ((r->src.y + r->height + ty) * yy + y0) * sy; |
dst.p.x = r->dst.x; |
v[3] = dst.f; |
v[7] = v[4] = ((r->src.x + tx) * xx + x0) * sx; |
dst.p.y = r->dst.y; |
v[6] = dst.f; |
v[8] = ((r->src.y + ty) * yy + y0) * sy; |
} |
sse2 fastcall static void |
emit_boxes_simple_source(const struct sna_composite_op *op, |
const BoxRec *box, int nbox, |
float *v) |
{ |
float xx = op->src.transform->matrix[0][0]; |
float x0 = op->src.transform->matrix[0][2]; |
float yy = op->src.transform->matrix[1][1]; |
float y0 = op->src.transform->matrix[1][2]; |
float sx = op->src.scale[0]; |
float sy = op->src.scale[1]; |
int16_t tx = op->src.offset[0]; |
int16_t ty = op->src.offset[1]; |
do { |
union { |
struct sna_coordinate p; |
float f; |
} dst; |
dst.p.x = box->x2; |
dst.p.y = box->y2; |
v[0] = dst.f; |
v[1] = ((box->x2 + tx) * xx + x0) * sx; |
v[5] = v[2] = ((box->y2 + ty) * yy + y0) * sy; |
dst.p.x = box->x1; |
v[3] = dst.f; |
v[7] = v[4] = ((box->x1 + tx) * xx + x0) * sx; |
dst.p.y = box->y1; |
v[6] = dst.f; |
v[8] = ((box->y1 + ty) * yy + y0) * sy; |
v += 9; |
box++; |
} while (--nbox); |
} |
sse2 fastcall static void |
emit_primitive_affine_source(struct sna *sna, |
const struct sna_composite_op *op, |
const struct sna_composite_rectangles *r) |
{ |
union { |
struct sna_coordinate p; |
float f; |
} dst; |
float *v; |
assert(op->floats_per_rect == 9); |
assert((sna->render.vertex_used % 3) == 0); |
v = sna->render.vertices + sna->render.vertex_used; |
sna->render.vertex_used += 9; |
dst.p.x = r->dst.x + r->width; |
dst.p.y = r->dst.y + r->height; |
v[0] = dst.f; |
_sna_get_transformed_scaled(op->src.offset[0] + r->src.x + r->width, |
op->src.offset[1] + r->src.y + r->height, |
op->src.transform, op->src.scale, |
&v[1], &v[2]); |
dst.p.x = r->dst.x; |
v[3] = dst.f; |
_sna_get_transformed_scaled(op->src.offset[0] + r->src.x, |
op->src.offset[1] + r->src.y + r->height, |
op->src.transform, op->src.scale, |
&v[4], &v[5]); |
dst.p.y = r->dst.y; |
v[6] = dst.f; |
_sna_get_transformed_scaled(op->src.offset[0] + r->src.x, |
op->src.offset[1] + r->src.y, |
op->src.transform, op->src.scale, |
&v[7], &v[8]); |
} |
sse2 fastcall static void |
emit_boxes_affine_source(const struct sna_composite_op *op, |
const BoxRec *box, int nbox, |
float *v) |
{ |
do { |
union { |
struct sna_coordinate p; |
float f; |
} dst; |
dst.p.x = box->x2; |
dst.p.y = box->y2; |
v[0] = dst.f; |
_sna_get_transformed_scaled(op->src.offset[0] + box->x2, |
op->src.offset[1] + box->y2, |
op->src.transform, op->src.scale, |
&v[1], &v[2]); |
dst.p.x = box->x1; |
v[3] = dst.f; |
_sna_get_transformed_scaled(op->src.offset[0] + box->x1, |
op->src.offset[1] + box->y2, |
op->src.transform, op->src.scale, |
&v[4], &v[5]); |
dst.p.y = box->y1; |
v[6] = dst.f; |
_sna_get_transformed_scaled(op->src.offset[0] + box->x1, |
op->src.offset[1] + box->y1, |
op->src.transform, op->src.scale, |
&v[7], &v[8]); |
box++; |
v += 9; |
} while (--nbox); |
} |
sse2 fastcall static void |
emit_primitive_identity_mask(struct sna *sna, |
const struct sna_composite_op *op, |
const struct sna_composite_rectangles *r) |
{ |
union { |
struct sna_coordinate p; |
float f; |
} dst; |
float msk_x, msk_y; |
float w, h; |
float *v; |
msk_x = r->mask.x + op->mask.offset[0]; |
msk_y = r->mask.y + op->mask.offset[1]; |
w = r->width; |
h = r->height; |
DBG(("%s: dst=(%d, %d), mask=(%f, %f) x (%f, %f)\n", |
__FUNCTION__, r->dst.x, r->dst.y, msk_x, msk_y, w, h)); |
assert(op->floats_per_rect == 12); |
assert((sna->render.vertex_used % 4) == 0); |
v = sna->render.vertices + sna->render.vertex_used; |
sna->render.vertex_used += 12; |
dst.p.x = r->dst.x + r->width; |
dst.p.y = r->dst.y + r->height; |
v[0] = dst.f; |
v[2] = (msk_x + w) * op->mask.scale[0]; |
v[7] = v[3] = (msk_y + h) * op->mask.scale[1]; |
dst.p.x = r->dst.x; |
v[4] = dst.f; |
v[10] = v[6] = msk_x * op->mask.scale[0]; |
dst.p.y = r->dst.y; |
v[8] = dst.f; |
v[11] = msk_y * op->mask.scale[1]; |
v[9] = v[5] = v[1] = .5; |
} |
sse2 fastcall static void |
emit_boxes_identity_mask(const struct sna_composite_op *op, |
const BoxRec *box, int nbox, |
float *v) |
{ |
float msk_x = op->mask.offset[0]; |
float msk_y = op->mask.offset[1]; |
do { |
union { |
struct sna_coordinate p; |
float f; |
} dst; |
dst.p.x = box->x2; |
dst.p.y = box->y2; |
v[0] = dst.f; |
v[2] = (msk_x + box->x2) * op->mask.scale[0]; |
v[7] = v[3] = (msk_y + box->y2) * op->mask.scale[1]; |
dst.p.x = box->x1; |
v[4] = dst.f; |
v[10] = v[6] = (msk_x + box->x1) * op->mask.scale[0]; |
dst.p.y = box->y1; |
v[8] = dst.f; |
v[11] = (msk_y + box->y1) * op->mask.scale[1]; |
v[9] = v[5] = v[1] = .5; |
v += 12; |
box++; |
} while (--nbox); |
} |
sse2 fastcall static void |
emit_primitive_linear_identity_mask(struct sna *sna, |
const struct sna_composite_op *op, |
const struct sna_composite_rectangles *r) |
{ |
union { |
struct sna_coordinate p; |
float f; |
} dst; |
float msk_x, msk_y; |
float w, h; |
float *v; |
msk_x = r->mask.x + op->mask.offset[0]; |
msk_y = r->mask.y + op->mask.offset[1]; |
w = r->width; |
h = r->height; |
DBG(("%s: dst=(%d, %d), mask=(%f, %f) x (%f, %f)\n", |
__FUNCTION__, r->dst.x, r->dst.y, msk_x, msk_y, w, h)); |
assert(op->floats_per_rect == 12); |
assert((sna->render.vertex_used % 4) == 0); |
v = sna->render.vertices + sna->render.vertex_used; |
sna->render.vertex_used += 12; |
dst.p.x = r->dst.x + r->width; |
dst.p.y = r->dst.y + r->height; |
v[0] = dst.f; |
v[2] = (msk_x + w) * op->mask.scale[0]; |
v[7] = v[3] = (msk_y + h) * op->mask.scale[1]; |
dst.p.x = r->dst.x; |
v[4] = dst.f; |
v[10] = v[6] = msk_x * op->mask.scale[0]; |
dst.p.y = r->dst.y; |
v[8] = dst.f; |
v[11] = msk_y * op->mask.scale[1]; |
v[1] = compute_linear(&op->src, r->src.x+r->width, r->src.y+r->height); |
v[5] = compute_linear(&op->src, r->src.x, r->src.y+r->height); |
v[9] = compute_linear(&op->src, r->src.x, r->src.y); |
} |
sse2 fastcall static void |
emit_boxes_linear_identity_mask(const struct sna_composite_op *op, |
const BoxRec *box, int nbox, |
float *v) |
{ |
float msk_x = op->mask.offset[0]; |
float msk_y = op->mask.offset[1]; |
do { |
union { |
struct sna_coordinate p; |
float f; |
} dst; |
dst.p.x = box->x2; |
dst.p.y = box->y2; |
v[0] = dst.f; |
v[2] = (msk_x + box->x2) * op->mask.scale[0]; |
v[7] = v[3] = (msk_y + box->y2) * op->mask.scale[1]; |
dst.p.x = box->x1; |
v[4] = dst.f; |
v[10] = v[6] = (msk_x + box->x1) * op->mask.scale[0]; |
dst.p.y = box->y1; |
v[8] = dst.f; |
v[11] = (msk_y + box->y1) * op->mask.scale[1]; |
v[1] = compute_linear(&op->src, box->x2, box->y2); |
v[5] = compute_linear(&op->src, box->x1, box->y2); |
v[9] = compute_linear(&op->src, box->x1, box->y1); |
v += 12; |
box++; |
} while (--nbox); |
} |
sse2 fastcall static void |
emit_primitive_identity_source_mask(struct sna *sna, |
const struct sna_composite_op *op, |
const struct sna_composite_rectangles *r) |
{ |
union { |
struct sna_coordinate p; |
float f; |
} dst; |
float src_x, src_y; |
float msk_x, msk_y; |
float w, h; |
float *v; |
src_x = r->src.x + op->src.offset[0]; |
src_y = r->src.y + op->src.offset[1]; |
msk_x = r->mask.x + op->mask.offset[0]; |
msk_y = r->mask.y + op->mask.offset[1]; |
w = r->width; |
h = r->height; |
assert(op->floats_per_rect == 15); |
assert((sna->render.vertex_used % 5) == 0); |
v = sna->render.vertices + sna->render.vertex_used; |
sna->render.vertex_used += 15; |
dst.p.x = r->dst.x + r->width; |
dst.p.y = r->dst.y + r->height; |
v[0] = dst.f; |
v[1] = (src_x + w) * op->src.scale[0]; |
v[2] = (src_y + h) * op->src.scale[1]; |
v[3] = (msk_x + w) * op->mask.scale[0]; |
v[4] = (msk_y + h) * op->mask.scale[1]; |
dst.p.x = r->dst.x; |
v[5] = dst.f; |
v[6] = src_x * op->src.scale[0]; |
v[7] = v[2]; |
v[8] = msk_x * op->mask.scale[0]; |
v[9] = v[4]; |
dst.p.y = r->dst.y; |
v[10] = dst.f; |
v[11] = v[6]; |
v[12] = src_y * op->src.scale[1]; |
v[13] = v[8]; |
v[14] = msk_y * op->mask.scale[1]; |
} |
sse2 fastcall static void |
emit_primitive_simple_source_identity(struct sna *sna, |
const struct sna_composite_op *op, |
const struct sna_composite_rectangles *r) |
{ |
float *v; |
union { |
struct sna_coordinate p; |
float f; |
} dst; |
float xx = op->src.transform->matrix[0][0]; |
float x0 = op->src.transform->matrix[0][2]; |
float yy = op->src.transform->matrix[1][1]; |
float y0 = op->src.transform->matrix[1][2]; |
float sx = op->src.scale[0]; |
float sy = op->src.scale[1]; |
int16_t tx = op->src.offset[0]; |
int16_t ty = op->src.offset[1]; |
float msk_x = r->mask.x + op->mask.offset[0]; |
float msk_y = r->mask.y + op->mask.offset[1]; |
float w = r->width, h = r->height; |
assert(op->floats_per_rect == 15); |
assert((sna->render.vertex_used % 5) == 0); |
v = sna->render.vertices + sna->render.vertex_used; |
sna->render.vertex_used += 3*5; |
dst.p.x = r->dst.x + r->width; |
dst.p.y = r->dst.y + r->height; |
v[0] = dst.f; |
v[1] = ((r->src.x + r->width + tx) * xx + x0) * sx; |
v[2] = ((r->src.y + r->height + ty) * yy + y0) * sy; |
v[3] = (msk_x + w) * op->mask.scale[0]; |
v[4] = (msk_y + h) * op->mask.scale[1]; |
dst.p.x = r->dst.x; |
v[5] = dst.f; |
v[6] = ((r->src.x + tx) * xx + x0) * sx; |
v[7] = v[2]; |
v[8] = msk_x * op->mask.scale[0]; |
v[9] = v[4]; |
dst.p.y = r->dst.y; |
v[10] = dst.f; |
v[11] = v[6]; |
v[12] = ((r->src.y + ty) * yy + y0) * sy; |
v[13] = v[8]; |
v[14] = msk_y * op->mask.scale[1]; |
} |
sse2 fastcall static void |
emit_primitive_affine_source_identity(struct sna *sna, |
const struct sna_composite_op *op, |
const struct sna_composite_rectangles *r) |
{ |
float *v; |
union { |
struct sna_coordinate p; |
float f; |
} dst; |
float msk_x = r->mask.x + op->mask.offset[0]; |
float msk_y = r->mask.y + op->mask.offset[1]; |
float w = r->width, h = r->height; |
assert(op->floats_per_rect == 15); |
assert((sna->render.vertex_used % 5) == 0); |
v = sna->render.vertices + sna->render.vertex_used; |
sna->render.vertex_used += 3*5; |
dst.p.x = r->dst.x + r->width; |
dst.p.y = r->dst.y + r->height; |
v[0] = dst.f; |
_sna_get_transformed_scaled(op->src.offset[0] + r->src.x + r->width, |
op->src.offset[1] + r->src.y + r->height, |
op->src.transform, op->src.scale, |
&v[1], &v[2]); |
v[3] = (msk_x + w) * op->mask.scale[0]; |
v[4] = (msk_y + h) * op->mask.scale[1]; |
dst.p.x = r->dst.x; |
v[5] = dst.f; |
_sna_get_transformed_scaled(op->src.offset[0] + r->src.x, |
op->src.offset[1] + r->src.y + r->height, |
op->src.transform, op->src.scale, |
&v[6], &v[7]); |
v[8] = msk_x * op->mask.scale[0]; |
v[9] = v[4]; |
dst.p.y = r->dst.y; |
v[10] = dst.f; |
_sna_get_transformed_scaled(op->src.offset[0] + r->src.x, |
op->src.offset[1] + r->src.y, |
op->src.transform, op->src.scale, |
&v[11], &v[12]); |
v[13] = v[8]; |
v[14] = msk_y * op->mask.scale[1]; |
} |
/* SSE4_2 */ |
#if defined(sse4_2) |
sse4_2 fastcall static void |
emit_primitive_linear__sse4_2(struct sna *sna, |
const struct sna_composite_op *op, |
const struct sna_composite_rectangles *r) |
{ |
float *v; |
union { |
struct sna_coordinate p; |
float f; |
} dst; |
assert(op->floats_per_rect == 6); |
assert((sna->render.vertex_used % 2) == 0); |
v = sna->render.vertices + sna->render.vertex_used; |
sna->render.vertex_used += 6; |
assert(sna->render.vertex_used <= sna->render.vertex_size); |
dst.p.x = r->dst.x + r->width; |
dst.p.y = r->dst.y + r->height; |
v[0] = dst.f; |
dst.p.x = r->dst.x; |
v[2] = dst.f; |
dst.p.y = r->dst.y; |
v[4] = dst.f; |
v[1] = compute_linear(&op->src, r->src.x+r->width, r->src.y+r->height); |
v[3] = compute_linear(&op->src, r->src.x, r->src.y+r->height); |
v[5] = compute_linear(&op->src, r->src.x, r->src.y); |
} |
sse4_2 fastcall static void |
emit_boxes_linear__sse4_2(const struct sna_composite_op *op, |
const BoxRec *box, int nbox, |
float *v) |
{ |
union { |
struct sna_coordinate p; |
float f; |
} dst; |
do { |
dst.p.x = box->x2; |
dst.p.y = box->y2; |
v[0] = dst.f; |
dst.p.x = box->x1; |
v[2] = dst.f; |
dst.p.y = box->y1; |
v[4] = dst.f; |
v[1] = compute_linear(&op->src, box->x2, box->y2); |
v[3] = compute_linear(&op->src, box->x1, box->y2); |
v[5] = compute_linear(&op->src, box->x1, box->y1); |
v += 6; |
box++; |
} while (--nbox); |
} |
sse4_2 fastcall static void |
emit_primitive_identity_source__sse4_2(struct sna *sna, |
const struct sna_composite_op *op, |
const struct sna_composite_rectangles *r) |
{ |
union { |
struct sna_coordinate p; |
float f; |
} dst; |
float *v; |
assert(op->floats_per_rect == 9); |
assert((sna->render.vertex_used % 3) == 0); |
v = sna->render.vertices + sna->render.vertex_used; |
sna->render.vertex_used += 9; |
dst.p.x = r->dst.x + r->width; |
dst.p.y = r->dst.y + r->height; |
v[0] = dst.f; |
dst.p.x = r->dst.x; |
v[3] = dst.f; |
dst.p.y = r->dst.y; |
v[6] = dst.f; |
v[7] = v[4] = (r->src.x + op->src.offset[0]) * op->src.scale[0]; |
v[1] = v[4] + r->width * op->src.scale[0]; |
v[8] = (r->src.y + op->src.offset[1]) * op->src.scale[1]; |
v[5] = v[2] = v[8] + r->height * op->src.scale[1]; |
} |
sse4_2 fastcall static void |
emit_boxes_identity_source__sse4_2(const struct sna_composite_op *op, |
const BoxRec *box, int nbox, |
float *v) |
{ |
do { |
union { |
struct sna_coordinate p; |
float f; |
} dst; |
dst.p.x = box->x2; |
dst.p.y = box->y2; |
v[0] = dst.f; |
dst.p.x = box->x1; |
v[3] = dst.f; |
dst.p.y = box->y1; |
v[6] = dst.f; |
v[7] = v[4] = (box->x1 + op->src.offset[0]) * op->src.scale[0]; |
v[1] = (box->x2 + op->src.offset[0]) * op->src.scale[0]; |
v[8] = (box->y1 + op->src.offset[1]) * op->src.scale[1]; |
v[2] = v[5] = (box->y2 + op->src.offset[1]) * op->src.scale[1]; |
v += 9; |
box++; |
} while (--nbox); |
} |
sse4_2 fastcall static void |
emit_primitive_simple_source__sse4_2(struct sna *sna, |
const struct sna_composite_op *op, |
const struct sna_composite_rectangles *r) |
{ |
float *v; |
union { |
struct sna_coordinate p; |
float f; |
} dst; |
float xx = op->src.transform->matrix[0][0]; |
float x0 = op->src.transform->matrix[0][2]; |
float yy = op->src.transform->matrix[1][1]; |
float y0 = op->src.transform->matrix[1][2]; |
float sx = op->src.scale[0]; |
float sy = op->src.scale[1]; |
int16_t tx = op->src.offset[0]; |
int16_t ty = op->src.offset[1]; |
assert(op->floats_per_rect == 9); |
assert((sna->render.vertex_used % 3) == 0); |
v = sna->render.vertices + sna->render.vertex_used; |
sna->render.vertex_used += 3*3; |
dst.p.x = r->dst.x + r->width; |
dst.p.y = r->dst.y + r->height; |
v[0] = dst.f; |
v[1] = ((r->src.x + r->width + tx) * xx + x0) * sx; |
v[5] = v[2] = ((r->src.y + r->height + ty) * yy + y0) * sy; |
dst.p.x = r->dst.x; |
v[3] = dst.f; |
v[7] = v[4] = ((r->src.x + tx) * xx + x0) * sx; |
dst.p.y = r->dst.y; |
v[6] = dst.f; |
v[8] = ((r->src.y + ty) * yy + y0) * sy; |
} |
sse4_2 fastcall static void |
emit_boxes_simple_source__sse4_2(const struct sna_composite_op *op, |
const BoxRec *box, int nbox, |
float *v) |
{ |
float xx = op->src.transform->matrix[0][0]; |
float x0 = op->src.transform->matrix[0][2]; |
float yy = op->src.transform->matrix[1][1]; |
float y0 = op->src.transform->matrix[1][2]; |
float sx = op->src.scale[0]; |
float sy = op->src.scale[1]; |
int16_t tx = op->src.offset[0]; |
int16_t ty = op->src.offset[1]; |
do { |
union { |
struct sna_coordinate p; |
float f; |
} dst; |
dst.p.x = box->x2; |
dst.p.y = box->y2; |
v[0] = dst.f; |
v[1] = ((box->x2 + tx) * xx + x0) * sx; |
v[5] = v[2] = ((box->y2 + ty) * yy + y0) * sy; |
dst.p.x = box->x1; |
v[3] = dst.f; |
v[7] = v[4] = ((box->x1 + tx) * xx + x0) * sx; |
dst.p.y = box->y1; |
v[6] = dst.f; |
v[8] = ((box->y1 + ty) * yy + y0) * sy; |
v += 9; |
box++; |
} while (--nbox); |
} |
sse4_2 fastcall static void |
emit_primitive_identity_mask__sse4_2(struct sna *sna, |
const struct sna_composite_op *op, |
const struct sna_composite_rectangles *r) |
{ |
union { |
struct sna_coordinate p; |
float f; |
} dst; |
float msk_x, msk_y; |
float w, h; |
float *v; |
msk_x = r->mask.x + op->mask.offset[0]; |
msk_y = r->mask.y + op->mask.offset[1]; |
w = r->width; |
h = r->height; |
DBG(("%s: dst=(%d, %d), mask=(%f, %f) x (%f, %f)\n", |
__FUNCTION__, r->dst.x, r->dst.y, msk_x, msk_y, w, h)); |
assert(op->floats_per_rect == 12); |
assert((sna->render.vertex_used % 4) == 0); |
v = sna->render.vertices + sna->render.vertex_used; |
sna->render.vertex_used += 12; |
dst.p.x = r->dst.x + r->width; |
dst.p.y = r->dst.y + r->height; |
v[0] = dst.f; |
v[2] = (msk_x + w) * op->mask.scale[0]; |
v[7] = v[3] = (msk_y + h) * op->mask.scale[1]; |
dst.p.x = r->dst.x; |
v[4] = dst.f; |
v[10] = v[6] = msk_x * op->mask.scale[0]; |
dst.p.y = r->dst.y; |
v[8] = dst.f; |
v[11] = msk_y * op->mask.scale[1]; |
v[9] = v[5] = v[1] = .5; |
} |
sse4_2 fastcall static void |
emit_boxes_identity_mask__sse4_2(const struct sna_composite_op *op, |
const BoxRec *box, int nbox, |
float *v) |
{ |
float msk_x = op->mask.offset[0]; |
float msk_y = op->mask.offset[1]; |
do { |
union { |
struct sna_coordinate p; |
float f; |
} dst; |
dst.p.x = box->x2; |
dst.p.y = box->y2; |
v[0] = dst.f; |
v[2] = (msk_x + box->x2) * op->mask.scale[0]; |
v[7] = v[3] = (msk_y + box->y2) * op->mask.scale[1]; |
dst.p.x = box->x1; |
v[4] = dst.f; |
v[10] = v[6] = (msk_x + box->x1) * op->mask.scale[0]; |
dst.p.y = box->y1; |
v[8] = dst.f; |
v[11] = (msk_y + box->y1) * op->mask.scale[1]; |
v[9] = v[5] = v[1] = .5; |
v += 12; |
box++; |
} while (--nbox); |
} |
sse4_2 fastcall static void |
emit_primitive_linear_identity_mask__sse4_2(struct sna *sna, |
const struct sna_composite_op *op, |
const struct sna_composite_rectangles *r) |
{ |
union { |
struct sna_coordinate p; |
float f; |
} dst; |
float msk_x, msk_y; |
float w, h; |
float *v; |
msk_x = r->mask.x + op->mask.offset[0]; |
msk_y = r->mask.y + op->mask.offset[1]; |
w = r->width; |
h = r->height; |
DBG(("%s: dst=(%d, %d), mask=(%f, %f) x (%f, %f)\n", |
__FUNCTION__, r->dst.x, r->dst.y, msk_x, msk_y, w, h)); |
assert(op->floats_per_rect == 12); |
assert((sna->render.vertex_used % 4) == 0); |
v = sna->render.vertices + sna->render.vertex_used; |
sna->render.vertex_used += 12; |
dst.p.x = r->dst.x + r->width; |
dst.p.y = r->dst.y + r->height; |
v[0] = dst.f; |
v[2] = (msk_x + w) * op->mask.scale[0]; |
v[7] = v[3] = (msk_y + h) * op->mask.scale[1]; |
dst.p.x = r->dst.x; |
v[4] = dst.f; |
v[10] = v[6] = msk_x * op->mask.scale[0]; |
dst.p.y = r->dst.y; |
v[8] = dst.f; |
v[11] = msk_y * op->mask.scale[1]; |
v[1] = compute_linear(&op->src, r->src.x+r->width, r->src.y+r->height); |
v[5] = compute_linear(&op->src, r->src.x, r->src.y+r->height); |
v[9] = compute_linear(&op->src, r->src.x, r->src.y); |
} |
sse4_2 fastcall static void |
emit_boxes_linear_identity_mask__sse4_2(const struct sna_composite_op *op, |
const BoxRec *box, int nbox, |
float *v) |
{ |
float msk_x = op->mask.offset[0]; |
float msk_y = op->mask.offset[1]; |
do { |
union { |
struct sna_coordinate p; |
float f; |
} dst; |
dst.p.x = box->x2; |
dst.p.y = box->y2; |
v[0] = dst.f; |
v[2] = (msk_x + box->x2) * op->mask.scale[0]; |
v[7] = v[3] = (msk_y + box->y2) * op->mask.scale[1]; |
dst.p.x = box->x1; |
v[4] = dst.f; |
v[10] = v[6] = (msk_x + box->x1) * op->mask.scale[0]; |
dst.p.y = box->y1; |
v[8] = dst.f; |
v[11] = (msk_y + box->y1) * op->mask.scale[1]; |
v[1] = compute_linear(&op->src, box->x2, box->y2); |
v[5] = compute_linear(&op->src, box->x1, box->y2); |
v[9] = compute_linear(&op->src, box->x1, box->y1); |
v += 12; |
box++; |
} while (--nbox); |
} |
#endif |
/* AVX2 */ |
#if defined(avx2) |
avx2 fastcall static void |
emit_primitive_linear__avx2(struct sna *sna, |
const struct sna_composite_op *op, |
const struct sna_composite_rectangles *r) |
{ |
float *v; |
union { |
struct sna_coordinate p; |
float f; |
} dst; |
assert(op->floats_per_rect == 6); |
assert((sna->render.vertex_used % 2) == 0); |
v = sna->render.vertices + sna->render.vertex_used; |
sna->render.vertex_used += 6; |
assert(sna->render.vertex_used <= sna->render.vertex_size); |
dst.p.x = r->dst.x + r->width; |
dst.p.y = r->dst.y + r->height; |
v[0] = dst.f; |
dst.p.x = r->dst.x; |
v[2] = dst.f; |
dst.p.y = r->dst.y; |
v[4] = dst.f; |
v[1] = compute_linear(&op->src, r->src.x+r->width, r->src.y+r->height); |
v[3] = compute_linear(&op->src, r->src.x, r->src.y+r->height); |
v[5] = compute_linear(&op->src, r->src.x, r->src.y); |
} |
avx2 fastcall static void |
emit_boxes_linear__avx2(const struct sna_composite_op *op, |
const BoxRec *box, int nbox, |
float *v) |
{ |
union { |
struct sna_coordinate p; |
float f; |
} dst; |
do { |
dst.p.x = box->x2; |
dst.p.y = box->y2; |
v[0] = dst.f; |
dst.p.x = box->x1; |
v[2] = dst.f; |
dst.p.y = box->y1; |
v[4] = dst.f; |
v[1] = compute_linear(&op->src, box->x2, box->y2); |
v[3] = compute_linear(&op->src, box->x1, box->y2); |
v[5] = compute_linear(&op->src, box->x1, box->y1); |
v += 6; |
box++; |
} while (--nbox); |
} |
avx2 fastcall static void |
emit_primitive_identity_source__avx2(struct sna *sna, |
const struct sna_composite_op *op, |
const struct sna_composite_rectangles *r) |
{ |
union { |
struct sna_coordinate p; |
float f; |
} dst; |
float *v; |
assert(op->floats_per_rect == 9); |
assert((sna->render.vertex_used % 3) == 0); |
v = sna->render.vertices + sna->render.vertex_used; |
sna->render.vertex_used += 9; |
dst.p.x = r->dst.x + r->width; |
dst.p.y = r->dst.y + r->height; |
v[0] = dst.f; |
dst.p.x = r->dst.x; |
v[3] = dst.f; |
dst.p.y = r->dst.y; |
v[6] = dst.f; |
v[7] = v[4] = (r->src.x + op->src.offset[0]) * op->src.scale[0]; |
v[1] = v[4] + r->width * op->src.scale[0]; |
v[8] = (r->src.y + op->src.offset[1]) * op->src.scale[1]; |
v[5] = v[2] = v[8] + r->height * op->src.scale[1]; |
} |
avx2 fastcall static void |
emit_boxes_identity_source__avx2(const struct sna_composite_op *op, |
const BoxRec *box, int nbox, |
float *v) |
{ |
do { |
union { |
struct sna_coordinate p; |
float f; |
} dst; |
dst.p.x = box->x2; |
dst.p.y = box->y2; |
v[0] = dst.f; |
dst.p.x = box->x1; |
v[3] = dst.f; |
dst.p.y = box->y1; |
v[6] = dst.f; |
v[7] = v[4] = (box->x1 + op->src.offset[0]) * op->src.scale[0]; |
v[1] = (box->x2 + op->src.offset[0]) * op->src.scale[0]; |
v[8] = (box->y1 + op->src.offset[1]) * op->src.scale[1]; |
v[2] = v[5] = (box->y2 + op->src.offset[1]) * op->src.scale[1]; |
v += 9; |
box++; |
} while (--nbox); |
} |
avx2 fastcall static void |
emit_primitive_simple_source__avx2(struct sna *sna, |
const struct sna_composite_op *op, |
const struct sna_composite_rectangles *r) |
{ |
float *v; |
union { |
struct sna_coordinate p; |
float f; |
} dst; |
float xx = op->src.transform->matrix[0][0]; |
float x0 = op->src.transform->matrix[0][2]; |
float yy = op->src.transform->matrix[1][1]; |
float y0 = op->src.transform->matrix[1][2]; |
float sx = op->src.scale[0]; |
float sy = op->src.scale[1]; |
int16_t tx = op->src.offset[0]; |
int16_t ty = op->src.offset[1]; |
assert(op->floats_per_rect == 9); |
assert((sna->render.vertex_used % 3) == 0); |
v = sna->render.vertices + sna->render.vertex_used; |
sna->render.vertex_used += 3*3; |
dst.p.x = r->dst.x + r->width; |
dst.p.y = r->dst.y + r->height; |
v[0] = dst.f; |
v[1] = ((r->src.x + r->width + tx) * xx + x0) * sx; |
v[5] = v[2] = ((r->src.y + r->height + ty) * yy + y0) * sy; |
dst.p.x = r->dst.x; |
v[3] = dst.f; |
v[7] = v[4] = ((r->src.x + tx) * xx + x0) * sx; |
dst.p.y = r->dst.y; |
v[6] = dst.f; |
v[8] = ((r->src.y + ty) * yy + y0) * sy; |
} |
avx2 fastcall static void |
emit_boxes_simple_source__avx2(const struct sna_composite_op *op, |
const BoxRec *box, int nbox, |
float *v) |
{ |
float xx = op->src.transform->matrix[0][0]; |
float x0 = op->src.transform->matrix[0][2]; |
float yy = op->src.transform->matrix[1][1]; |
float y0 = op->src.transform->matrix[1][2]; |
float sx = op->src.scale[0]; |
float sy = op->src.scale[1]; |
int16_t tx = op->src.offset[0]; |
int16_t ty = op->src.offset[1]; |
do { |
union { |
struct sna_coordinate p; |
float f; |
} dst; |
dst.p.x = box->x2; |
dst.p.y = box->y2; |
v[0] = dst.f; |
v[1] = ((box->x2 + tx) * xx + x0) * sx; |
v[5] = v[2] = ((box->y2 + ty) * yy + y0) * sy; |
dst.p.x = box->x1; |
v[3] = dst.f; |
v[7] = v[4] = ((box->x1 + tx) * xx + x0) * sx; |
dst.p.y = box->y1; |
v[6] = dst.f; |
v[8] = ((box->y1 + ty) * yy + y0) * sy; |
v += 9; |
box++; |
} while (--nbox); |
} |
avx2 fastcall static void |
emit_primitive_identity_mask__avx2(struct sna *sna, |
const struct sna_composite_op *op, |
const struct sna_composite_rectangles *r) |
{ |
union { |
struct sna_coordinate p; |
float f; |
} dst; |
float msk_x, msk_y; |
float w, h; |
float *v; |
msk_x = r->mask.x + op->mask.offset[0]; |
msk_y = r->mask.y + op->mask.offset[1]; |
w = r->width; |
h = r->height; |
DBG(("%s: dst=(%d, %d), mask=(%f, %f) x (%f, %f)\n", |
__FUNCTION__, r->dst.x, r->dst.y, msk_x, msk_y, w, h)); |
assert(op->floats_per_rect == 12); |
assert((sna->render.vertex_used % 4) == 0); |
v = sna->render.vertices + sna->render.vertex_used; |
sna->render.vertex_used += 12; |
dst.p.x = r->dst.x + r->width; |
dst.p.y = r->dst.y + r->height; |
v[0] = dst.f; |
v[2] = (msk_x + w) * op->mask.scale[0]; |
v[7] = v[3] = (msk_y + h) * op->mask.scale[1]; |
dst.p.x = r->dst.x; |
v[4] = dst.f; |
v[10] = v[6] = msk_x * op->mask.scale[0]; |
dst.p.y = r->dst.y; |
v[8] = dst.f; |
v[11] = msk_y * op->mask.scale[1]; |
v[9] = v[5] = v[1] = .5; |
} |
avx2 fastcall static void |
emit_boxes_identity_mask__avx2(const struct sna_composite_op *op, |
const BoxRec *box, int nbox, |
float *v) |
{ |
float msk_x = op->mask.offset[0]; |
float msk_y = op->mask.offset[1]; |
do { |
union { |
struct sna_coordinate p; |
float f; |
} dst; |
dst.p.x = box->x2; |
dst.p.y = box->y2; |
v[0] = dst.f; |
v[2] = (msk_x + box->x2) * op->mask.scale[0]; |
v[7] = v[3] = (msk_y + box->y2) * op->mask.scale[1]; |
dst.p.x = box->x1; |
v[4] = dst.f; |
v[10] = v[6] = (msk_x + box->x1) * op->mask.scale[0]; |
dst.p.y = box->y1; |
v[8] = dst.f; |
v[11] = (msk_y + box->y1) * op->mask.scale[1]; |
v[9] = v[5] = v[1] = .5; |
v += 12; |
box++; |
} while (--nbox); |
} |
avx2 fastcall static void |
emit_primitive_linear_identity_mask__avx2(struct sna *sna, |
const struct sna_composite_op *op, |
const struct sna_composite_rectangles *r) |
{ |
union { |
struct sna_coordinate p; |
float f; |
} dst; |
float msk_x, msk_y; |
float w, h; |
float *v; |
msk_x = r->mask.x + op->mask.offset[0]; |
msk_y = r->mask.y + op->mask.offset[1]; |
w = r->width; |
h = r->height; |
DBG(("%s: dst=(%d, %d), mask=(%f, %f) x (%f, %f)\n", |
__FUNCTION__, r->dst.x, r->dst.y, msk_x, msk_y, w, h)); |
assert(op->floats_per_rect == 12); |
assert((sna->render.vertex_used % 4) == 0); |
v = sna->render.vertices + sna->render.vertex_used; |
sna->render.vertex_used += 12; |
dst.p.x = r->dst.x + r->width; |
dst.p.y = r->dst.y + r->height; |
v[0] = dst.f; |
v[2] = (msk_x + w) * op->mask.scale[0]; |
v[7] = v[3] = (msk_y + h) * op->mask.scale[1]; |
dst.p.x = r->dst.x; |
v[4] = dst.f; |
v[10] = v[6] = msk_x * op->mask.scale[0]; |
dst.p.y = r->dst.y; |
v[8] = dst.f; |
v[11] = msk_y * op->mask.scale[1]; |
v[1] = compute_linear(&op->src, r->src.x+r->width, r->src.y+r->height); |
v[5] = compute_linear(&op->src, r->src.x, r->src.y+r->height); |
v[9] = compute_linear(&op->src, r->src.x, r->src.y); |
} |
avx2 fastcall static void |
emit_boxes_linear_identity_mask__avx2(const struct sna_composite_op *op, |
const BoxRec *box, int nbox, |
float *v) |
{ |
float msk_x = op->mask.offset[0]; |
float msk_y = op->mask.offset[1]; |
do { |
union { |
struct sna_coordinate p; |
float f; |
} dst; |
dst.p.x = box->x2; |
dst.p.y = box->y2; |
v[0] = dst.f; |
v[2] = (msk_x + box->x2) * op->mask.scale[0]; |
v[7] = v[3] = (msk_y + box->y2) * op->mask.scale[1]; |
dst.p.x = box->x1; |
v[4] = dst.f; |
v[10] = v[6] = (msk_x + box->x1) * op->mask.scale[0]; |
dst.p.y = box->y1; |
v[8] = dst.f; |
v[11] = (msk_y + box->y1) * op->mask.scale[1]; |
v[1] = compute_linear(&op->src, box->x2, box->y2); |
v[5] = compute_linear(&op->src, box->x1, box->y2); |
v[9] = compute_linear(&op->src, box->x1, box->y1); |
v += 12; |
box++; |
} while (--nbox); |
} |
#endif |
unsigned gen4_choose_composite_emitter(struct sna *sna, struct sna_composite_op *tmp) |
{ |
unsigned vb; |
if (tmp->mask.bo) { |
if (tmp->mask.transform == NULL) { |
if (tmp->src.is_solid) { |
DBG(("%s: solid, identity mask\n", __FUNCTION__)); |
#if defined(avx2) |
if (sna->cpu_features & AVX2) { |
tmp->prim_emit = emit_primitive_identity_mask__avx2; |
} else |
#endif |
#if defined(sse4_2) |
if (sna->cpu_features & SSE4_2) { |
tmp->prim_emit = emit_primitive_identity_mask__sse4_2; |
} else |
#endif |
{ |
tmp->prim_emit = emit_primitive_identity_mask; |
} |
tmp->floats_per_vertex = 4; |
vb = 1 | 2 << 2; |
} else if (tmp->src.is_linear) { |
DBG(("%s: linear, identity mask\n", __FUNCTION__)); |
#if defined(avx2) |
if (sna->cpu_features & AVX2) { |
tmp->prim_emit = emit_primitive_linear_identity_mask__avx2; |
} else |
#endif |
#if defined(sse4_2) |
if (sna->cpu_features & SSE4_2) { |
tmp->prim_emit = emit_primitive_linear_identity_mask__sse4_2; |
} else |
#endif |
{ |
tmp->prim_emit = emit_primitive_linear_identity_mask; |
} |
tmp->floats_per_vertex = 4; |
vb = 1 | 2 << 2; |
} else if (tmp->src.transform == NULL) { |
DBG(("%s: identity source, identity mask\n", __FUNCTION__)); |
tmp->prim_emit = emit_primitive_identity_source_mask; |
tmp->floats_per_vertex = 5; |
vb = 2 << 2 | 2; |
} else if (tmp->src.is_affine) { |
tmp->src.scale[0] /= tmp->src.transform->matrix[2][2]; |
tmp->src.scale[1] /= tmp->src.transform->matrix[2][2]; |
if (!sna_affine_transform_is_rotation(tmp->src.transform)) { |
DBG(("%s: simple src, identity mask\n", __FUNCTION__)); |
tmp->prim_emit = emit_primitive_simple_source_identity; |
} else { |
DBG(("%s: affine src, identity mask\n", __FUNCTION__)); |
tmp->prim_emit = emit_primitive_affine_source_identity; |
} |
tmp->floats_per_vertex = 5; |
vb = 2 << 2 | 2; |
} else { |
DBG(("%s: projective source, identity mask\n", __FUNCTION__)); |
tmp->prim_emit = emit_primitive_mask; |
tmp->floats_per_vertex = 6; |
vb = 2 << 2 | 3; |
} |
} else { |
tmp->prim_emit = emit_primitive_mask; |
tmp->floats_per_vertex = 1; |
vb = 0; |
if (tmp->mask.is_solid) { |
tmp->floats_per_vertex += 1; |
vb |= 1 << 2; |
} else if (tmp->mask.is_affine) { |
tmp->floats_per_vertex += 2; |
vb |= 2 << 2; |
}else { |
tmp->floats_per_vertex += 3; |
vb |= 3 << 2; |
} |
if (tmp->src.is_solid) { |
tmp->floats_per_vertex += 1; |
vb |= 1; |
} else if (tmp->src.is_affine) { |
tmp->floats_per_vertex += 2; |
vb |= 2 ; |
}else { |
tmp->floats_per_vertex += 3; |
vb |= 3; |
} |
DBG(("%s: general mask: floats-per-vertex=%d, vb=%x\n", |
__FUNCTION__,tmp->floats_per_vertex, vb)); |
} |
} else { |
#if 0 |
if (tmp->src.is_solid) { |
DBG(("%s: solid, no mask\n", __FUNCTION__)); |
tmp->prim_emit = emit_primitive_solid; |
if (tmp->src.is_opaque && tmp->op == PictOpOver) |
tmp->op = PictOpSrc; |
tmp->floats_per_vertex = 2; |
vb = 1; |
} else if (tmp->src.is_linear) { |
DBG(("%s: linear, no mask\n", __FUNCTION__)); |
#if defined(avx2) |
if (sna->cpu_features & AVX2) { |
tmp->prim_emit = emit_primitive_linear__avx2; |
} else |
#endif |
#if defined(sse4_2) |
if (sna->cpu_features & SSE4_2) { |
tmp->prim_emit = emit_primitive_linear__sse4_2; |
} else |
#endif |
{ |
tmp->prim_emit = emit_primitive_linear; |
} |
tmp->floats_per_vertex = 2; |
vb = 1; |
} else if (tmp->src.transform == NULL) { |
DBG(("%s: identity src, no mask\n", __FUNCTION__)); |
#if defined(avx2) |
if (sna->cpu_features & AVX2) { |
tmp->prim_emit = emit_primitive_identity_source__avx2; |
} else |
#endif |
#if defined(sse4_2) |
if (sna->cpu_features & SSE4_2) { |
tmp->prim_emit = emit_primitive_identity_source__sse4_2; |
} else |
#endif |
{ |
tmp->prim_emit = emit_primitive_identity_source; |
} |
tmp->floats_per_vertex = 3; |
vb = 2; |
} else if (tmp->src.is_affine) { |
tmp->src.scale[0] /= tmp->src.transform->matrix[2][2]; |
tmp->src.scale[1] /= tmp->src.transform->matrix[2][2]; |
if (!sna_affine_transform_is_rotation(tmp->src.transform)) { |
DBG(("%s: simple src, no mask\n", __FUNCTION__)); |
#if defined(avx2) |
if (sna->cpu_features & AVX2) { |
tmp->prim_emit = emit_primitive_simple_source__avx2; |
} else |
#endif |
#if defined(sse4_2) |
if (sna->cpu_features & SSE4_2) { |
tmp->prim_emit = emit_primitive_simple_source__sse4_2; |
} else |
#endif |
{ |
tmp->prim_emit = emit_primitive_simple_source; |
} |
} else { |
DBG(("%s: affine src, no mask\n", __FUNCTION__)); |
tmp->prim_emit = emit_primitive_affine_source; |
} |
tmp->floats_per_vertex = 3; |
vb = 2; |
} else { |
DBG(("%s: projective src, no mask\n", __FUNCTION__)); |
assert(!tmp->src.is_solid); |
tmp->prim_emit = emit_primitive; |
tmp->floats_per_vertex = 4; |
vb = 3; |
} |
#endif |
} |
tmp->floats_per_rect = 3 * tmp->floats_per_vertex; |
return vb; |
} |
/drivers/video/Intel-2D/sna/gen4_vertex.h |
---|
0,0 → 1,15 |
#ifndef GEN4_VERTEX_H |
#define GEN4_VERTEX_H |
#include "compiler.h" |
#include "sna.h" |
#include "sna_render.h" |
void gen4_vertex_flush(struct sna *sna); |
int gen4_vertex_finish(struct sna *sna); |
void gen4_vertex_close(struct sna *sna); |
unsigned gen4_choose_composite_emitter(struct sna *sna, struct sna_composite_op *tmp); |
#endif /* GEN4_VERTEX_H */ |
/drivers/video/Intel-2D/sna/gen5_render.c |
---|
0,0 → 1,2473 |
/* |
* Copyright © 2006,2008,2011 Intel Corporation |
* Copyright © 2007 Red Hat, Inc. |
* |
* Permission is hereby granted, free of charge, to any person obtaining a |
* copy of this software and associated documentation files (the "Software"), |
* to deal in the Software without restriction, including without limitation |
* the rights to use, copy, modify, merge, publish, distribute, sublicense, |
* and/or sell copies of the Software, and to permit persons to whom the |
* Software is furnished to do so, subject to the following conditions: |
* |
* The above copyright notice and this permission notice (including the next |
* paragraph) shall be included in all copies or substantial portions of the |
* Software. |
* |
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL |
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, |
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE |
* SOFTWARE. |
* |
* Authors: |
* Wang Zhenyu <zhenyu.z.wang@sna.com> |
* Eric Anholt <eric@anholt.net> |
* Carl Worth <cworth@redhat.com> |
* Keith Packard <keithp@keithp.com> |
* Chris Wilson <chris@chris-wilson.co.uk> |
* |
*/ |
#ifdef HAVE_CONFIG_H |
#include "config.h" |
#endif |
#include "sna.h" |
#include "sna_reg.h" |
#include "sna_render.h" |
#include "sna_render_inline.h" |
//#include "sna_video.h" |
#include "brw/brw.h" |
#include "gen5_render.h" |
#include "gen4_source.h" |
#include "gen4_vertex.h" |
#define NO_COMPOSITE 0 |
#define NO_COMPOSITE_SPANS 0 |
#define PREFER_BLT_FILL 1 |
#define DBG_NO_STATE_CACHE 0 |
#define DBG_NO_SURFACE_CACHE 0 |
#define MAX_3D_SIZE 8192 |
#define GEN5_GRF_BLOCKS(nreg) ((nreg + 15) / 16 - 1) |
/* Set up a default static partitioning of the URB, which is supposed to |
* allow anything we would want to do, at potentially lower performance. |
*/ |
#define URB_CS_ENTRY_SIZE 1 |
#define URB_CS_ENTRIES 0 |
#define URB_VS_ENTRY_SIZE 1 |
#define URB_VS_ENTRIES 256 /* minimum of 8 */ |
#define URB_GS_ENTRY_SIZE 0 |
#define URB_GS_ENTRIES 0 |
#define URB_CLIP_ENTRY_SIZE 0 |
#define URB_CLIP_ENTRIES 0 |
#define URB_SF_ENTRY_SIZE 2 |
#define URB_SF_ENTRIES 64 |
/* |
* this program computes dA/dx and dA/dy for the texture coordinates along |
* with the base texture coordinate. It was extracted from the Mesa driver |
*/ |
#define SF_KERNEL_NUM_GRF 16 |
#define SF_MAX_THREADS 48 |
#define PS_KERNEL_NUM_GRF 32 |
#define PS_MAX_THREADS 72 |
static const uint32_t ps_kernel_packed_static[][4] = { |
#include "exa_wm_xy.g5b" |
#include "exa_wm_src_affine.g5b" |
#include "exa_wm_src_sample_argb.g5b" |
#include "exa_wm_yuv_rgb.g5b" |
#include "exa_wm_write.g5b" |
}; |
static const uint32_t ps_kernel_planar_static[][4] = { |
#include "exa_wm_xy.g5b" |
#include "exa_wm_src_affine.g5b" |
#include "exa_wm_src_sample_planar.g5b" |
#include "exa_wm_yuv_rgb.g5b" |
#include "exa_wm_write.g5b" |
}; |
#define NOKERNEL(kernel_enum, func, masked) \ |
[kernel_enum] = {func, 0, masked} |
#define KERNEL(kernel_enum, kernel, masked) \ |
[kernel_enum] = {&kernel, sizeof(kernel), masked} |
static const struct wm_kernel_info { |
const void *data; |
unsigned int size; |
bool has_mask; |
} wm_kernels[] = { |
NOKERNEL(WM_KERNEL, brw_wm_kernel__affine, false), |
NOKERNEL(WM_KERNEL_P, brw_wm_kernel__projective, false), |
NOKERNEL(WM_KERNEL_MASK, brw_wm_kernel__affine_mask, true), |
NOKERNEL(WM_KERNEL_MASK_P, brw_wm_kernel__projective_mask, true), |
NOKERNEL(WM_KERNEL_MASKCA, brw_wm_kernel__affine_mask_ca, true), |
NOKERNEL(WM_KERNEL_MASKCA_P, brw_wm_kernel__projective_mask_ca, true), |
NOKERNEL(WM_KERNEL_MASKSA, brw_wm_kernel__affine_mask_sa, true), |
NOKERNEL(WM_KERNEL_MASKSA_P, brw_wm_kernel__projective_mask_sa, true), |
NOKERNEL(WM_KERNEL_OPACITY, brw_wm_kernel__affine_opacity, true), |
NOKERNEL(WM_KERNEL_OPACITY_P, brw_wm_kernel__projective_opacity, true), |
KERNEL(WM_KERNEL_VIDEO_PLANAR, ps_kernel_planar_static, false), |
KERNEL(WM_KERNEL_VIDEO_PACKED, ps_kernel_packed_static, false), |
}; |
#undef KERNEL |
static const struct blendinfo { |
bool src_alpha; |
uint32_t src_blend; |
uint32_t dst_blend; |
} gen5_blend_op[] = { |
/* Clear */ {0, GEN5_BLENDFACTOR_ZERO, GEN5_BLENDFACTOR_ZERO}, |
/* Src */ {0, GEN5_BLENDFACTOR_ONE, GEN5_BLENDFACTOR_ZERO}, |
/* Dst */ {0, GEN5_BLENDFACTOR_ZERO, GEN5_BLENDFACTOR_ONE}, |
/* Over */ {1, GEN5_BLENDFACTOR_ONE, GEN5_BLENDFACTOR_INV_SRC_ALPHA}, |
/* OverReverse */ {0, GEN5_BLENDFACTOR_INV_DST_ALPHA, GEN5_BLENDFACTOR_ONE}, |
/* In */ {0, GEN5_BLENDFACTOR_DST_ALPHA, GEN5_BLENDFACTOR_ZERO}, |
/* InReverse */ {1, GEN5_BLENDFACTOR_ZERO, GEN5_BLENDFACTOR_SRC_ALPHA}, |
/* Out */ {0, GEN5_BLENDFACTOR_INV_DST_ALPHA, GEN5_BLENDFACTOR_ZERO}, |
/* OutReverse */ {1, GEN5_BLENDFACTOR_ZERO, GEN5_BLENDFACTOR_INV_SRC_ALPHA}, |
/* Atop */ {1, GEN5_BLENDFACTOR_DST_ALPHA, GEN5_BLENDFACTOR_INV_SRC_ALPHA}, |
/* AtopReverse */ {1, GEN5_BLENDFACTOR_INV_DST_ALPHA, GEN5_BLENDFACTOR_SRC_ALPHA}, |
/* Xor */ {1, GEN5_BLENDFACTOR_INV_DST_ALPHA, GEN5_BLENDFACTOR_INV_SRC_ALPHA}, |
/* Add */ {0, GEN5_BLENDFACTOR_ONE, GEN5_BLENDFACTOR_ONE}, |
}; |
/** |
* Highest-valued BLENDFACTOR used in gen5_blend_op. |
* |
* This leaves out GEN5_BLENDFACTOR_INV_DST_COLOR, |
* GEN5_BLENDFACTOR_INV_CONST_{COLOR,ALPHA}, |
* GEN5_BLENDFACTOR_INV_SRC1_{COLOR,ALPHA} |
*/ |
#define GEN5_BLENDFACTOR_COUNT (GEN5_BLENDFACTOR_INV_DST_ALPHA + 1) |
#define BLEND_OFFSET(s, d) \ |
(((s) * GEN5_BLENDFACTOR_COUNT + (d)) * 64) |
#define SAMPLER_OFFSET(sf, se, mf, me, k) \ |
((((((sf) * EXTEND_COUNT + (se)) * FILTER_COUNT + (mf)) * EXTEND_COUNT + (me)) * KERNEL_COUNT + (k)) * 64) |
static bool |
gen5_emit_pipelined_pointers(struct sna *sna, |
const struct sna_composite_op *op, |
int blend, int kernel); |
#define OUT_BATCH(v) batch_emit(sna, v) |
#define OUT_VERTEX(x,y) vertex_emit_2s(sna, x,y) |
#define OUT_VERTEX_F(v) vertex_emit(sna, v) |
static inline bool too_large(int width, int height) |
{ |
return width > MAX_3D_SIZE || height > MAX_3D_SIZE; |
} |
static int |
gen5_choose_composite_kernel(int op, bool has_mask, bool is_ca, bool is_affine) |
{ |
int base; |
if (has_mask) { |
if (is_ca) { |
if (gen5_blend_op[op].src_alpha) |
base = WM_KERNEL_MASKSA; |
else |
base = WM_KERNEL_MASKCA; |
} else |
base = WM_KERNEL_MASK; |
} else |
base = WM_KERNEL; |
return base + !is_affine; |
} |
static bool gen5_magic_ca_pass(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
struct gen5_render_state *state = &sna->render_state.gen5; |
if (!op->need_magic_ca_pass) |
return false; |
assert(sna->render.vertex_index > sna->render.vertex_start); |
DBG(("%s: CA fixup\n", __FUNCTION__)); |
assert(op->mask.bo != NULL); |
assert(op->has_component_alpha); |
gen5_emit_pipelined_pointers |
(sna, op, PictOpAdd, |
gen5_choose_composite_kernel(PictOpAdd, |
true, true, op->is_affine)); |
OUT_BATCH(GEN5_3DPRIMITIVE | |
GEN5_3DPRIMITIVE_VERTEX_SEQUENTIAL | |
(_3DPRIM_RECTLIST << GEN5_3DPRIMITIVE_TOPOLOGY_SHIFT) | |
(0 << 9) | |
4); |
OUT_BATCH(sna->render.vertex_index - sna->render.vertex_start); |
OUT_BATCH(sna->render.vertex_start); |
OUT_BATCH(1); /* single instance */ |
OUT_BATCH(0); /* start instance location */ |
OUT_BATCH(0); /* index buffer offset, ignored */ |
state->last_primitive = sna->kgem.nbatch; |
return true; |
} |
static uint32_t gen5_get_blend(int op, |
bool has_component_alpha, |
uint32_t dst_format) |
{ |
uint32_t src, dst; |
src = GEN5_BLENDFACTOR_ONE; //gen6_blend_op[op].src_blend; |
dst = GEN5_BLENDFACTOR_INV_SRC_ALPHA; //gen6_blend_op[op].dst_blend; |
#if 0 |
/* If there's no dst alpha channel, adjust the blend op so that we'll treat |
* it as always 1. |
*/ |
if (PICT_FORMAT_A(dst_format) == 0) { |
if (src == GEN5_BLENDFACTOR_DST_ALPHA) |
src = GEN5_BLENDFACTOR_ONE; |
else if (src == GEN5_BLENDFACTOR_INV_DST_ALPHA) |
src = GEN5_BLENDFACTOR_ZERO; |
} |
/* If the source alpha is being used, then we should only be in a |
* case where the source blend factor is 0, and the source blend |
* value is the mask channels multiplied by the source picture's alpha. |
*/ |
if (has_component_alpha && gen5_blend_op[op].src_alpha) { |
if (dst == GEN5_BLENDFACTOR_SRC_ALPHA) |
dst = GEN5_BLENDFACTOR_SRC_COLOR; |
else if (dst == GEN5_BLENDFACTOR_INV_SRC_ALPHA) |
dst = GEN5_BLENDFACTOR_INV_SRC_COLOR; |
} |
#endif |
DBG(("blend op=%d, dst=%x [A=%d] => src=%d, dst=%d => offset=%x\n", |
op, dst_format, PICT_FORMAT_A(dst_format), |
src, dst, BLEND_OFFSET(src, dst))); |
return BLEND_OFFSET(src, dst); |
} |
static uint32_t gen5_get_card_format(PictFormat format) |
{ |
switch (format) { |
default: |
return -1; |
case PICT_a8r8g8b8: |
return GEN5_SURFACEFORMAT_B8G8R8A8_UNORM; |
case PICT_x8r8g8b8: |
return GEN5_SURFACEFORMAT_B8G8R8X8_UNORM; |
case PICT_a8b8g8r8: |
return GEN5_SURFACEFORMAT_R8G8B8A8_UNORM; |
case PICT_x8b8g8r8: |
return GEN5_SURFACEFORMAT_R8G8B8X8_UNORM; |
case PICT_a2r10g10b10: |
return GEN5_SURFACEFORMAT_B10G10R10A2_UNORM; |
case PICT_x2r10g10b10: |
return GEN5_SURFACEFORMAT_B10G10R10X2_UNORM; |
case PICT_r8g8b8: |
return GEN5_SURFACEFORMAT_R8G8B8_UNORM; |
case PICT_r5g6b5: |
return GEN5_SURFACEFORMAT_B5G6R5_UNORM; |
case PICT_a1r5g5b5: |
return GEN5_SURFACEFORMAT_B5G5R5A1_UNORM; |
case PICT_a8: |
return GEN5_SURFACEFORMAT_A8_UNORM; |
case PICT_a4r4g4b4: |
return GEN5_SURFACEFORMAT_B4G4R4A4_UNORM; |
} |
} |
static uint32_t gen5_get_dest_format(PictFormat format) |
{ |
return GEN5_SURFACEFORMAT_B8G8R8A8_UNORM; |
#if 0 |
switch (format) { |
default: |
return -1; |
case PICT_a8r8g8b8: |
case PICT_x8r8g8b8: |
return GEN5_SURFACEFORMAT_B8G8R8A8_UNORM; |
case PICT_a8b8g8r8: |
case PICT_x8b8g8r8: |
return GEN5_SURFACEFORMAT_R8G8B8A8_UNORM; |
case PICT_a2r10g10b10: |
case PICT_x2r10g10b10: |
return GEN5_SURFACEFORMAT_B10G10R10A2_UNORM; |
case PICT_r5g6b5: |
return GEN5_SURFACEFORMAT_B5G6R5_UNORM; |
case PICT_x1r5g5b5: |
case PICT_a1r5g5b5: |
return GEN5_SURFACEFORMAT_B5G5R5A1_UNORM; |
case PICT_a8: |
return GEN5_SURFACEFORMAT_A8_UNORM; |
case PICT_a4r4g4b4: |
case PICT_x4r4g4b4: |
return GEN5_SURFACEFORMAT_B4G4R4A4_UNORM; |
} |
#endif |
} |
typedef struct gen5_surface_state_padded { |
struct gen5_surface_state state; |
char pad[32 - sizeof(struct gen5_surface_state)]; |
} gen5_surface_state_padded; |
static void null_create(struct sna_static_stream *stream) |
{ |
/* A bunch of zeros useful for legacy border color and depth-stencil */ |
sna_static_stream_map(stream, 64, 64); |
} |
static void |
sampler_state_init(struct gen5_sampler_state *sampler_state, |
sampler_filter_t filter, |
sampler_extend_t extend) |
{ |
sampler_state->ss0.lod_preclamp = 1; /* GL mode */ |
/* We use the legacy mode to get the semantics specified by |
* the Render extension. */ |
sampler_state->ss0.border_color_mode = GEN5_BORDER_COLOR_MODE_LEGACY; |
switch (filter) { |
default: |
case SAMPLER_FILTER_NEAREST: |
sampler_state->ss0.min_filter = GEN5_MAPFILTER_NEAREST; |
sampler_state->ss0.mag_filter = GEN5_MAPFILTER_NEAREST; |
break; |
case SAMPLER_FILTER_BILINEAR: |
sampler_state->ss0.min_filter = GEN5_MAPFILTER_LINEAR; |
sampler_state->ss0.mag_filter = GEN5_MAPFILTER_LINEAR; |
break; |
} |
switch (extend) { |
default: |
case SAMPLER_EXTEND_NONE: |
sampler_state->ss1.r_wrap_mode = GEN5_TEXCOORDMODE_CLAMP_BORDER; |
sampler_state->ss1.s_wrap_mode = GEN5_TEXCOORDMODE_CLAMP_BORDER; |
sampler_state->ss1.t_wrap_mode = GEN5_TEXCOORDMODE_CLAMP_BORDER; |
break; |
case SAMPLER_EXTEND_REPEAT: |
sampler_state->ss1.r_wrap_mode = GEN5_TEXCOORDMODE_WRAP; |
sampler_state->ss1.s_wrap_mode = GEN5_TEXCOORDMODE_WRAP; |
sampler_state->ss1.t_wrap_mode = GEN5_TEXCOORDMODE_WRAP; |
break; |
case SAMPLER_EXTEND_PAD: |
sampler_state->ss1.r_wrap_mode = GEN5_TEXCOORDMODE_CLAMP; |
sampler_state->ss1.s_wrap_mode = GEN5_TEXCOORDMODE_CLAMP; |
sampler_state->ss1.t_wrap_mode = GEN5_TEXCOORDMODE_CLAMP; |
break; |
case SAMPLER_EXTEND_REFLECT: |
sampler_state->ss1.r_wrap_mode = GEN5_TEXCOORDMODE_MIRROR; |
sampler_state->ss1.s_wrap_mode = GEN5_TEXCOORDMODE_MIRROR; |
sampler_state->ss1.t_wrap_mode = GEN5_TEXCOORDMODE_MIRROR; |
break; |
} |
} |
static uint32_t |
gen5_tiling_bits(uint32_t tiling) |
{ |
switch (tiling) { |
default: assert(0); |
case I915_TILING_NONE: return 0; |
case I915_TILING_X: return GEN5_SURFACE_TILED; |
case I915_TILING_Y: return GEN5_SURFACE_TILED | GEN5_SURFACE_TILED_Y; |
} |
} |
/** |
* Sets up the common fields for a surface state buffer for the given |
* picture in the given surface state buffer. |
*/ |
static uint32_t |
gen5_bind_bo(struct sna *sna, |
struct kgem_bo *bo, |
uint32_t width, |
uint32_t height, |
uint32_t format, |
bool is_dst) |
{ |
uint32_t domains; |
uint16_t offset; |
uint32_t *ss; |
/* After the first bind, we manage the cache domains within the batch */ |
if (!DBG_NO_SURFACE_CACHE) { |
offset = kgem_bo_get_binding(bo, format | is_dst << 31); |
if (offset) { |
if (is_dst) |
kgem_bo_mark_dirty(bo); |
return offset * sizeof(uint32_t); |
} |
} |
offset = sna->kgem.surface -= |
sizeof(struct gen5_surface_state_padded) / sizeof(uint32_t); |
ss = sna->kgem.batch + offset; |
ss[0] = (GEN5_SURFACE_2D << GEN5_SURFACE_TYPE_SHIFT | |
GEN5_SURFACE_BLEND_ENABLED | |
format << GEN5_SURFACE_FORMAT_SHIFT); |
if (is_dst) { |
ss[0] |= GEN5_SURFACE_RC_READ_WRITE; |
domains = I915_GEM_DOMAIN_RENDER << 16 | I915_GEM_DOMAIN_RENDER; |
} else |
domains = I915_GEM_DOMAIN_SAMPLER << 16; |
ss[1] = kgem_add_reloc(&sna->kgem, offset + 1, bo, domains, 0); |
ss[2] = ((width - 1) << GEN5_SURFACE_WIDTH_SHIFT | |
(height - 1) << GEN5_SURFACE_HEIGHT_SHIFT); |
ss[3] = (gen5_tiling_bits(bo->tiling) | |
(bo->pitch - 1) << GEN5_SURFACE_PITCH_SHIFT); |
ss[4] = 0; |
ss[5] = 0; |
kgem_bo_set_binding(bo, format | is_dst << 31, offset); |
DBG(("[%x] bind bo(handle=%d, addr=%d), format=%d, width=%d, height=%d, pitch=%d, tiling=%d -> %s\n", |
offset, bo->handle, ss[1], |
format, width, height, bo->pitch, bo->tiling, |
domains & 0xffff ? "render" : "sampler")); |
return offset * sizeof(uint32_t); |
} |
static void gen5_emit_vertex_buffer(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
int id = op->u.gen5.ve_id; |
assert((sna->render.vb_id & (1 << id)) == 0); |
OUT_BATCH(GEN5_3DSTATE_VERTEX_BUFFERS | 3); |
OUT_BATCH(id << VB0_BUFFER_INDEX_SHIFT | VB0_VERTEXDATA | |
(4*op->floats_per_vertex << VB0_BUFFER_PITCH_SHIFT)); |
assert(sna->render.nvertex_reloc < ARRAY_SIZE(sna->render.vertex_reloc)); |
sna->render.vertex_reloc[sna->render.nvertex_reloc++] = sna->kgem.nbatch; |
OUT_BATCH(0); |
OUT_BATCH(~0); /* max address: disabled */ |
OUT_BATCH(0); |
sna->render.vb_id |= 1 << id; |
} |
static void gen5_emit_primitive(struct sna *sna) |
{ |
if (sna->kgem.nbatch == sna->render_state.gen5.last_primitive) { |
sna->render.vertex_offset = sna->kgem.nbatch - 5; |
return; |
} |
OUT_BATCH(GEN5_3DPRIMITIVE | |
GEN5_3DPRIMITIVE_VERTEX_SEQUENTIAL | |
(_3DPRIM_RECTLIST << GEN5_3DPRIMITIVE_TOPOLOGY_SHIFT) | |
(0 << 9) | |
4); |
sna->render.vertex_offset = sna->kgem.nbatch; |
OUT_BATCH(0); /* vertex count, to be filled in later */ |
OUT_BATCH(sna->render.vertex_index); |
OUT_BATCH(1); /* single instance */ |
OUT_BATCH(0); /* start instance location */ |
OUT_BATCH(0); /* index buffer offset, ignored */ |
sna->render.vertex_start = sna->render.vertex_index; |
sna->render_state.gen5.last_primitive = sna->kgem.nbatch; |
} |
static bool gen5_rectangle_begin(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
int id = op->u.gen5.ve_id; |
int ndwords; |
if (sna_vertex_wait__locked(&sna->render) && sna->render.vertex_offset) |
return true; |
ndwords = op->need_magic_ca_pass ? 20 : 6; |
if ((sna->render.vb_id & (1 << id)) == 0) |
ndwords += 5; |
if (!kgem_check_batch(&sna->kgem, ndwords)) |
return false; |
if ((sna->render.vb_id & (1 << id)) == 0) |
gen5_emit_vertex_buffer(sna, op); |
if (sna->render.vertex_offset == 0) |
gen5_emit_primitive(sna); |
return true; |
} |
static int gen5_get_rectangles__flush(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
/* Preventing discarding new vbo after lock contention */ |
if (sna_vertex_wait__locked(&sna->render)) { |
int rem = vertex_space(sna); |
if (rem > op->floats_per_rect) |
return rem; |
} |
if (!kgem_check_batch(&sna->kgem, op->need_magic_ca_pass ? 20 : 6)) |
return 0; |
if (!kgem_check_reloc_and_exec(&sna->kgem, 2)) |
return 0; |
if (sna->render.vertex_offset) { |
gen4_vertex_flush(sna); |
if (gen5_magic_ca_pass(sna, op)) |
gen5_emit_pipelined_pointers(sna, op, op->op, |
op->u.gen5.wm_kernel); |
} |
return gen4_vertex_finish(sna); |
} |
inline static int gen5_get_rectangles(struct sna *sna, |
const struct sna_composite_op *op, |
int want, |
void (*emit_state)(struct sna *sna, |
const struct sna_composite_op *op)) |
{ |
int rem; |
assert(want); |
start: |
rem = vertex_space(sna); |
if (unlikely(rem < op->floats_per_rect)) { |
DBG(("flushing vbo for %s: %d < %d\n", |
__FUNCTION__, rem, op->floats_per_rect)); |
rem = gen5_get_rectangles__flush(sna, op); |
if (unlikely (rem == 0)) |
goto flush; |
} |
if (unlikely(sna->render.vertex_offset == 0)) { |
if (!gen5_rectangle_begin(sna, op)) |
goto flush; |
else |
goto start; |
} |
assert(rem <= vertex_space(sna)); |
assert(op->floats_per_rect <= rem); |
if (want > 1 && want * op->floats_per_rect > rem) |
want = rem / op->floats_per_rect; |
sna->render.vertex_index += 3*want; |
return want; |
flush: |
if (sna->render.vertex_offset) { |
gen4_vertex_flush(sna); |
gen5_magic_ca_pass(sna, op); |
} |
sna_vertex_wait__locked(&sna->render); |
_kgem_submit(&sna->kgem); |
emit_state(sna, op); |
goto start; |
} |
static uint32_t * |
gen5_composite_get_binding_table(struct sna *sna, |
uint16_t *offset) |
{ |
sna->kgem.surface -= |
sizeof(struct gen5_surface_state_padded) / sizeof(uint32_t); |
DBG(("%s(%x)\n", __FUNCTION__, 4*sna->kgem.surface)); |
/* Clear all surplus entries to zero in case of prefetch */ |
*offset = sna->kgem.surface; |
return memset(sna->kgem.batch + sna->kgem.surface, |
0, sizeof(struct gen5_surface_state_padded)); |
} |
static void |
gen5_emit_urb(struct sna *sna) |
{ |
int urb_vs_start, urb_vs_size; |
int urb_gs_start, urb_gs_size; |
int urb_clip_start, urb_clip_size; |
int urb_sf_start, urb_sf_size; |
int urb_cs_start, urb_cs_size; |
urb_vs_start = 0; |
urb_vs_size = URB_VS_ENTRIES * URB_VS_ENTRY_SIZE; |
urb_gs_start = urb_vs_start + urb_vs_size; |
urb_gs_size = URB_GS_ENTRIES * URB_GS_ENTRY_SIZE; |
urb_clip_start = urb_gs_start + urb_gs_size; |
urb_clip_size = URB_CLIP_ENTRIES * URB_CLIP_ENTRY_SIZE; |
urb_sf_start = urb_clip_start + urb_clip_size; |
urb_sf_size = URB_SF_ENTRIES * URB_SF_ENTRY_SIZE; |
urb_cs_start = urb_sf_start + urb_sf_size; |
urb_cs_size = URB_CS_ENTRIES * URB_CS_ENTRY_SIZE; |
OUT_BATCH(GEN5_URB_FENCE | |
UF0_CS_REALLOC | |
UF0_SF_REALLOC | |
UF0_CLIP_REALLOC | |
UF0_GS_REALLOC | |
UF0_VS_REALLOC | |
1); |
OUT_BATCH(((urb_clip_start + urb_clip_size) << UF1_CLIP_FENCE_SHIFT) | |
((urb_gs_start + urb_gs_size) << UF1_GS_FENCE_SHIFT) | |
((urb_vs_start + urb_vs_size) << UF1_VS_FENCE_SHIFT)); |
OUT_BATCH(((urb_cs_start + urb_cs_size) << UF2_CS_FENCE_SHIFT) | |
((urb_sf_start + urb_sf_size) << UF2_SF_FENCE_SHIFT)); |
/* Constant buffer state */ |
OUT_BATCH(GEN5_CS_URB_STATE | 0); |
OUT_BATCH((URB_CS_ENTRY_SIZE - 1) << 4 | URB_CS_ENTRIES << 0); |
} |
static void |
gen5_emit_state_base_address(struct sna *sna) |
{ |
assert(sna->render_state.gen5.general_bo->proxy == NULL); |
OUT_BATCH(GEN5_STATE_BASE_ADDRESS | 6); |
OUT_BATCH(kgem_add_reloc(&sna->kgem, /* general */ |
sna->kgem.nbatch, |
sna->render_state.gen5.general_bo, |
I915_GEM_DOMAIN_INSTRUCTION << 16, |
BASE_ADDRESS_MODIFY)); |
OUT_BATCH(kgem_add_reloc(&sna->kgem, /* surface */ |
sna->kgem.nbatch, |
NULL, |
I915_GEM_DOMAIN_INSTRUCTION << 16, |
BASE_ADDRESS_MODIFY)); |
OUT_BATCH(0); /* media */ |
OUT_BATCH(kgem_add_reloc(&sna->kgem, /* instruction */ |
sna->kgem.nbatch, |
sna->render_state.gen5.general_bo, |
I915_GEM_DOMAIN_INSTRUCTION << 16, |
BASE_ADDRESS_MODIFY)); |
/* upper bounds, all disabled */ |
OUT_BATCH(BASE_ADDRESS_MODIFY); |
OUT_BATCH(0); |
OUT_BATCH(BASE_ADDRESS_MODIFY); |
} |
static void |
gen5_emit_invariant(struct sna *sna) |
{ |
/* Ironlake errata workaround: Before disabling the clipper, |
* you have to MI_FLUSH to get the pipeline idle. |
* |
* However, the kernel flushes the pipeline between batches, |
* so we should be safe.... |
* |
* On the other hand, after using BLT we must use a non-pipelined |
* operation... |
*/ |
if (sna->kgem.nreloc) |
OUT_BATCH(MI_FLUSH | MI_INHIBIT_RENDER_CACHE_FLUSH); |
OUT_BATCH(GEN5_PIPELINE_SELECT | PIPELINE_SELECT_3D); |
gen5_emit_state_base_address(sna); |
sna->render_state.gen5.needs_invariant = false; |
} |
static void |
gen5_get_batch(struct sna *sna, const struct sna_composite_op *op) |
{ |
kgem_set_mode(&sna->kgem, KGEM_RENDER, op->dst.bo); |
if (!kgem_check_batch_with_surfaces(&sna->kgem, 150, 4)) { |
DBG(("%s: flushing batch: %d < %d+%d\n", |
__FUNCTION__, sna->kgem.surface - sna->kgem.nbatch, |
150, 4*8)); |
kgem_submit(&sna->kgem); |
_kgem_set_mode(&sna->kgem, KGEM_RENDER); |
} |
if (sna->render_state.gen5.needs_invariant) |
gen5_emit_invariant(sna); |
} |
static void |
gen5_align_vertex(struct sna *sna, const struct sna_composite_op *op) |
{ |
assert(op->floats_per_rect == 3*op->floats_per_vertex); |
if (op->floats_per_vertex != sna->render_state.gen5.floats_per_vertex) { |
if (sna->render.vertex_size - sna->render.vertex_used < 2*op->floats_per_rect) |
gen4_vertex_finish(sna); |
DBG(("aligning vertex: was %d, now %d floats per vertex, %d->%d\n", |
sna->render_state.gen5.floats_per_vertex, |
op->floats_per_vertex, |
sna->render.vertex_index, |
(sna->render.vertex_used + op->floats_per_vertex - 1) / op->floats_per_vertex)); |
sna->render.vertex_index = (sna->render.vertex_used + op->floats_per_vertex - 1) / op->floats_per_vertex; |
sna->render.vertex_used = sna->render.vertex_index * op->floats_per_vertex; |
sna->render_state.gen5.floats_per_vertex = op->floats_per_vertex; |
} |
} |
static void |
gen5_emit_binding_table(struct sna *sna, uint16_t offset) |
{ |
if (!DBG_NO_STATE_CACHE && |
sna->render_state.gen5.surface_table == offset) |
return; |
sna->render_state.gen5.surface_table = offset; |
/* Binding table pointers */ |
OUT_BATCH(GEN5_3DSTATE_BINDING_TABLE_POINTERS | 4); |
OUT_BATCH(0); /* vs */ |
OUT_BATCH(0); /* gs */ |
OUT_BATCH(0); /* clip */ |
OUT_BATCH(0); /* sf */ |
/* Only the PS uses the binding table */ |
OUT_BATCH(offset*4); |
} |
static bool |
gen5_emit_pipelined_pointers(struct sna *sna, |
const struct sna_composite_op *op, |
int blend, int kernel) |
{ |
uint16_t sp, bp; |
uint32_t key; |
DBG(("%s: has_mask=%d, src=(%d, %d), mask=(%d, %d),kernel=%d, blend=%d, ca=%d, format=%x\n", |
__FUNCTION__, op->u.gen5.ve_id & 2, |
op->src.filter, op->src.repeat, |
op->mask.filter, op->mask.repeat, |
kernel, blend, op->has_component_alpha, (int)op->dst.format)); |
sp = SAMPLER_OFFSET(op->src.filter, op->src.repeat, |
op->mask.filter, op->mask.repeat, |
kernel); |
bp = gen5_get_blend(blend, op->has_component_alpha, op->dst.format); |
key = sp | (uint32_t)bp << 16 | (op->mask.bo != NULL) << 31; |
DBG(("%s: sp=%d, bp=%d, key=%08x (current sp=%d, bp=%d, key=%08x)\n", |
__FUNCTION__, sp, bp, key, |
sna->render_state.gen5.last_pipelined_pointers & 0xffff, |
(sna->render_state.gen5.last_pipelined_pointers >> 16) & 0x7fff, |
sna->render_state.gen5.last_pipelined_pointers)); |
if (key == sna->render_state.gen5.last_pipelined_pointers) |
return false; |
OUT_BATCH(GEN5_3DSTATE_PIPELINED_POINTERS | 5); |
OUT_BATCH(sna->render_state.gen5.vs); |
OUT_BATCH(GEN5_GS_DISABLE); /* passthrough */ |
OUT_BATCH(GEN5_CLIP_DISABLE); /* passthrough */ |
OUT_BATCH(sna->render_state.gen5.sf[op->mask.bo != NULL]); |
OUT_BATCH(sna->render_state.gen5.wm + sp); |
OUT_BATCH(sna->render_state.gen5.cc + bp); |
bp = (sna->render_state.gen5.last_pipelined_pointers & 0x7fff0000) != ((uint32_t)bp << 16); |
sna->render_state.gen5.last_pipelined_pointers = key; |
gen5_emit_urb(sna); |
return bp; |
} |
static bool |
gen5_emit_drawing_rectangle(struct sna *sna, const struct sna_composite_op *op) |
{ |
uint32_t limit = (op->dst.height - 1) << 16 | (op->dst.width - 1); |
uint32_t offset = (uint16_t)op->dst.y << 16 | (uint16_t)op->dst.x; |
assert(!too_large(op->dst.x, op->dst.y)); |
assert(!too_large(op->dst.width, op->dst.height)); |
if (!DBG_NO_STATE_CACHE && |
sna->render_state.gen5.drawrect_limit == limit && |
sna->render_state.gen5.drawrect_offset == offset) |
return false; |
sna->render_state.gen5.drawrect_offset = offset; |
sna->render_state.gen5.drawrect_limit = limit; |
OUT_BATCH(GEN5_3DSTATE_DRAWING_RECTANGLE | (4 - 2)); |
OUT_BATCH(0x00000000); |
OUT_BATCH(limit); |
OUT_BATCH(offset); |
return true; |
} |
static void |
gen5_emit_vertex_elements(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
/* |
* vertex data in vertex buffer |
* position: (x, y) |
* texture coordinate 0: (u0, v0) if (is_affine is true) else (u0, v0, w0) |
* texture coordinate 1 if (has_mask is true): same as above |
*/ |
struct gen5_render_state *render = &sna->render_state.gen5; |
int id = op->u.gen5.ve_id; |
bool has_mask = id >> 2; |
uint32_t format, dw; |
if (!DBG_NO_STATE_CACHE && render->ve_id == id) |
return; |
DBG(("%s: changing %d -> %d\n", __FUNCTION__, render->ve_id, id)); |
render->ve_id = id; |
/* The VUE layout |
* dword 0-3: pad (0.0, 0.0, 0.0. 0.0) |
* dword 4-7: position (x, y, 1.0, 1.0), |
* dword 8-11: texture coordinate 0 (u0, v0, w0, 1.0) |
* dword 12-15: texture coordinate 1 (u1, v1, w1, 1.0) |
* |
* dword 4-15 are fetched from vertex buffer |
*/ |
OUT_BATCH(GEN5_3DSTATE_VERTEX_ELEMENTS | |
((2 * (has_mask ? 4 : 3)) + 1 - 2)); |
OUT_BATCH((id << VE0_VERTEX_BUFFER_INDEX_SHIFT) | VE0_VALID | |
(GEN5_SURFACEFORMAT_R32G32B32A32_FLOAT << VE0_FORMAT_SHIFT) | |
(0 << VE0_OFFSET_SHIFT)); |
OUT_BATCH((VFCOMPONENT_STORE_0 << VE1_VFCOMPONENT_0_SHIFT) | |
(VFCOMPONENT_STORE_0 << VE1_VFCOMPONENT_1_SHIFT) | |
(VFCOMPONENT_STORE_0 << VE1_VFCOMPONENT_2_SHIFT) | |
(VFCOMPONENT_STORE_0 << VE1_VFCOMPONENT_3_SHIFT)); |
/* x,y */ |
OUT_BATCH(id << VE0_VERTEX_BUFFER_INDEX_SHIFT | VE0_VALID | |
GEN5_SURFACEFORMAT_R16G16_SSCALED << VE0_FORMAT_SHIFT | |
0 << VE0_OFFSET_SHIFT); |
OUT_BATCH(VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_0_SHIFT | |
VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_1_SHIFT | |
VFCOMPONENT_STORE_1_FLT << VE1_VFCOMPONENT_2_SHIFT | |
VFCOMPONENT_STORE_1_FLT << VE1_VFCOMPONENT_3_SHIFT); |
/* u0, v0, w0 */ |
DBG(("%s: id=%d, first channel %d floats, offset=4b\n", __FUNCTION__, |
id, id & 3)); |
dw = VFCOMPONENT_STORE_1_FLT << VE1_VFCOMPONENT_3_SHIFT; |
switch (id & 3) { |
default: |
assert(0); |
case 0: |
format = GEN5_SURFACEFORMAT_R16G16_SSCALED << VE0_FORMAT_SHIFT; |
dw |= VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_0_SHIFT; |
dw |= VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_1_SHIFT; |
dw |= VFCOMPONENT_STORE_1_FLT << VE1_VFCOMPONENT_2_SHIFT; |
break; |
case 1: |
format = GEN5_SURFACEFORMAT_R32_FLOAT << VE0_FORMAT_SHIFT; |
dw |= VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_0_SHIFT; |
dw |= VFCOMPONENT_STORE_0 << VE1_VFCOMPONENT_1_SHIFT; |
dw |= VFCOMPONENT_STORE_1_FLT << VE1_VFCOMPONENT_2_SHIFT; |
break; |
case 2: |
format = GEN5_SURFACEFORMAT_R32G32_FLOAT << VE0_FORMAT_SHIFT; |
dw |= VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_0_SHIFT; |
dw |= VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_1_SHIFT; |
dw |= VFCOMPONENT_STORE_1_FLT << VE1_VFCOMPONENT_2_SHIFT; |
break; |
case 3: |
format = GEN5_SURFACEFORMAT_R32G32B32_FLOAT << VE0_FORMAT_SHIFT; |
dw |= VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_0_SHIFT; |
dw |= VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_1_SHIFT; |
dw |= VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_2_SHIFT; |
break; |
} |
OUT_BATCH(id << VE0_VERTEX_BUFFER_INDEX_SHIFT | VE0_VALID | |
format | 4 << VE0_OFFSET_SHIFT); |
OUT_BATCH(dw); |
/* u1, v1, w1 */ |
if (has_mask) { |
unsigned offset = 4 + ((id & 3) ?: 1) * sizeof(float); |
DBG(("%s: id=%x, second channel %d floats, offset=%db\n", __FUNCTION__, |
id, id >> 2, offset)); |
dw = VFCOMPONENT_STORE_1_FLT << VE1_VFCOMPONENT_3_SHIFT; |
switch (id >> 2) { |
case 1: |
format = GEN5_SURFACEFORMAT_R32_FLOAT << VE0_FORMAT_SHIFT; |
dw |= VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_0_SHIFT; |
dw |= VFCOMPONENT_STORE_0 << VE1_VFCOMPONENT_1_SHIFT; |
dw |= VFCOMPONENT_STORE_1_FLT << VE1_VFCOMPONENT_2_SHIFT; |
break; |
default: |
assert(0); |
case 2: |
format = GEN5_SURFACEFORMAT_R32G32_FLOAT << VE0_FORMAT_SHIFT; |
dw |= VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_0_SHIFT; |
dw |= VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_1_SHIFT; |
dw |= VFCOMPONENT_STORE_1_FLT << VE1_VFCOMPONENT_2_SHIFT; |
break; |
case 3: |
format = GEN5_SURFACEFORMAT_R32G32B32_FLOAT << VE0_FORMAT_SHIFT; |
dw |= VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_0_SHIFT; |
dw |= VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_1_SHIFT; |
dw |= VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_2_SHIFT; |
break; |
} |
OUT_BATCH(id << VE0_VERTEX_BUFFER_INDEX_SHIFT | VE0_VALID | |
format | offset << VE0_OFFSET_SHIFT); |
OUT_BATCH(dw); |
} |
} |
inline static void |
gen5_emit_pipe_flush(struct sna *sna) |
{ |
OUT_BATCH(GEN5_PIPE_CONTROL | (4 - 2)); |
OUT_BATCH(GEN5_PIPE_CONTROL_WC_FLUSH); |
OUT_BATCH(0); |
OUT_BATCH(0); |
} |
static void |
gen5_emit_state(struct sna *sna, |
const struct sna_composite_op *op, |
uint16_t offset) |
{ |
bool flush = false; |
assert(op->dst.bo->exec); |
/* drawrect must be first for Ironlake BLT workaround */ |
if (gen5_emit_drawing_rectangle(sna, op)) |
offset &= ~1; |
gen5_emit_binding_table(sna, offset & ~1); |
if (gen5_emit_pipelined_pointers(sna, op, op->op, op->u.gen5.wm_kernel)){ |
DBG(("%s: changed blend state, flush required? %d\n", |
__FUNCTION__, (offset & 1) && op->op > PictOpSrc)); |
flush = (offset & 1) && op->op > PictOpSrc; |
} |
gen5_emit_vertex_elements(sna, op); |
if (kgem_bo_is_dirty(op->src.bo) || kgem_bo_is_dirty(op->mask.bo)) { |
DBG(("%s: flushing dirty (%d, %d)\n", __FUNCTION__, |
kgem_bo_is_dirty(op->src.bo), |
kgem_bo_is_dirty(op->mask.bo))); |
OUT_BATCH(MI_FLUSH); |
kgem_clear_dirty(&sna->kgem); |
kgem_bo_mark_dirty(op->dst.bo); |
flush = false; |
} |
if (flush) { |
DBG(("%s: forcing flush\n", __FUNCTION__)); |
gen5_emit_pipe_flush(sna); |
} |
} |
static void gen5_bind_surfaces(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
bool dirty = kgem_bo_is_dirty(op->dst.bo); |
uint32_t *binding_table; |
uint16_t offset; |
gen5_get_batch(sna, op); |
binding_table = gen5_composite_get_binding_table(sna, &offset); |
binding_table[0] = |
gen5_bind_bo(sna, |
op->dst.bo, op->dst.width, op->dst.height, |
gen5_get_dest_format(op->dst.format), |
true); |
binding_table[1] = |
gen5_bind_bo(sna, |
op->src.bo, op->src.width, op->src.height, |
op->src.card_format, |
false); |
if (op->mask.bo) { |
assert(op->u.gen5.ve_id >> 2); |
binding_table[2] = |
gen5_bind_bo(sna, |
op->mask.bo, |
op->mask.width, |
op->mask.height, |
op->mask.card_format, |
false); |
} |
if (sna->kgem.surface == offset && |
*(uint64_t *)(sna->kgem.batch + sna->render_state.gen5.surface_table) == *(uint64_t*)binding_table && |
(op->mask.bo == NULL || |
sna->kgem.batch[sna->render_state.gen5.surface_table+2] == binding_table[2])) { |
sna->kgem.surface += sizeof(struct gen5_surface_state_padded) / sizeof(uint32_t); |
offset = sna->render_state.gen5.surface_table; |
} |
gen5_emit_state(sna, op, offset | dirty); |
} |
fastcall static void |
gen5_render_composite_blt(struct sna *sna, |
const struct sna_composite_op *op, |
const struct sna_composite_rectangles *r) |
{ |
DBG(("%s: src=(%d, %d)+(%d, %d), mask=(%d, %d)+(%d, %d), dst=(%d, %d)+(%d, %d), size=(%d, %d)\n", |
__FUNCTION__, |
r->src.x, r->src.y, op->src.offset[0], op->src.offset[1], |
r->mask.x, r->mask.y, op->mask.offset[0], op->mask.offset[1], |
r->dst.x, r->dst.y, op->dst.x, op->dst.y, |
r->width, r->height)); |
gen5_get_rectangles(sna, op, 1, gen5_bind_surfaces); |
op->prim_emit(sna, op, r); |
} |
#if 0 |
fastcall static void |
gen5_render_composite_box(struct sna *sna, |
const struct sna_composite_op *op, |
const BoxRec *box) |
{ |
struct sna_composite_rectangles r; |
DBG((" %s: (%d, %d), (%d, %d)\n", |
__FUNCTION__, |
box->x1, box->y1, box->x2, box->y2)); |
gen5_get_rectangles(sna, op, 1, gen5_bind_surfaces); |
r.dst.x = box->x1; |
r.dst.y = box->y1; |
r.width = box->x2 - box->x1; |
r.height = box->y2 - box->y1; |
r.mask = r.src = r.dst; |
op->prim_emit(sna, op, &r); |
} |
static void |
gen5_render_composite_boxes__blt(struct sna *sna, |
const struct sna_composite_op *op, |
const BoxRec *box, int nbox) |
{ |
DBG(("%s(%d) delta=(%d, %d), src=(%d, %d)/(%d, %d), mask=(%d, %d)/(%d, %d)\n", |
__FUNCTION__, nbox, op->dst.x, op->dst.y, |
op->src.offset[0], op->src.offset[1], |
op->src.width, op->src.height, |
op->mask.offset[0], op->mask.offset[1], |
op->mask.width, op->mask.height)); |
do { |
int nbox_this_time; |
nbox_this_time = gen5_get_rectangles(sna, op, nbox, |
gen5_bind_surfaces); |
nbox -= nbox_this_time; |
do { |
struct sna_composite_rectangles r; |
DBG((" %s: (%d, %d), (%d, %d)\n", |
__FUNCTION__, |
box->x1, box->y1, box->x2, box->y2)); |
r.dst.x = box->x1; |
r.dst.y = box->y1; |
r.width = box->x2 - box->x1; |
r.height = box->y2 - box->y1; |
r.mask = r.src = r.dst; |
op->prim_emit(sna, op, &r); |
box++; |
} while (--nbox_this_time); |
} while (nbox); |
} |
static void |
gen5_render_composite_boxes(struct sna *sna, |
const struct sna_composite_op *op, |
const BoxRec *box, int nbox) |
{ |
DBG(("%s: nbox=%d\n", __FUNCTION__, nbox)); |
do { |
int nbox_this_time; |
float *v; |
nbox_this_time = gen5_get_rectangles(sna, op, nbox, |
gen5_bind_surfaces); |
assert(nbox_this_time); |
nbox -= nbox_this_time; |
v = sna->render.vertices + sna->render.vertex_used; |
sna->render.vertex_used += nbox_this_time * op->floats_per_rect; |
op->emit_boxes(op, box, nbox_this_time, v); |
box += nbox_this_time; |
} while (nbox); |
} |
static void |
gen5_render_composite_boxes__thread(struct sna *sna, |
const struct sna_composite_op *op, |
const BoxRec *box, int nbox) |
{ |
DBG(("%s: nbox=%d\n", __FUNCTION__, nbox)); |
sna_vertex_lock(&sna->render); |
do { |
int nbox_this_time; |
float *v; |
nbox_this_time = gen5_get_rectangles(sna, op, nbox, |
gen5_bind_surfaces); |
assert(nbox_this_time); |
nbox -= nbox_this_time; |
v = sna->render.vertices + sna->render.vertex_used; |
sna->render.vertex_used += nbox_this_time * op->floats_per_rect; |
sna_vertex_acquire__locked(&sna->render); |
sna_vertex_unlock(&sna->render); |
op->emit_boxes(op, box, nbox_this_time, v); |
box += nbox_this_time; |
sna_vertex_lock(&sna->render); |
sna_vertex_release__locked(&sna->render); |
} while (nbox); |
sna_vertex_unlock(&sna->render); |
} |
#ifndef MAX |
#define MAX(a,b) ((a) > (b) ? (a) : (b)) |
#endif |
static uint32_t gen5_bind_video_source(struct sna *sna, |
struct kgem_bo *src_bo, |
uint32_t src_offset, |
int src_width, |
int src_height, |
int src_pitch, |
uint32_t src_surf_format) |
{ |
struct gen5_surface_state *ss; |
sna->kgem.surface -= sizeof(struct gen5_surface_state_padded) / sizeof(uint32_t); |
ss = memset(sna->kgem.batch + sna->kgem.surface, 0, sizeof(*ss)); |
ss->ss0.surface_type = GEN5_SURFACE_2D; |
ss->ss0.surface_format = src_surf_format; |
ss->ss0.color_blend = 1; |
ss->ss1.base_addr = |
kgem_add_reloc(&sna->kgem, |
sna->kgem.surface + 1, |
src_bo, |
I915_GEM_DOMAIN_SAMPLER << 16, |
src_offset); |
ss->ss2.width = src_width - 1; |
ss->ss2.height = src_height - 1; |
ss->ss3.pitch = src_pitch - 1; |
return sna->kgem.surface * sizeof(uint32_t); |
} |
static void gen5_video_bind_surfaces(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
bool dirty = kgem_bo_is_dirty(op->dst.bo); |
struct sna_video_frame *frame = op->priv; |
uint32_t src_surf_format; |
uint32_t src_surf_base[6]; |
int src_width[6]; |
int src_height[6]; |
int src_pitch[6]; |
uint32_t *binding_table; |
uint16_t offset; |
int n_src, n; |
src_surf_base[0] = 0; |
src_surf_base[1] = 0; |
src_surf_base[2] = frame->VBufOffset; |
src_surf_base[3] = frame->VBufOffset; |
src_surf_base[4] = frame->UBufOffset; |
src_surf_base[5] = frame->UBufOffset; |
if (is_planar_fourcc(frame->id)) { |
src_surf_format = GEN5_SURFACEFORMAT_R8_UNORM; |
src_width[1] = src_width[0] = frame->width; |
src_height[1] = src_height[0] = frame->height; |
src_pitch[1] = src_pitch[0] = frame->pitch[1]; |
src_width[4] = src_width[5] = src_width[2] = src_width[3] = |
frame->width / 2; |
src_height[4] = src_height[5] = src_height[2] = src_height[3] = |
frame->height / 2; |
src_pitch[4] = src_pitch[5] = src_pitch[2] = src_pitch[3] = |
frame->pitch[0]; |
n_src = 6; |
} else { |
if (frame->id == FOURCC_UYVY) |
src_surf_format = GEN5_SURFACEFORMAT_YCRCB_SWAPY; |
else |
src_surf_format = GEN5_SURFACEFORMAT_YCRCB_NORMAL; |
src_width[0] = frame->width; |
src_height[0] = frame->height; |
src_pitch[0] = frame->pitch[0]; |
n_src = 1; |
} |
gen5_get_batch(sna, op); |
binding_table = gen5_composite_get_binding_table(sna, &offset); |
binding_table[0] = |
gen5_bind_bo(sna, |
op->dst.bo, op->dst.width, op->dst.height, |
gen5_get_dest_format(op->dst.format), |
true); |
for (n = 0; n < n_src; n++) { |
binding_table[1+n] = |
gen5_bind_video_source(sna, |
frame->bo, |
src_surf_base[n], |
src_width[n], |
src_height[n], |
src_pitch[n], |
src_surf_format); |
} |
gen5_emit_state(sna, op, offset | dirty); |
} |
static bool |
gen5_render_video(struct sna *sna, |
struct sna_video *video, |
struct sna_video_frame *frame, |
RegionPtr dstRegion, |
PixmapPtr pixmap) |
{ |
struct sna_composite_op tmp; |
int dst_width = dstRegion->extents.x2 - dstRegion->extents.x1; |
int dst_height = dstRegion->extents.y2 - dstRegion->extents.y1; |
int src_width = frame->src.x2 - frame->src.x1; |
int src_height = frame->src.y2 - frame->src.y1; |
float src_offset_x, src_offset_y; |
float src_scale_x, src_scale_y; |
int nbox, pix_xoff, pix_yoff; |
struct sna_pixmap *priv; |
BoxPtr box; |
DBG(("%s: %dx%d -> %dx%d\n", __FUNCTION__, |
src_width, src_height, dst_width, dst_height)); |
priv = sna_pixmap_force_to_gpu(pixmap, MOVE_READ | MOVE_WRITE); |
if (priv == NULL) |
return false; |
memset(&tmp, 0, sizeof(tmp)); |
tmp.op = PictOpSrc; |
tmp.dst.pixmap = pixmap; |
tmp.dst.width = pixmap->drawable.width; |
tmp.dst.height = pixmap->drawable.height; |
tmp.dst.format = sna_format_for_depth(pixmap->drawable.depth); |
tmp.dst.bo = priv->gpu_bo; |
if (src_width == dst_width && src_height == dst_height) |
tmp.src.filter = SAMPLER_FILTER_NEAREST; |
else |
tmp.src.filter = SAMPLER_FILTER_BILINEAR; |
tmp.src.repeat = SAMPLER_EXTEND_PAD; |
tmp.src.bo = frame->bo; |
tmp.mask.bo = NULL; |
tmp.u.gen5.wm_kernel = |
is_planar_fourcc(frame->id) ? WM_KERNEL_VIDEO_PLANAR : WM_KERNEL_VIDEO_PACKED; |
tmp.u.gen5.ve_id = 2; |
tmp.is_affine = true; |
tmp.floats_per_vertex = 3; |
tmp.floats_per_rect = 9; |
tmp.priv = frame; |
if (!kgem_check_bo(&sna->kgem, tmp.dst.bo, frame->bo, NULL)) { |
kgem_submit(&sna->kgem); |
assert(kgem_check_bo(&sna->kgem, tmp.dst.bo, frame->bo, NULL)); |
} |
gen5_video_bind_surfaces(sna, &tmp); |
gen5_align_vertex(sna, &tmp); |
/* Set up the offset for translating from the given region (in screen |
* coordinates) to the backing pixmap. |
*/ |
#ifdef COMPOSITE |
pix_xoff = -pixmap->screen_x + pixmap->drawable.x; |
pix_yoff = -pixmap->screen_y + pixmap->drawable.y; |
#else |
pix_xoff = 0; |
pix_yoff = 0; |
#endif |
src_scale_x = (float)src_width / dst_width / frame->width; |
src_offset_x = (float)frame->src.x1 / frame->width - dstRegion->extents.x1 * src_scale_x; |
src_scale_y = (float)src_height / dst_height / frame->height; |
src_offset_y = (float)frame->src.y1 / frame->height - dstRegion->extents.y1 * src_scale_y; |
box = REGION_RECTS(dstRegion); |
nbox = REGION_NUM_RECTS(dstRegion); |
while (nbox--) { |
BoxRec r; |
r.x1 = box->x1 + pix_xoff; |
r.x2 = box->x2 + pix_xoff; |
r.y1 = box->y1 + pix_yoff; |
r.y2 = box->y2 + pix_yoff; |
gen5_get_rectangles(sna, &tmp, 1, gen5_video_bind_surfaces); |
OUT_VERTEX(r.x2, r.y2); |
OUT_VERTEX_F(box->x2 * src_scale_x + src_offset_x); |
OUT_VERTEX_F(box->y2 * src_scale_y + src_offset_y); |
OUT_VERTEX(r.x1, r.y2); |
OUT_VERTEX_F(box->x1 * src_scale_x + src_offset_x); |
OUT_VERTEX_F(box->y2 * src_scale_y + src_offset_y); |
OUT_VERTEX(r.x1, r.y1); |
OUT_VERTEX_F(box->x1 * src_scale_x + src_offset_x); |
OUT_VERTEX_F(box->y1 * src_scale_y + src_offset_y); |
if (!DAMAGE_IS_ALL(priv->gpu_damage)) { |
sna_damage_add_box(&priv->gpu_damage, &r); |
sna_damage_subtract_box(&priv->cpu_damage, &r); |
} |
box++; |
} |
gen4_vertex_flush(sna); |
return true; |
} |
#endif |
static void |
gen5_render_composite_done(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
if (sna->render.vertex_offset) { |
gen4_vertex_flush(sna); |
gen5_magic_ca_pass(sna,op); |
} |
DBG(("%s()\n", __FUNCTION__)); |
} |
#if 0 |
static bool |
gen5_composite_set_target(struct sna *sna, |
struct sna_composite_op *op, |
PicturePtr dst, |
int x, int y, int w, int h, |
bool partial) |
{ |
BoxRec box; |
op->dst.pixmap = get_drawable_pixmap(dst->pDrawable); |
op->dst.width = op->dst.pixmap->drawable.width; |
op->dst.height = op->dst.pixmap->drawable.height; |
op->dst.format = dst->format; |
if (w && h) { |
box.x1 = x; |
box.y1 = y; |
box.x2 = x + w; |
box.y2 = y + h; |
} else |
sna_render_picture_extents(dst, &box); |
op->dst.bo = sna_drawable_use_bo (dst->pDrawable, |
PREFER_GPU | FORCE_GPU | RENDER_GPU, |
&box, &op->damage); |
if (op->dst.bo == NULL) |
return false; |
get_drawable_deltas(dst->pDrawable, op->dst.pixmap, |
&op->dst.x, &op->dst.y); |
DBG(("%s: pixmap=%p, format=%08x, size=%dx%d, pitch=%d, delta=(%d,%d),damage=%p\n", |
__FUNCTION__, |
op->dst.pixmap, (int)op->dst.format, |
op->dst.width, op->dst.height, |
op->dst.bo->pitch, |
op->dst.x, op->dst.y, |
op->damage ? *op->damage : (void *)-1)); |
assert(op->dst.bo->proxy == NULL); |
if (too_large(op->dst.width, op->dst.height) && |
!sna_render_composite_redirect(sna, op, x, y, w, h, partial)) |
return false; |
return true; |
} |
static bool |
gen5_render_composite(struct sna *sna, |
uint8_t op, |
PicturePtr src, |
PicturePtr mask, |
PicturePtr dst, |
int16_t src_x, int16_t src_y, |
int16_t msk_x, int16_t msk_y, |
int16_t dst_x, int16_t dst_y, |
int16_t width, int16_t height, |
struct sna_composite_op *tmp) |
{ |
DBG(("%s: %dx%d, current mode=%d\n", __FUNCTION__, |
width, height, sna->kgem.mode)); |
if (op >= ARRAY_SIZE(gen5_blend_op)) { |
DBG(("%s: unhandled blend op %d\n", __FUNCTION__, op)); |
return false; |
} |
if (mask == NULL && |
try_blt(sna, dst, src, width, height) && |
sna_blt_composite(sna, op, |
src, dst, |
src_x, src_y, |
dst_x, dst_y, |
width, height, |
tmp, false)) |
return true; |
if (gen5_composite_fallback(sna, src, mask, dst)) |
return false; |
if (need_tiling(sna, width, height)) |
return sna_tiling_composite(op, src, mask, dst, |
src_x, src_y, |
msk_x, msk_y, |
dst_x, dst_y, |
width, height, |
tmp); |
if (!gen5_composite_set_target(sna, tmp, dst, |
dst_x, dst_y, width, height, |
op > PictOpSrc || dst->pCompositeClip->data)) { |
DBG(("%s: failed to set composite target\n", __FUNCTION__)); |
return false; |
} |
DBG(("%s: preparing source\n", __FUNCTION__)); |
tmp->op = op; |
switch (gen5_composite_picture(sna, src, &tmp->src, |
src_x, src_y, |
width, height, |
dst_x, dst_y, |
dst->polyMode == PolyModePrecise)) { |
case -1: |
DBG(("%s: failed to prepare source picture\n", __FUNCTION__)); |
goto cleanup_dst; |
case 0: |
if (!gen4_channel_init_solid(sna, &tmp->src, 0)) |
goto cleanup_dst; |
/* fall through to fixup */ |
case 1: |
if (mask == NULL && |
sna_blt_composite__convert(sna, |
dst_x, dst_y, width, height, |
tmp)) |
return true; |
gen5_composite_channel_convert(&tmp->src); |
break; |
} |
tmp->is_affine = tmp->src.is_affine; |
tmp->has_component_alpha = false; |
tmp->need_magic_ca_pass = false; |
if (mask) { |
if (mask->componentAlpha && PICT_FORMAT_RGB(mask->format)) { |
tmp->has_component_alpha = true; |
/* Check if it's component alpha that relies on a source alpha and on |
* the source value. We can only get one of those into the single |
* source value that we get to blend with. |
*/ |
if (gen5_blend_op[op].src_alpha && |
(gen5_blend_op[op].src_blend != GEN5_BLENDFACTOR_ZERO)) { |
if (op != PictOpOver) { |
DBG(("%s: unhandled CA blend op %d\n", __FUNCTION__, op)); |
goto cleanup_src; |
} |
tmp->need_magic_ca_pass = true; |
tmp->op = PictOpOutReverse; |
} |
} |
if (!reuse_source(sna, |
src, &tmp->src, src_x, src_y, |
mask, &tmp->mask, msk_x, msk_y)) { |
DBG(("%s: preparing mask\n", __FUNCTION__)); |
switch (gen5_composite_picture(sna, mask, &tmp->mask, |
msk_x, msk_y, |
width, height, |
dst_x, dst_y, |
dst->polyMode == PolyModePrecise)) { |
case -1: |
DBG(("%s: failed to prepare mask picture\n", __FUNCTION__)); |
goto cleanup_src; |
case 0: |
if (!gen4_channel_init_solid(sna, &tmp->mask, 0)) |
goto cleanup_src; |
/* fall through to fixup */ |
case 1: |
gen5_composite_channel_convert(&tmp->mask); |
break; |
} |
} |
tmp->is_affine &= tmp->mask.is_affine; |
} |
tmp->u.gen5.wm_kernel = |
gen5_choose_composite_kernel(tmp->op, |
tmp->mask.bo != NULL, |
tmp->has_component_alpha, |
tmp->is_affine); |
tmp->u.gen5.ve_id = gen4_choose_composite_emitter(sna, tmp); |
tmp->blt = gen5_render_composite_blt; |
tmp->box = gen5_render_composite_box; |
tmp->boxes = gen5_render_composite_boxes__blt; |
if (tmp->emit_boxes) { |
tmp->boxes = gen5_render_composite_boxes; |
tmp->thread_boxes = gen5_render_composite_boxes__thread; |
} |
tmp->done = gen5_render_composite_done; |
if (!kgem_check_bo(&sna->kgem, |
tmp->dst.bo, tmp->src.bo, tmp->mask.bo, NULL)) { |
kgem_submit(&sna->kgem); |
if (!kgem_check_bo(&sna->kgem, |
tmp->dst.bo, tmp->src.bo, tmp->mask.bo, NULL)) |
goto cleanup_mask; |
} |
gen5_bind_surfaces(sna, tmp); |
gen5_align_vertex(sna, tmp); |
return true; |
cleanup_mask: |
if (tmp->mask.bo) |
kgem_bo_destroy(&sna->kgem, tmp->mask.bo); |
cleanup_src: |
if (tmp->src.bo) |
kgem_bo_destroy(&sna->kgem, tmp->src.bo); |
cleanup_dst: |
if (tmp->redirect.real_bo) |
kgem_bo_destroy(&sna->kgem, tmp->dst.bo); |
return false; |
} |
#if !NO_COMPOSITE_SPANS |
fastcall static void |
gen5_render_composite_spans_box(struct sna *sna, |
const struct sna_composite_spans_op *op, |
const BoxRec *box, float opacity) |
{ |
DBG(("%s: src=+(%d, %d), opacity=%f, dst=+(%d, %d), box=(%d, %d) x (%d, %d)\n", |
__FUNCTION__, |
op->base.src.offset[0], op->base.src.offset[1], |
opacity, |
op->base.dst.x, op->base.dst.y, |
box->x1, box->y1, |
box->x2 - box->x1, |
box->y2 - box->y1)); |
gen5_get_rectangles(sna, &op->base, 1, gen5_bind_surfaces); |
op->prim_emit(sna, op, box, opacity); |
} |
static void |
gen5_render_composite_spans_boxes(struct sna *sna, |
const struct sna_composite_spans_op *op, |
const BoxRec *box, int nbox, |
float opacity) |
{ |
DBG(("%s: nbox=%d, src=+(%d, %d), opacity=%f, dst=+(%d, %d)\n", |
__FUNCTION__, nbox, |
op->base.src.offset[0], op->base.src.offset[1], |
opacity, |
op->base.dst.x, op->base.dst.y)); |
do { |
int nbox_this_time; |
nbox_this_time = gen5_get_rectangles(sna, &op->base, nbox, |
gen5_bind_surfaces); |
nbox -= nbox_this_time; |
do { |
DBG((" %s: (%d, %d) x (%d, %d)\n", __FUNCTION__, |
box->x1, box->y1, |
box->x2 - box->x1, |
box->y2 - box->y1)); |
op->prim_emit(sna, op, box++, opacity); |
} while (--nbox_this_time); |
} while (nbox); |
} |
fastcall static void |
gen5_render_composite_spans_boxes__thread(struct sna *sna, |
const struct sna_composite_spans_op *op, |
const struct sna_opacity_box *box, |
int nbox) |
{ |
DBG(("%s: nbox=%d, src=+(%d, %d), dst=+(%d, %d)\n", |
__FUNCTION__, nbox, |
op->base.src.offset[0], op->base.src.offset[1], |
op->base.dst.x, op->base.dst.y)); |
sna_vertex_lock(&sna->render); |
do { |
int nbox_this_time; |
float *v; |
nbox_this_time = gen5_get_rectangles(sna, &op->base, nbox, |
gen5_bind_surfaces); |
assert(nbox_this_time); |
nbox -= nbox_this_time; |
v = sna->render.vertices + sna->render.vertex_used; |
sna->render.vertex_used += nbox_this_time * op->base.floats_per_rect; |
sna_vertex_acquire__locked(&sna->render); |
sna_vertex_unlock(&sna->render); |
op->emit_boxes(op, box, nbox_this_time, v); |
box += nbox_this_time; |
sna_vertex_lock(&sna->render); |
sna_vertex_release__locked(&sna->render); |
} while (nbox); |
sna_vertex_unlock(&sna->render); |
} |
fastcall static void |
gen5_render_composite_spans_done(struct sna *sna, |
const struct sna_composite_spans_op *op) |
{ |
if (sna->render.vertex_offset) |
gen4_vertex_flush(sna); |
DBG(("%s()\n", __FUNCTION__)); |
kgem_bo_destroy(&sna->kgem, op->base.src.bo); |
sna_render_composite_redirect_done(sna, &op->base); |
} |
static bool |
gen5_check_composite_spans(struct sna *sna, |
uint8_t op, PicturePtr src, PicturePtr dst, |
int16_t width, int16_t height, |
unsigned flags) |
{ |
DBG(("%s: op=%d, width=%d, height=%d, flags=%x\n", |
__FUNCTION__, op, width, height, flags)); |
if (op >= ARRAY_SIZE(gen5_blend_op)) |
return false; |
if (gen5_composite_fallback(sna, src, NULL, dst)) { |
DBG(("%s: operation would fallback\n", __FUNCTION__)); |
return false; |
} |
if (need_tiling(sna, width, height) && |
!is_gpu(sna, dst->pDrawable, PREFER_GPU_SPANS)) { |
DBG(("%s: fallback, tiled operation not on GPU\n", |
__FUNCTION__)); |
return false; |
} |
if ((flags & COMPOSITE_SPANS_RECTILINEAR) == 0) { |
struct sna_pixmap *priv = sna_pixmap_from_drawable(dst->pDrawable); |
assert(priv); |
if (priv->cpu_bo && kgem_bo_is_busy(priv->cpu_bo)) |
return true; |
if (flags & COMPOSITE_SPANS_INPLACE_HINT) |
return false; |
if ((sna->render.prefer_gpu & PREFER_GPU_SPANS) == 0 && |
dst->format == PICT_a8) |
return false; |
return priv->gpu_bo && kgem_bo_is_busy(priv->gpu_bo); |
} |
return true; |
} |
static bool |
gen5_render_composite_spans(struct sna *sna, |
uint8_t op, |
PicturePtr src, |
PicturePtr dst, |
int16_t src_x, int16_t src_y, |
int16_t dst_x, int16_t dst_y, |
int16_t width, int16_t height, |
unsigned flags, |
struct sna_composite_spans_op *tmp) |
{ |
DBG(("%s: %dx%d with flags=%x, current mode=%d\n", __FUNCTION__, |
width, height, flags, sna->kgem.ring)); |
assert(gen5_check_composite_spans(sna, op, src, dst, width, height, flags)); |
if (need_tiling(sna, width, height)) { |
DBG(("%s: tiling, operation (%dx%d) too wide for pipeline\n", |
__FUNCTION__, width, height)); |
return sna_tiling_composite_spans(op, src, dst, |
src_x, src_y, dst_x, dst_y, |
width, height, flags, tmp); |
} |
tmp->base.op = op; |
if (!gen5_composite_set_target(sna, &tmp->base, dst, |
dst_x, dst_y, width, height, |
true)) |
return false; |
switch (gen5_composite_picture(sna, src, &tmp->base.src, |
src_x, src_y, |
width, height, |
dst_x, dst_y, |
dst->polyMode == PolyModePrecise)) { |
case -1: |
goto cleanup_dst; |
case 0: |
if (!gen4_channel_init_solid(sna, &tmp->base.src, 0)) |
goto cleanup_dst; |
/* fall through to fixup */ |
case 1: |
gen5_composite_channel_convert(&tmp->base.src); |
break; |
} |
tmp->base.mask.bo = NULL; |
tmp->base.is_affine = tmp->base.src.is_affine; |
tmp->base.has_component_alpha = false; |
tmp->base.need_magic_ca_pass = false; |
tmp->base.u.gen5.ve_id = gen4_choose_spans_emitter(sna, tmp); |
tmp->base.u.gen5.wm_kernel = WM_KERNEL_OPACITY | !tmp->base.is_affine; |
tmp->box = gen5_render_composite_spans_box; |
tmp->boxes = gen5_render_composite_spans_boxes; |
if (tmp->emit_boxes) |
tmp->thread_boxes = gen5_render_composite_spans_boxes__thread; |
tmp->done = gen5_render_composite_spans_done; |
if (!kgem_check_bo(&sna->kgem, |
tmp->base.dst.bo, tmp->base.src.bo, |
NULL)) { |
kgem_submit(&sna->kgem); |
if (!kgem_check_bo(&sna->kgem, |
tmp->base.dst.bo, tmp->base.src.bo, |
NULL)) |
goto cleanup_src; |
} |
gen5_bind_surfaces(sna, &tmp->base); |
gen5_align_vertex(sna, &tmp->base); |
return true; |
cleanup_src: |
if (tmp->base.src.bo) |
kgem_bo_destroy(&sna->kgem, tmp->base.src.bo); |
cleanup_dst: |
if (tmp->base.redirect.real_bo) |
kgem_bo_destroy(&sna->kgem, tmp->base.dst.bo); |
return false; |
} |
#endif |
static bool |
gen5_render_copy_boxes(struct sna *sna, uint8_t alu, |
PixmapPtr src, struct kgem_bo *src_bo, int16_t src_dx, int16_t src_dy, |
PixmapPtr dst, struct kgem_bo *dst_bo, int16_t dst_dx, int16_t dst_dy, |
const BoxRec *box, int n, unsigned flags) |
{ |
struct sna_composite_op tmp; |
DBG(("%s alu=%d, src=%ld:handle=%d, dst=%ld:handle=%d boxes=%d x [((%d, %d), (%d, %d))...], flags=%x\n", |
__FUNCTION__, alu, |
src->drawable.serialNumber, src_bo->handle, |
dst->drawable.serialNumber, dst_bo->handle, |
n, box->x1, box->y1, box->x2, box->y2, |
flags)); |
if (sna_blt_compare_depth(&src->drawable, &dst->drawable) && |
sna_blt_copy_boxes(sna, alu, |
src_bo, src_dx, src_dy, |
dst_bo, dst_dx, dst_dy, |
dst->drawable.bitsPerPixel, |
box, n)) |
return true; |
if (!(alu == GXcopy || alu == GXclear) || src_bo == dst_bo) { |
fallback_blt: |
if (!sna_blt_compare_depth(&src->drawable, &dst->drawable)) |
return false; |
return sna_blt_copy_boxes_fallback(sna, alu, |
src, src_bo, src_dx, src_dy, |
dst, dst_bo, dst_dx, dst_dy, |
box, n); |
} |
memset(&tmp, 0, sizeof(tmp)); |
if (dst->drawable.depth == src->drawable.depth) { |
tmp.dst.format = sna_render_format_for_depth(dst->drawable.depth); |
tmp.src.pict_format = tmp.dst.format; |
} else { |
tmp.dst.format = sna_format_for_depth(dst->drawable.depth); |
tmp.src.pict_format = sna_format_for_depth(src->drawable.depth); |
} |
if (!gen5_check_format(tmp.src.pict_format)) { |
DBG(("%s: unsupported source format, %x, use BLT\n", |
__FUNCTION__, tmp.src.pict_format)); |
goto fallback_blt; |
} |
DBG(("%s (%d, %d)->(%d, %d) x %d\n", |
__FUNCTION__, src_dx, src_dy, dst_dx, dst_dy, n)); |
tmp.op = alu == GXcopy ? PictOpSrc : PictOpClear; |
tmp.dst.pixmap = dst; |
tmp.dst.width = dst->drawable.width; |
tmp.dst.height = dst->drawable.height; |
tmp.dst.x = tmp.dst.y = 0; |
tmp.dst.bo = dst_bo; |
tmp.damage = NULL; |
sna_render_composite_redirect_init(&tmp); |
if (too_large(tmp.dst.width, tmp.dst.height)) { |
BoxRec extents = box[0]; |
int i; |
for (i = 1; i < n; i++) { |
if (box[i].x1 < extents.x1) |
extents.x1 = box[i].x1; |
if (box[i].y1 < extents.y1) |
extents.y1 = box[i].y1; |
if (box[i].x2 > extents.x2) |
extents.x2 = box[i].x2; |
if (box[i].y2 > extents.y2) |
extents.y2 = box[i].y2; |
} |
if (!sna_render_composite_redirect(sna, &tmp, |
extents.x1 + dst_dx, |
extents.y1 + dst_dy, |
extents.x2 - extents.x1, |
extents.y2 - extents.y1, |
n > 1)) |
goto fallback_tiled; |
} |
tmp.src.filter = SAMPLER_FILTER_NEAREST; |
tmp.src.repeat = SAMPLER_EXTEND_NONE; |
tmp.src.card_format = gen5_get_card_format(tmp.src.pict_format); |
if (too_large(src->drawable.width, src->drawable.height)) { |
BoxRec extents = box[0]; |
int i; |
for (i = 1; i < n; i++) { |
if (box[i].x1 < extents.x1) |
extents.x1 = box[i].x1; |
if (box[i].y1 < extents.y1) |
extents.y1 = box[i].y1; |
if (box[i].x2 > extents.x2) |
extents.x2 = box[i].x2; |
if (box[i].y2 > extents.y2) |
extents.y2 = box[i].y2; |
} |
if (!sna_render_pixmap_partial(sna, src, src_bo, &tmp.src, |
extents.x1 + src_dx, |
extents.y1 + src_dy, |
extents.x2 - extents.x1, |
extents.y2 - extents.y1)) |
goto fallback_tiled_dst; |
} else { |
tmp.src.bo = kgem_bo_reference(src_bo); |
tmp.src.width = src->drawable.width; |
tmp.src.height = src->drawable.height; |
tmp.src.offset[0] = tmp.src.offset[1] = 0; |
tmp.src.scale[0] = 1.f/src->drawable.width; |
tmp.src.scale[1] = 1.f/src->drawable.height; |
} |
tmp.is_affine = true; |
tmp.floats_per_vertex = 3; |
tmp.floats_per_rect = 9; |
tmp.u.gen5.wm_kernel = WM_KERNEL; |
tmp.u.gen5.ve_id = 2; |
if (!kgem_check_bo(&sna->kgem, dst_bo, src_bo, NULL)) { |
kgem_submit(&sna->kgem); |
if (!kgem_check_bo(&sna->kgem, dst_bo, src_bo, NULL)) { |
DBG(("%s: aperture check failed\n", __FUNCTION__)); |
goto fallback_tiled_src; |
} |
} |
dst_dx += tmp.dst.x; |
dst_dy += tmp.dst.y; |
tmp.dst.x = tmp.dst.y = 0; |
src_dx += tmp.src.offset[0]; |
src_dy += tmp.src.offset[1]; |
gen5_copy_bind_surfaces(sna, &tmp); |
gen5_align_vertex(sna, &tmp); |
do { |
int n_this_time; |
n_this_time = gen5_get_rectangles(sna, &tmp, n, |
gen5_copy_bind_surfaces); |
n -= n_this_time; |
do { |
DBG((" (%d, %d) -> (%d, %d) + (%d, %d)\n", |
box->x1 + src_dx, box->y1 + src_dy, |
box->x1 + dst_dx, box->y1 + dst_dy, |
box->x2 - box->x1, box->y2 - box->y1)); |
OUT_VERTEX(box->x2 + dst_dx, box->y2 + dst_dy); |
OUT_VERTEX_F((box->x2 + src_dx) * tmp.src.scale[0]); |
OUT_VERTEX_F((box->y2 + src_dy) * tmp.src.scale[1]); |
OUT_VERTEX(box->x1 + dst_dx, box->y2 + dst_dy); |
OUT_VERTEX_F((box->x1 + src_dx) * tmp.src.scale[0]); |
OUT_VERTEX_F((box->y2 + src_dy) * tmp.src.scale[1]); |
OUT_VERTEX(box->x1 + dst_dx, box->y1 + dst_dy); |
OUT_VERTEX_F((box->x1 + src_dx) * tmp.src.scale[0]); |
OUT_VERTEX_F((box->y1 + src_dy) * tmp.src.scale[1]); |
box++; |
} while (--n_this_time); |
} while (n); |
gen4_vertex_flush(sna); |
sna_render_composite_redirect_done(sna, &tmp); |
kgem_bo_destroy(&sna->kgem, tmp.src.bo); |
return true; |
fallback_tiled_src: |
kgem_bo_destroy(&sna->kgem, tmp.src.bo); |
fallback_tiled_dst: |
if (tmp.redirect.real_bo) |
kgem_bo_destroy(&sna->kgem, tmp.dst.bo); |
fallback_tiled: |
if (sna_blt_compare_depth(&src->drawable, &dst->drawable) && |
sna_blt_copy_boxes(sna, alu, |
src_bo, src_dx, src_dy, |
dst_bo, dst_dx, dst_dy, |
dst->drawable.bitsPerPixel, |
box, n)) |
return true; |
DBG(("%s: tiled fallback\n", __FUNCTION__)); |
return sna_tiling_copy_boxes(sna, alu, |
src, src_bo, src_dx, src_dy, |
dst, dst_bo, dst_dx, dst_dy, |
box, n); |
} |
#endif |
static void |
gen5_render_flush(struct sna *sna) |
{ |
gen4_vertex_close(sna); |
assert(sna->render.vb_id == 0); |
assert(sna->render.vertex_offset == 0); |
} |
static void |
gen5_render_context_switch(struct kgem *kgem, |
int new_mode) |
{ |
if (!kgem->nbatch) |
return; |
/* WaNonPipelinedStateCommandFlush |
* |
* Ironlake has a limitation that a 3D or Media command can't |
* be the first command after a BLT, unless it's |
* non-pipelined. |
* |
* We do this by ensuring that the non-pipelined drawrect |
* is always emitted first following a switch from BLT. |
*/ |
if (kgem->mode == KGEM_BLT) { |
struct sna *sna = to_sna_from_kgem(kgem); |
DBG(("%s: forcing drawrect on next state emission\n", |
__FUNCTION__)); |
sna->render_state.gen5.drawrect_limit = -1; |
} |
if (kgem_ring_is_idle(kgem, kgem->ring)) { |
DBG(("%s: GPU idle, flushing\n", __FUNCTION__)); |
_kgem_submit(kgem); |
} |
} |
static void |
discard_vbo(struct sna *sna) |
{ |
kgem_bo_destroy(&sna->kgem, sna->render.vbo); |
sna->render.vbo = NULL; |
sna->render.vertices = sna->render.vertex_data; |
sna->render.vertex_size = ARRAY_SIZE(sna->render.vertex_data); |
sna->render.vertex_used = 0; |
sna->render.vertex_index = 0; |
} |
static void |
gen5_render_retire(struct kgem *kgem) |
{ |
struct sna *sna; |
sna = container_of(kgem, struct sna, kgem); |
if (kgem->nbatch == 0 && sna->render.vbo && !kgem_bo_is_busy(sna->render.vbo)) { |
DBG(("%s: resetting idle vbo\n", __FUNCTION__)); |
sna->render.vertex_used = 0; |
sna->render.vertex_index = 0; |
} |
} |
static void |
gen5_render_expire(struct kgem *kgem) |
{ |
struct sna *sna; |
sna = container_of(kgem, struct sna, kgem); |
if (sna->render.vbo && !sna->render.vertex_used) { |
DBG(("%s: discarding vbo\n", __FUNCTION__)); |
discard_vbo(sna); |
} |
} |
static void gen5_render_reset(struct sna *sna) |
{ |
sna->render_state.gen5.needs_invariant = true; |
sna->render_state.gen5.ve_id = -1; |
sna->render_state.gen5.last_primitive = -1; |
sna->render_state.gen5.last_pipelined_pointers = 0; |
sna->render_state.gen5.drawrect_offset = -1; |
sna->render_state.gen5.drawrect_limit = -1; |
sna->render_state.gen5.surface_table = -1; |
if (sna->render.vbo && |
!kgem_bo_is_mappable(&sna->kgem, sna->render.vbo)) { |
DBG(("%s: discarding unmappable vbo\n", __FUNCTION__)); |
discard_vbo(sna); |
} |
sna->render.vertex_offset = 0; |
sna->render.nvertex_reloc = 0; |
sna->render.vb_id = 0; |
} |
static void gen5_render_fini(struct sna *sna) |
{ |
kgem_bo_destroy(&sna->kgem, sna->render_state.gen5.general_bo); |
} |
static uint32_t gen5_create_vs_unit_state(struct sna_static_stream *stream) |
{ |
struct gen5_vs_unit_state *vs = sna_static_stream_map(stream, sizeof(*vs), 32); |
/* Set up the vertex shader to be disabled (passthrough) */ |
vs->thread4.nr_urb_entries = URB_VS_ENTRIES >> 2; |
vs->thread4.urb_entry_allocation_size = URB_VS_ENTRY_SIZE - 1; |
vs->vs6.vs_enable = 0; |
vs->vs6.vert_cache_disable = 1; |
return sna_static_stream_offsetof(stream, vs); |
} |
static uint32_t gen5_create_sf_state(struct sna_static_stream *stream, |
uint32_t kernel) |
{ |
struct gen5_sf_unit_state *sf_state; |
sf_state = sna_static_stream_map(stream, sizeof(*sf_state), 32); |
sf_state->thread0.grf_reg_count = GEN5_GRF_BLOCKS(SF_KERNEL_NUM_GRF); |
sf_state->thread0.kernel_start_pointer = kernel >> 6; |
sf_state->thread3.const_urb_entry_read_length = 0; /* no const URBs */ |
sf_state->thread3.const_urb_entry_read_offset = 0; /* no const URBs */ |
sf_state->thread3.urb_entry_read_length = 1; /* 1 URB per vertex */ |
/* don't smash vertex header, read start from dw8 */ |
sf_state->thread3.urb_entry_read_offset = 1; |
sf_state->thread3.dispatch_grf_start_reg = 3; |
sf_state->thread4.max_threads = SF_MAX_THREADS - 1; |
sf_state->thread4.urb_entry_allocation_size = URB_SF_ENTRY_SIZE - 1; |
sf_state->thread4.nr_urb_entries = URB_SF_ENTRIES; |
sf_state->sf5.viewport_transform = false; /* skip viewport */ |
sf_state->sf6.cull_mode = GEN5_CULLMODE_NONE; |
sf_state->sf6.scissor = 0; |
sf_state->sf7.trifan_pv = 2; |
sf_state->sf6.dest_org_vbias = 0x8; |
sf_state->sf6.dest_org_hbias = 0x8; |
return sna_static_stream_offsetof(stream, sf_state); |
} |
static uint32_t gen5_create_sampler_state(struct sna_static_stream *stream, |
sampler_filter_t src_filter, |
sampler_extend_t src_extend, |
sampler_filter_t mask_filter, |
sampler_extend_t mask_extend) |
{ |
struct gen5_sampler_state *sampler_state; |
sampler_state = sna_static_stream_map(stream, |
sizeof(struct gen5_sampler_state) * 2, |
32); |
sampler_state_init(&sampler_state[0], src_filter, src_extend); |
sampler_state_init(&sampler_state[1], mask_filter, mask_extend); |
return sna_static_stream_offsetof(stream, sampler_state); |
} |
static void gen5_init_wm_state(struct gen5_wm_unit_state *state, |
bool has_mask, |
uint32_t kernel, |
uint32_t sampler) |
{ |
state->thread0.grf_reg_count = GEN5_GRF_BLOCKS(PS_KERNEL_NUM_GRF); |
state->thread0.kernel_start_pointer = kernel >> 6; |
state->thread1.single_program_flow = 0; |
/* scratch space is not used in our kernel */ |
state->thread2.scratch_space_base_pointer = 0; |
state->thread2.per_thread_scratch_space = 0; |
state->thread3.const_urb_entry_read_length = 0; |
state->thread3.const_urb_entry_read_offset = 0; |
state->thread3.urb_entry_read_offset = 0; |
/* wm kernel use urb from 3, see wm_program in compiler module */ |
state->thread3.dispatch_grf_start_reg = 3; /* must match kernel */ |
state->wm4.sampler_count = 0; /* hardware requirement */ |
state->wm4.sampler_state_pointer = sampler >> 5; |
state->wm5.max_threads = PS_MAX_THREADS - 1; |
state->wm5.transposed_urb_read = 0; |
state->wm5.thread_dispatch_enable = 1; |
/* just use 16-pixel dispatch (4 subspans), don't need to change kernel |
* start point |
*/ |
state->wm5.enable_16_pix = 1; |
state->wm5.enable_8_pix = 0; |
state->wm5.early_depth_test = 1; |
/* Each pair of attributes (src/mask coords) is two URB entries */ |
if (has_mask) { |
state->thread1.binding_table_entry_count = 3; /* 2 tex and fb */ |
state->thread3.urb_entry_read_length = 4; |
} else { |
state->thread1.binding_table_entry_count = 2; /* 1 tex and fb */ |
state->thread3.urb_entry_read_length = 2; |
} |
/* binding table entry count is only used for prefetching, |
* and it has to be set 0 for Ironlake |
*/ |
state->thread1.binding_table_entry_count = 0; |
} |
static uint32_t gen5_create_cc_unit_state(struct sna_static_stream *stream) |
{ |
uint8_t *ptr, *base; |
int i, j; |
base = ptr = |
sna_static_stream_map(stream, |
GEN5_BLENDFACTOR_COUNT*GEN5_BLENDFACTOR_COUNT*64, |
64); |
for (i = 0; i < GEN5_BLENDFACTOR_COUNT; i++) { |
for (j = 0; j < GEN5_BLENDFACTOR_COUNT; j++) { |
struct gen5_cc_unit_state *state = |
(struct gen5_cc_unit_state *)ptr; |
state->cc3.blend_enable = |
!(j == GEN5_BLENDFACTOR_ZERO && i == GEN5_BLENDFACTOR_ONE); |
state->cc5.logicop_func = 0xc; /* COPY */ |
state->cc5.ia_blend_function = GEN5_BLENDFUNCTION_ADD; |
/* Fill in alpha blend factors same as color, for the future. */ |
state->cc5.ia_src_blend_factor = i; |
state->cc5.ia_dest_blend_factor = j; |
state->cc6.blend_function = GEN5_BLENDFUNCTION_ADD; |
state->cc6.clamp_post_alpha_blend = 1; |
state->cc6.clamp_pre_alpha_blend = 1; |
state->cc6.src_blend_factor = i; |
state->cc6.dest_blend_factor = j; |
ptr += 64; |
} |
} |
return sna_static_stream_offsetof(stream, base); |
} |
static bool gen5_render_setup(struct sna *sna) |
{ |
struct gen5_render_state *state = &sna->render_state.gen5; |
struct sna_static_stream general; |
struct gen5_wm_unit_state_padded *wm_state; |
uint32_t sf[2], wm[KERNEL_COUNT]; |
int i, j, k, l, m; |
sna_static_stream_init(&general); |
/* Zero pad the start. If you see an offset of 0x0 in the batchbuffer |
* dumps, you know it points to zero. |
*/ |
null_create(&general); |
/* Set up the two SF states (one for blending with a mask, one without) */ |
sf[0] = sna_static_stream_compile_sf(sna, &general, brw_sf_kernel__nomask); |
sf[1] = sna_static_stream_compile_sf(sna, &general, brw_sf_kernel__mask); |
for (m = 0; m < KERNEL_COUNT; m++) { |
if (wm_kernels[m].size) { |
wm[m] = sna_static_stream_add(&general, |
wm_kernels[m].data, |
wm_kernels[m].size, |
64); |
} else { |
wm[m] = sna_static_stream_compile_wm(sna, &general, |
wm_kernels[m].data, |
16); |
} |
assert(wm[m]); |
} |
state->vs = gen5_create_vs_unit_state(&general); |
state->sf[0] = gen5_create_sf_state(&general, sf[0]); |
state->sf[1] = gen5_create_sf_state(&general, sf[1]); |
/* Set up the WM states: each filter/extend type for source and mask, per |
* kernel. |
*/ |
wm_state = sna_static_stream_map(&general, |
sizeof(*wm_state) * KERNEL_COUNT * |
FILTER_COUNT * EXTEND_COUNT * |
FILTER_COUNT * EXTEND_COUNT, |
64); |
state->wm = sna_static_stream_offsetof(&general, wm_state); |
for (i = 0; i < FILTER_COUNT; i++) { |
for (j = 0; j < EXTEND_COUNT; j++) { |
for (k = 0; k < FILTER_COUNT; k++) { |
for (l = 0; l < EXTEND_COUNT; l++) { |
uint32_t sampler_state; |
sampler_state = |
gen5_create_sampler_state(&general, |
i, j, |
k, l); |
for (m = 0; m < KERNEL_COUNT; m++) { |
gen5_init_wm_state(&wm_state->state, |
wm_kernels[m].has_mask, |
wm[m], sampler_state); |
wm_state++; |
} |
} |
} |
} |
} |
state->cc = gen5_create_cc_unit_state(&general); |
state->general_bo = sna_static_stream_fini(sna, &general); |
return state->general_bo != NULL; |
} |
const char *gen5_render_init(struct sna *sna, const char *backend) |
{ |
if (!gen5_render_setup(sna)) |
return backend; |
sna->kgem.context_switch = gen5_render_context_switch; |
sna->kgem.retire = gen5_render_retire; |
sna->kgem.expire = gen5_render_expire; |
#if 0 |
#if !NO_COMPOSITE |
sna->render.composite = gen5_render_composite; |
sna->render.prefer_gpu |= PREFER_GPU_RENDER; |
#endif |
#if !NO_COMPOSITE_SPANS |
sna->render.check_composite_spans = gen5_check_composite_spans; |
sna->render.composite_spans = gen5_render_composite_spans; |
if (sna->PciInfo->device_id == 0x0044) |
sna->render.prefer_gpu |= PREFER_GPU_SPANS; |
#endif |
sna->render.video = gen5_render_video; |
sna->render.copy_boxes = gen5_render_copy_boxes; |
sna->render.copy = gen5_render_copy; |
sna->render.fill_boxes = gen5_render_fill_boxes; |
sna->render.fill = gen5_render_fill; |
sna->render.fill_one = gen5_render_fill_one; |
#endif |
sna->render.blit_tex = gen5_blit_tex; |
sna->render.caps = HW_BIT_BLIT | HW_TEX_BLIT; |
sna->render.flush = gen5_render_flush; |
sna->render.reset = gen5_render_reset; |
sna->render.fini = gen5_render_fini; |
sna->render.max_3d_size = MAX_3D_SIZE; |
sna->render.max_3d_pitch = 1 << 18; |
return "Ironlake (gen5)"; |
}; |
static bool |
gen5_blit_tex(struct sna *sna, |
uint8_t op, bool scale, |
PixmapPtr src, struct kgem_bo *src_bo, |
PixmapPtr mask,struct kgem_bo *mask_bo, |
PixmapPtr dst, struct kgem_bo *dst_bo, |
int32_t src_x, int32_t src_y, |
int32_t msk_x, int32_t msk_y, |
int32_t dst_x, int32_t dst_y, |
int32_t width, int32_t height, |
struct sna_composite_op *tmp) |
{ |
DBG(("%s: %dx%d, current mode=%d\n", __FUNCTION__, |
width, height, sna->kgem.mode)); |
tmp->op = PictOpSrc; |
tmp->dst.pixmap = dst; |
tmp->dst.bo = dst_bo; |
tmp->dst.width = dst->drawable.width; |
tmp->dst.height = dst->drawable.height; |
tmp->dst.format = PICT_x8r8g8b8; |
tmp->src.repeat = RepeatNone; |
tmp->src.filter = PictFilterNearest; |
tmp->src.is_affine = true; |
tmp->src.bo = src_bo; |
tmp->src.pict_format = PICT_x8r8g8b8; |
tmp->src.card_format = gen5_get_card_format(tmp->src.pict_format); |
tmp->src.width = src->drawable.width; |
tmp->src.height = src->drawable.height; |
tmp->is_affine = tmp->src.is_affine; |
tmp->has_component_alpha = false; |
tmp->need_magic_ca_pass = false; |
tmp->mask.is_affine = true; |
tmp->mask.repeat = SAMPLER_EXTEND_NONE; |
tmp->mask.filter = SAMPLER_FILTER_NEAREST; |
tmp->mask.bo = mask_bo; |
tmp->mask.pict_format = PIXMAN_a8; |
tmp->mask.card_format = gen5_get_card_format(tmp->mask.pict_format); |
tmp->mask.width = mask->drawable.width; |
tmp->mask.height = mask->drawable.height; |
if( scale ) |
{ |
tmp->src.scale[0] = 1.f/width; |
tmp->src.scale[1] = 1.f/height; |
} |
else |
{ |
tmp->src.scale[0] = 1.f/src->drawable.width; |
tmp->src.scale[1] = 1.f/src->drawable.height; |
} |
tmp->mask.scale[0] = 1.f/mask->drawable.width; |
tmp->mask.scale[1] = 1.f/mask->drawable.height; |
tmp->u.gen5.wm_kernel = WM_KERNEL_MASK; |
// gen5_choose_composite_kernel(tmp->op, |
// tmp->mask.bo != NULL, |
// tmp->has_component_alpha, |
// tmp->is_affine); |
tmp->u.gen5.ve_id = gen4_choose_composite_emitter(sna, tmp); |
tmp->blt = gen5_render_composite_blt; |
tmp->done = gen5_render_composite_done; |
if (!kgem_check_bo(&sna->kgem, |
tmp->dst.bo, tmp->src.bo, tmp->mask.bo, NULL)) { |
kgem_submit(&sna->kgem); |
} |
gen5_bind_surfaces(sna, tmp); |
gen5_align_vertex(sna, tmp); |
return true; |
} |
/drivers/video/Intel-2D/sna/gen5_render.h |
---|
0,0 → 1,2772 |
/************************************************************************** |
* |
* Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. |
* All Rights Reserved. |
* |
* Permission is hereby granted, free of charge, to any person obtaining a |
* copy of this software and associated documentation files (the |
* "Software"), to deal in the Software without restriction, including |
* without limitation the rights to use, copy, modify, merge, publish, |
* distribute, sub license, and/or sell copies of the Software, and to |
* permit persons to whom the Software is furnished to do so, subject to |
* the following conditions: |
* |
* The above copyright notice and this permission notice (including the |
* next paragraph) shall be included in all copies or substantial portions |
* of the Software. |
* |
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS |
* OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF |
* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. |
* IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR |
* ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, |
* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE |
* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. |
* |
**************************************************************************/ |
#ifndef GEN5_RENDER_H |
#define GEN5_RENDER_H |
#define GEN5_3D(Pipeline,Opcode,Subopcode) ((3 << 29) | \ |
((Pipeline) << 27) | \ |
((Opcode) << 24) | \ |
((Subopcode) << 16)) |
#define GEN5_URB_FENCE GEN5_3D(0, 0, 0) |
#define GEN5_CS_URB_STATE GEN5_3D(0, 0, 1) |
#define GEN5_CONSTANT_BUFFER GEN5_3D(0, 0, 2) |
#define GEN5_STATE_PREFETCH GEN5_3D(0, 0, 3) |
#define GEN5_STATE_BASE_ADDRESS GEN5_3D(0, 1, 1) |
#define GEN5_STATE_SIP GEN5_3D(0, 1, 2) |
#define GEN5_PIPELINE_SELECT GEN5_3D(1, 1, 4) |
#define GEN5_MEDIA_STATE_POINTERS GEN5_3D(2, 0, 0) |
#define GEN5_MEDIA_OBJECT GEN5_3D(2, 1, 0) |
#define GEN5_3DSTATE_PIPELINED_POINTERS GEN5_3D(3, 0, 0) |
#define GEN5_3DSTATE_BINDING_TABLE_POINTERS GEN5_3D(3, 0, 1) |
# define GEN6_3DSTATE_BINDING_TABLE_MODIFY_PS (1 << 12)/* for GEN6 */ |
# define GEN6_3DSTATE_BINDING_TABLE_MODIFY_GS (1 << 9) /* for GEN6 */ |
# define GEN6_3DSTATE_BINDING_TABLE_MODIFY_VS (1 << 8) /* for GEN6 */ |
#define GEN5_3DSTATE_VERTEX_BUFFERS GEN5_3D(3, 0, 8) |
#define GEN5_3DSTATE_VERTEX_ELEMENTS GEN5_3D(3, 0, 9) |
#define GEN5_3DSTATE_INDEX_BUFFER GEN5_3D(3, 0, 0xa) |
#define GEN5_3DSTATE_VF_STATISTICS GEN5_3D(3, 0, 0xb) |
#define GEN5_3DSTATE_DRAWING_RECTANGLE GEN5_3D(3, 1, 0) |
#define GEN5_3DSTATE_CONSTANT_COLOR GEN5_3D(3, 1, 1) |
#define GEN5_3DSTATE_SAMPLER_PALETTE_LOAD GEN5_3D(3, 1, 2) |
#define GEN5_3DSTATE_CHROMA_KEY GEN5_3D(3, 1, 4) |
#define GEN5_3DSTATE_DEPTH_BUFFER GEN5_3D(3, 1, 5) |
# define GEN5_3DSTATE_DEPTH_BUFFER_TYPE_SHIFT 29 |
# define GEN5_3DSTATE_DEPTH_BUFFER_FORMAT_SHIFT 18 |
#define GEN5_3DSTATE_POLY_STIPPLE_OFFSET GEN5_3D(3, 1, 6) |
#define GEN5_3DSTATE_POLY_STIPPLE_PATTERN GEN5_3D(3, 1, 7) |
#define GEN5_3DSTATE_LINE_STIPPLE GEN5_3D(3, 1, 8) |
#define GEN5_3DSTATE_GLOBAL_DEPTH_OFFSET_CLAMP GEN5_3D(3, 1, 9) |
/* These two are BLC and CTG only, not BW or CL */ |
#define GEN5_3DSTATE_AA_LINE_PARAMS GEN5_3D(3, 1, 0xa) |
#define GEN5_3DSTATE_GS_SVB_INDEX GEN5_3D(3, 1, 0xb) |
#define GEN5_PIPE_CONTROL GEN5_3D(3, 2, 0) |
#define GEN5_3DPRIMITIVE GEN5_3D(3, 3, 0) |
#define GEN5_3DSTATE_CLEAR_PARAMS GEN5_3D(3, 1, 0x10) |
/* DW1 */ |
# define GEN5_3DSTATE_DEPTH_CLEAR_VALID (1 << 15) |
/* for GEN6+ */ |
#define GEN6_3DSTATE_SAMPLER_STATE_POINTERS GEN5_3D(3, 0, 0x02) |
# define GEN6_3DSTATE_SAMPLER_STATE_MODIFY_PS (1 << 12) |
# define GEN6_3DSTATE_SAMPLER_STATE_MODIFY_GS (1 << 9) |
# define GEN6_3DSTATE_SAMPLER_STATE_MODIFY_VS (1 << 8) |
#define GEN6_3DSTATE_URB GEN5_3D(3, 0, 0x05) |
/* DW1 */ |
# define GEN6_3DSTATE_URB_VS_SIZE_SHIFT 16 |
# define GEN6_3DSTATE_URB_VS_ENTRIES_SHIFT 0 |
/* DW2 */ |
# define GEN6_3DSTATE_URB_GS_ENTRIES_SHIFT 8 |
# define GEN6_3DSTATE_URB_GS_SIZE_SHIFT 0 |
#define GEN6_3DSTATE_VIEWPORT_STATE_POINTERS GEN5_3D(3, 0, 0x0d) |
# define GEN6_3DSTATE_VIEWPORT_STATE_MODIFY_CC (1 << 12) |
# define GEN6_3DSTATE_VIEWPORT_STATE_MODIFY_SF (1 << 11) |
# define GEN6_3DSTATE_VIEWPORT_STATE_MODIFY_CLIP (1 << 10) |
#define GEN6_3DSTATE_CC_STATE_POINTERS GEN5_3D(3, 0, 0x0e) |
#define GEN6_3DSTATE_VS GEN5_3D(3, 0, 0x10) |
#define GEN6_3DSTATE_GS GEN5_3D(3, 0, 0x11) |
/* DW4 */ |
# define GEN6_3DSTATE_GS_DISPATCH_START_GRF_SHIFT 0 |
#define GEN6_3DSTATE_CLIP GEN5_3D(3, 0, 0x12) |
#define GEN6_3DSTATE_SF GEN5_3D(3, 0, 0x13) |
/* DW1 */ |
# define GEN6_3DSTATE_SF_NUM_OUTPUTS_SHIFT 22 |
# define GEN6_3DSTATE_SF_URB_ENTRY_READ_LENGTH_SHIFT 11 |
# define GEN6_3DSTATE_SF_URB_ENTRY_READ_OFFSET_SHIFT 4 |
/* DW2 */ |
/* DW3 */ |
# define GEN6_3DSTATE_SF_CULL_BOTH (0 << 29) |
# define GEN6_3DSTATE_SF_CULL_NONE (1 << 29) |
# define GEN6_3DSTATE_SF_CULL_FRONT (2 << 29) |
# define GEN6_3DSTATE_SF_CULL_BACK (3 << 29) |
/* DW4 */ |
# define GEN6_3DSTATE_SF_TRI_PROVOKE_SHIFT 29 |
# define GEN6_3DSTATE_SF_LINE_PROVOKE_SHIFT 27 |
# define GEN6_3DSTATE_SF_TRIFAN_PROVOKE_SHIFT 25 |
#define GEN6_3DSTATE_WM GEN5_3D(3, 0, 0x14) |
/* DW2 */ |
# define GEN6_3DSTATE_WM_SAMPLER_COUNT_SHITF 27 |
# define GEN6_3DSTATE_WM_BINDING_TABLE_ENTRY_COUNT_SHIFT 18 |
/* DW4 */ |
# define GEN6_3DSTATE_WM_DISPATCH_START_GRF_0_SHIFT 16 |
/* DW5 */ |
# define GEN6_3DSTATE_WM_MAX_THREADS_SHIFT 25 |
# define GEN6_3DSTATE_WM_DISPATCH_ENABLE (1 << 19) |
# define GEN6_3DSTATE_WM_16_DISPATCH_ENABLE (1 << 1) |
# define GEN6_3DSTATE_WM_8_DISPATCH_ENABLE (1 << 0) |
/* DW6 */ |
# define GEN6_3DSTATE_WM_NUM_SF_OUTPUTS_SHIFT 20 |
# define GEN6_3DSTATE_WM_NONPERSPECTIVE_SAMPLE_BARYCENTRIC (1 << 15) |
# define GEN6_3DSTATE_WM_NONPERSPECTIVE_CENTROID_BARYCENTRIC (1 << 14) |
# define GEN6_3DSTATE_WM_NONPERSPECTIVE_PIXEL_BARYCENTRIC (1 << 13) |
# define GEN6_3DSTATE_WM_PERSPECTIVE_SAMPLE_BARYCENTRIC (1 << 12) |
# define GEN6_3DSTATE_WM_PERSPECTIVE_CENTROID_BARYCENTRIC (1 << 11) |
# define GEN6_3DSTATE_WM_PERSPECTIVE_PIXEL_BARYCENTRIC (1 << 10) |
#define GEN6_3DSTATE_CONSTANT_VS GEN5_3D(3, 0, 0x15) |
#define GEN6_3DSTATE_CONSTANT_GS GEN5_3D(3, 0, 0x16) |
#define GEN6_3DSTATE_CONSTANT_PS GEN5_3D(3, 0, 0x17) |
#define GEN6_3DSTATE_SAMPLE_MASK GEN5_3D(3, 0, 0x18) |
#define GEN6_3DSTATE_MULTISAMPLE GEN5_3D(3, 1, 0x0d) |
/* DW1 */ |
# define GEN6_3DSTATE_MULTISAMPLE_PIXEL_LOCATION_CENTER (0 << 4) |
# define GEN6_3DSTATE_MULTISAMPLE_PIXEL_LOCATION_UPPER_LEFT (1 << 4) |
# define GEN6_3DSTATE_MULTISAMPLE_NUMSAMPLES_1 (0 << 1) |
# define GEN6_3DSTATE_MULTISAMPLE_NUMSAMPLES_4 (2 << 1) |
# define GEN6_3DSTATE_MULTISAMPLE_NUMSAMPLES_8 (3 << 1) |
#define PIPELINE_SELECT_3D 0 |
#define PIPELINE_SELECT_MEDIA 1 |
#define UF0_CS_REALLOC (1 << 13) |
#define UF0_VFE_REALLOC (1 << 12) |
#define UF0_SF_REALLOC (1 << 11) |
#define UF0_CLIP_REALLOC (1 << 10) |
#define UF0_GS_REALLOC (1 << 9) |
#define UF0_VS_REALLOC (1 << 8) |
#define UF1_CLIP_FENCE_SHIFT 20 |
#define UF1_GS_FENCE_SHIFT 10 |
#define UF1_VS_FENCE_SHIFT 0 |
#define UF2_CS_FENCE_SHIFT 20 |
#define UF2_VFE_FENCE_SHIFT 10 |
#define UF2_SF_FENCE_SHIFT 0 |
/* for GEN5_STATE_BASE_ADDRESS */ |
#define BASE_ADDRESS_MODIFY (1 << 0) |
/* for GEN5_3DSTATE_PIPELINED_POINTERS */ |
#define GEN5_GS_DISABLE 0 |
#define GEN5_GS_ENABLE 1 |
#define GEN5_CLIP_DISABLE 0 |
#define GEN5_CLIP_ENABLE 1 |
/* for GEN5_PIPE_CONTROL */ |
#define GEN5_PIPE_CONTROL_NOWRITE (0 << 14) |
#define GEN5_PIPE_CONTROL_WRITE_QWORD (1 << 14) |
#define GEN5_PIPE_CONTROL_WRITE_DEPTH (2 << 14) |
#define GEN5_PIPE_CONTROL_WRITE_TIME (3 << 14) |
#define GEN5_PIPE_CONTROL_DEPTH_STALL (1 << 13) |
#define GEN5_PIPE_CONTROL_WC_FLUSH (1 << 12) |
#define GEN5_PIPE_CONTROL_IS_FLUSH (1 << 11) |
#define GEN5_PIPE_CONTROL_TC_FLUSH (1 << 10) |
#define GEN5_PIPE_CONTROL_NOTIFY_ENABLE (1 << 8) |
#define GEN5_PIPE_CONTROL_GLOBAL_GTT (1 << 2) |
#define GEN5_PIPE_CONTROL_LOCAL_PGTT (0 << 2) |
#define GEN5_PIPE_CONTROL_DEPTH_CACHE_FLUSH (1 << 0) |
/* VERTEX_BUFFER_STATE Structure */ |
#define VB0_BUFFER_INDEX_SHIFT 27 |
#define GEN6_VB0_BUFFER_INDEX_SHIFT 26 |
#define VB0_VERTEXDATA (0 << 26) |
#define VB0_INSTANCEDATA (1 << 26) |
#define GEN6_VB0_VERTEXDATA (0 << 20) |
#define GEN6_VB0_INSTANCEDATA (1 << 20) |
#define VB0_BUFFER_PITCH_SHIFT 0 |
/* VERTEX_ELEMENT_STATE Structure */ |
#define VE0_VERTEX_BUFFER_INDEX_SHIFT 27 |
#define GEN6_VE0_VERTEX_BUFFER_INDEX_SHIFT 26 /* for GEN6 */ |
#define VE0_VALID (1 << 26) |
#define GEN6_VE0_VALID (1 << 25) /* for GEN6 */ |
#define VE0_FORMAT_SHIFT 16 |
#define VE0_OFFSET_SHIFT 0 |
#define VE1_VFCOMPONENT_0_SHIFT 28 |
#define VE1_VFCOMPONENT_1_SHIFT 24 |
#define VE1_VFCOMPONENT_2_SHIFT 20 |
#define VE1_VFCOMPONENT_3_SHIFT 16 |
#define VE1_DESTINATION_ELEMENT_OFFSET_SHIFT 0 |
/* 3DPRIMITIVE bits */ |
#define GEN5_3DPRIMITIVE_VERTEX_SEQUENTIAL (0 << 15) |
#define GEN5_3DPRIMITIVE_VERTEX_RANDOM (1 << 15) |
/* Primitive types are in gen5_defines.h */ |
#define GEN5_3DPRIMITIVE_TOPOLOGY_SHIFT 10 |
#define GEN5_SVG_CTL 0x7400 |
#define GEN5_SVG_CTL_GS_BA (0 << 8) |
#define GEN5_SVG_CTL_SS_BA (1 << 8) |
#define GEN5_SVG_CTL_IO_BA (2 << 8) |
#define GEN5_SVG_CTL_GS_AUB (3 << 8) |
#define GEN5_SVG_CTL_IO_AUB (4 << 8) |
#define GEN5_SVG_CTL_SIP (5 << 8) |
#define GEN5_SVG_RDATA 0x7404 |
#define GEN5_SVG_WORK_CTL 0x7408 |
#define GEN5_VF_CTL 0x7500 |
#define GEN5_VF_CTL_SNAPSHOT_COMPLETE (1 << 31) |
#define GEN5_VF_CTL_SNAPSHOT_MUX_SELECT_THREADID (0 << 8) |
#define GEN5_VF_CTL_SNAPSHOT_MUX_SELECT_VF_DEBUG (1 << 8) |
#define GEN5_VF_CTL_SNAPSHOT_TYPE_VERTEX_SEQUENCE (0 << 4) |
#define GEN5_VF_CTL_SNAPSHOT_TYPE_VERTEX_INDEX (1 << 4) |
#define GEN5_VF_CTL_SKIP_INITIAL_PRIMITIVES (1 << 3) |
#define GEN5_VF_CTL_MAX_PRIMITIVES_LIMIT_ENABLE (1 << 2) |
#define GEN5_VF_CTL_VERTEX_RANGE_LIMIT_ENABLE (1 << 1) |
#define GEN5_VF_CTL_SNAPSHOT_ENABLE (1 << 0) |
#define GEN5_VF_STRG_VAL 0x7504 |
#define GEN5_VF_STR_VL_OVR 0x7508 |
#define GEN5_VF_VC_OVR 0x750c |
#define GEN5_VF_STR_PSKIP 0x7510 |
#define GEN5_VF_MAX_PRIM 0x7514 |
#define GEN5_VF_RDATA 0x7518 |
#define GEN5_VS_CTL 0x7600 |
#define GEN5_VS_CTL_SNAPSHOT_COMPLETE (1 << 31) |
#define GEN5_VS_CTL_SNAPSHOT_MUX_VERTEX_0 (0 << 8) |
#define GEN5_VS_CTL_SNAPSHOT_MUX_VERTEX_1 (1 << 8) |
#define GEN5_VS_CTL_SNAPSHOT_MUX_VALID_COUNT (2 << 8) |
#define GEN5_VS_CTL_SNAPSHOT_MUX_VS_KERNEL_POINTER (3 << 8) |
#define GEN5_VS_CTL_SNAPSHOT_ALL_THREADS (1 << 2) |
#define GEN5_VS_CTL_THREAD_SNAPSHOT_ENABLE (1 << 1) |
#define GEN5_VS_CTL_SNAPSHOT_ENABLE (1 << 0) |
#define GEN5_VS_STRG_VAL 0x7604 |
#define GEN5_VS_RDATA 0x7608 |
#define GEN5_SF_CTL 0x7b00 |
#define GEN5_SF_CTL_SNAPSHOT_COMPLETE (1 << 31) |
#define GEN5_SF_CTL_SNAPSHOT_MUX_VERTEX_0_FF_ID (0 << 8) |
#define GEN5_SF_CTL_SNAPSHOT_MUX_VERTEX_0_REL_COUNT (1 << 8) |
#define GEN5_SF_CTL_SNAPSHOT_MUX_VERTEX_1_FF_ID (2 << 8) |
#define GEN5_SF_CTL_SNAPSHOT_MUX_VERTEX_1_REL_COUNT (3 << 8) |
#define GEN5_SF_CTL_SNAPSHOT_MUX_VERTEX_2_FF_ID (4 << 8) |
#define GEN5_SF_CTL_SNAPSHOT_MUX_VERTEX_2_REL_COUNT (5 << 8) |
#define GEN5_SF_CTL_SNAPSHOT_MUX_VERTEX_COUNT (6 << 8) |
#define GEN5_SF_CTL_SNAPSHOT_MUX_SF_KERNEL_POINTER (7 << 8) |
#define GEN5_SF_CTL_MIN_MAX_PRIMITIVE_RANGE_ENABLE (1 << 4) |
#define GEN5_SF_CTL_DEBUG_CLIP_RECTANGLE_ENABLE (1 << 3) |
#define GEN5_SF_CTL_SNAPSHOT_ALL_THREADS (1 << 2) |
#define GEN5_SF_CTL_THREAD_SNAPSHOT_ENABLE (1 << 1) |
#define GEN5_SF_CTL_SNAPSHOT_ENABLE (1 << 0) |
#define GEN5_SF_STRG_VAL 0x7b04 |
#define GEN5_SF_RDATA 0x7b18 |
#define GEN5_WIZ_CTL 0x7c00 |
#define GEN5_WIZ_CTL_SNAPSHOT_COMPLETE (1 << 31) |
#define GEN5_WIZ_CTL_SUBSPAN_INSTANCE_SHIFT 16 |
#define GEN5_WIZ_CTL_SNAPSHOT_MUX_WIZ_KERNEL_POINTER (0 << 8) |
#define GEN5_WIZ_CTL_SNAPSHOT_MUX_SUBSPAN_INSTANCE (1 << 8) |
#define GEN5_WIZ_CTL_SNAPSHOT_MUX_PRIMITIVE_SEQUENCE (2 << 8) |
#define GEN5_WIZ_CTL_SINGLE_SUBSPAN_DISPATCH (1 << 6) |
#define GEN5_WIZ_CTL_IGNORE_COLOR_SCOREBOARD_STALLS (1 << 5) |
#define GEN5_WIZ_CTL_ENABLE_SUBSPAN_INSTANCE_COMPARE (1 << 4) |
#define GEN5_WIZ_CTL_USE_UPSTREAM_SNAPSHOT_FLAG (1 << 3) |
#define GEN5_WIZ_CTL_SNAPSHOT_ALL_THREADS (1 << 2) |
#define GEN5_WIZ_CTL_THREAD_SNAPSHOT_ENABLE (1 << 1) |
#define GEN5_WIZ_CTL_SNAPSHOT_ENABLE (1 << 0) |
#define GEN5_WIZ_STRG_VAL 0x7c04 |
#define GEN5_WIZ_RDATA 0x7c18 |
#define GEN5_TS_CTL 0x7e00 |
#define GEN5_TS_CTL_SNAPSHOT_COMPLETE (1 << 31) |
#define GEN5_TS_CTL_SNAPSHOT_MESSAGE_ERROR (0 << 8) |
#define GEN5_TS_CTL_SNAPSHOT_INTERFACE_DESCRIPTOR (3 << 8) |
#define GEN5_TS_CTL_SNAPSHOT_ALL_CHILD_THREADS (1 << 2) |
#define GEN5_TS_CTL_SNAPSHOT_ALL_ROOT_THREADS (1 << 1) |
#define GEN5_TS_CTL_SNAPSHOT_ENABLE (1 << 0) |
#define GEN5_TS_STRG_VAL 0x7e04 |
#define GEN5_TS_RDATA 0x7e08 |
#define GEN5_TD_CTL 0x8000 |
#define GEN5_TD_CTL_MUX_SHIFT 8 |
#define GEN5_TD_CTL_EXTERNAL_HALT_R0_DEBUG_MATCH (1 << 7) |
#define GEN5_TD_CTL_FORCE_EXTERNAL_HALT (1 << 6) |
#define GEN5_TD_CTL_EXCEPTION_MASK_OVERRIDE (1 << 5) |
#define GEN5_TD_CTL_FORCE_THREAD_BREAKPOINT_ENABLE (1 << 4) |
#define GEN5_TD_CTL_BREAKPOINT_ENABLE (1 << 2) |
#define GEN5_TD_CTL2 0x8004 |
#define GEN5_TD_CTL2_ILLEGAL_OPCODE_EXCEPTION_OVERRIDE (1 << 28) |
#define GEN5_TD_CTL2_MASKSTACK_EXCEPTION_OVERRIDE (1 << 26) |
#define GEN5_TD_CTL2_SOFTWARE_EXCEPTION_OVERRIDE (1 << 25) |
#define GEN5_TD_CTL2_ACTIVE_THREAD_LIMIT_SHIFT 16 |
#define GEN5_TD_CTL2_ACTIVE_THREAD_LIMIT_ENABLE (1 << 8) |
#define GEN5_TD_CTL2_THREAD_SPAWNER_EXECUTION_MASK_ENABLE (1 << 7) |
#define GEN5_TD_CTL2_WIZ_EXECUTION_MASK_ENABLE (1 << 6) |
#define GEN5_TD_CTL2_SF_EXECUTION_MASK_ENABLE (1 << 5) |
#define GEN5_TD_CTL2_CLIPPER_EXECUTION_MASK_ENABLE (1 << 4) |
#define GEN5_TD_CTL2_GS_EXECUTION_MASK_ENABLE (1 << 3) |
#define GEN5_TD_CTL2_VS_EXECUTION_MASK_ENABLE (1 << 0) |
#define GEN5_TD_VF_VS_EMSK 0x8008 |
#define GEN5_TD_GS_EMSK 0x800c |
#define GEN5_TD_CLIP_EMSK 0x8010 |
#define GEN5_TD_SF_EMSK 0x8014 |
#define GEN5_TD_WIZ_EMSK 0x8018 |
#define GEN5_TD_0_6_EHTRG_VAL 0x801c |
#define GEN5_TD_0_7_EHTRG_VAL 0x8020 |
#define GEN5_TD_0_6_EHTRG_MSK 0x8024 |
#define GEN5_TD_0_7_EHTRG_MSK 0x8028 |
#define GEN5_TD_RDATA 0x802c |
#define GEN5_TD_TS_EMSK 0x8030 |
#define GEN5_EU_CTL 0x8800 |
#define GEN5_EU_CTL_SELECT_SHIFT 16 |
#define GEN5_EU_CTL_DATA_MUX_SHIFT 8 |
#define GEN5_EU_ATT_0 0x8810 |
#define GEN5_EU_ATT_1 0x8814 |
#define GEN5_EU_ATT_DATA_0 0x8820 |
#define GEN5_EU_ATT_DATA_1 0x8824 |
#define GEN5_EU_ATT_CLR_0 0x8830 |
#define GEN5_EU_ATT_CLR_1 0x8834 |
#define GEN5_EU_RDATA 0x8840 |
/* 3D state: |
*/ |
#define _3DOP_3DSTATE_PIPELINED 0x0 |
#define _3DOP_3DSTATE_NONPIPELINED 0x1 |
#define _3DOP_3DCONTROL 0x2 |
#define _3DOP_3DPRIMITIVE 0x3 |
#define _3DSTATE_PIPELINED_POINTERS 0x00 |
#define _3DSTATE_BINDING_TABLE_POINTERS 0x01 |
#define _3DSTATE_VERTEX_BUFFERS 0x08 |
#define _3DSTATE_VERTEX_ELEMENTS 0x09 |
#define _3DSTATE_INDEX_BUFFER 0x0A |
#define _3DSTATE_VF_STATISTICS 0x0B |
#define _3DSTATE_DRAWING_RECTANGLE 0x00 |
#define _3DSTATE_CONSTANT_COLOR 0x01 |
#define _3DSTATE_SAMPLER_PALETTE_LOAD 0x02 |
#define _3DSTATE_CHROMA_KEY 0x04 |
#define _3DSTATE_DEPTH_BUFFER 0x05 |
#define _3DSTATE_POLY_STIPPLE_OFFSET 0x06 |
#define _3DSTATE_POLY_STIPPLE_PATTERN 0x07 |
#define _3DSTATE_LINE_STIPPLE 0x08 |
#define _3DSTATE_GLOBAL_DEPTH_OFFSET_CLAMP 0x09 |
#define _3DCONTROL 0x00 |
#define _3DPRIMITIVE 0x00 |
#define _3DPRIM_POINTLIST 0x01 |
#define _3DPRIM_LINELIST 0x02 |
#define _3DPRIM_LINESTRIP 0x03 |
#define _3DPRIM_TRILIST 0x04 |
#define _3DPRIM_TRISTRIP 0x05 |
#define _3DPRIM_TRIFAN 0x06 |
#define _3DPRIM_QUADLIST 0x07 |
#define _3DPRIM_QUADSTRIP 0x08 |
#define _3DPRIM_LINELIST_ADJ 0x09 |
#define _3DPRIM_LINESTRIP_ADJ 0x0A |
#define _3DPRIM_TRILIST_ADJ 0x0B |
#define _3DPRIM_TRISTRIP_ADJ 0x0C |
#define _3DPRIM_TRISTRIP_REVERSE 0x0D |
#define _3DPRIM_POLYGON 0x0E |
#define _3DPRIM_RECTLIST 0x0F |
#define _3DPRIM_LINELOOP 0x10 |
#define _3DPRIM_POINTLIST_BF 0x11 |
#define _3DPRIM_LINESTRIP_CONT 0x12 |
#define _3DPRIM_LINESTRIP_BF 0x13 |
#define _3DPRIM_LINESTRIP_CONT_BF 0x14 |
#define _3DPRIM_TRIFAN_NOSTIPPLE 0x15 |
#define _3DPRIM_VERTEXBUFFER_ACCESS_SEQUENTIAL 0 |
#define _3DPRIM_VERTEXBUFFER_ACCESS_RANDOM 1 |
#define GEN5_ANISORATIO_2 0 |
#define GEN5_ANISORATIO_4 1 |
#define GEN5_ANISORATIO_6 2 |
#define GEN5_ANISORATIO_8 3 |
#define GEN5_ANISORATIO_10 4 |
#define GEN5_ANISORATIO_12 5 |
#define GEN5_ANISORATIO_14 6 |
#define GEN5_ANISORATIO_16 7 |
#define GEN5_BLENDFACTOR_ONE 0x1 |
#define GEN5_BLENDFACTOR_SRC_COLOR 0x2 |
#define GEN5_BLENDFACTOR_SRC_ALPHA 0x3 |
#define GEN5_BLENDFACTOR_DST_ALPHA 0x4 |
#define GEN5_BLENDFACTOR_DST_COLOR 0x5 |
#define GEN5_BLENDFACTOR_SRC_ALPHA_SATURATE 0x6 |
#define GEN5_BLENDFACTOR_CONST_COLOR 0x7 |
#define GEN5_BLENDFACTOR_CONST_ALPHA 0x8 |
#define GEN5_BLENDFACTOR_SRC1_COLOR 0x9 |
#define GEN5_BLENDFACTOR_SRC1_ALPHA 0x0A |
#define GEN5_BLENDFACTOR_ZERO 0x11 |
#define GEN5_BLENDFACTOR_INV_SRC_COLOR 0x12 |
#define GEN5_BLENDFACTOR_INV_SRC_ALPHA 0x13 |
#define GEN5_BLENDFACTOR_INV_DST_ALPHA 0x14 |
#define GEN5_BLENDFACTOR_INV_DST_COLOR 0x15 |
#define GEN5_BLENDFACTOR_INV_CONST_COLOR 0x17 |
#define GEN5_BLENDFACTOR_INV_CONST_ALPHA 0x18 |
#define GEN5_BLENDFACTOR_INV_SRC1_COLOR 0x19 |
#define GEN5_BLENDFACTOR_INV_SRC1_ALPHA 0x1A |
#define GEN5_BLENDFUNCTION_ADD 0 |
#define GEN5_BLENDFUNCTION_SUBTRACT 1 |
#define GEN5_BLENDFUNCTION_REVERSE_SUBTRACT 2 |
#define GEN5_BLENDFUNCTION_MIN 3 |
#define GEN5_BLENDFUNCTION_MAX 4 |
#define GEN5_ALPHATEST_FORMAT_UNORM8 0 |
#define GEN5_ALPHATEST_FORMAT_FLOAT32 1 |
#define GEN5_CHROMAKEY_KILL_ON_ANY_MATCH 0 |
#define GEN5_CHROMAKEY_REPLACE_BLACK 1 |
#define GEN5_CLIP_API_OGL 0 |
#define GEN5_CLIP_API_DX 1 |
#define GEN5_CLIPMODE_NORMAL 0 |
#define GEN5_CLIPMODE_CLIP_ALL 1 |
#define GEN5_CLIPMODE_CLIP_NON_REJECTED 2 |
#define GEN5_CLIPMODE_REJECT_ALL 3 |
#define GEN5_CLIPMODE_ACCEPT_ALL 4 |
#define GEN5_CLIP_NDCSPACE 0 |
#define GEN5_CLIP_SCREENSPACE 1 |
#define GEN5_COMPAREFUNCTION_ALWAYS 0 |
#define GEN5_COMPAREFUNCTION_NEVER 1 |
#define GEN5_COMPAREFUNCTION_LESS 2 |
#define GEN5_COMPAREFUNCTION_EQUAL 3 |
#define GEN5_COMPAREFUNCTION_LEQUAL 4 |
#define GEN5_COMPAREFUNCTION_GREATER 5 |
#define GEN5_COMPAREFUNCTION_NOTEQUAL 6 |
#define GEN5_COMPAREFUNCTION_GEQUAL 7 |
#define GEN5_COVERAGE_PIXELS_HALF 0 |
#define GEN5_COVERAGE_PIXELS_1 1 |
#define GEN5_COVERAGE_PIXELS_2 2 |
#define GEN5_COVERAGE_PIXELS_4 3 |
#define GEN5_CULLMODE_BOTH 0 |
#define GEN5_CULLMODE_NONE 1 |
#define GEN5_CULLMODE_FRONT 2 |
#define GEN5_CULLMODE_BACK 3 |
#define GEN5_DEFAULTCOLOR_R8G8B8A8_UNORM 0 |
#define GEN5_DEFAULTCOLOR_R32G32B32A32_FLOAT 1 |
#define GEN5_DEPTHFORMAT_D32_FLOAT_S8X24_UINT 0 |
#define GEN5_DEPTHFORMAT_D32_FLOAT 1 |
#define GEN5_DEPTHFORMAT_D24_UNORM_S8_UINT 2 |
#define GEN5_DEPTHFORMAT_D16_UNORM 5 |
#define GEN5_FLOATING_POINT_IEEE_754 0 |
#define GEN5_FLOATING_POINT_NON_IEEE_754 1 |
#define GEN5_FRONTWINDING_CW 0 |
#define GEN5_FRONTWINDING_CCW 1 |
#define GEN5_INDEX_BYTE 0 |
#define GEN5_INDEX_WORD 1 |
#define GEN5_INDEX_DWORD 2 |
#define GEN5_LOGICOPFUNCTION_CLEAR 0 |
#define GEN5_LOGICOPFUNCTION_NOR 1 |
#define GEN5_LOGICOPFUNCTION_AND_INVERTED 2 |
#define GEN5_LOGICOPFUNCTION_COPY_INVERTED 3 |
#define GEN5_LOGICOPFUNCTION_AND_REVERSE 4 |
#define GEN5_LOGICOPFUNCTION_INVERT 5 |
#define GEN5_LOGICOPFUNCTION_XOR 6 |
#define GEN5_LOGICOPFUNCTION_NAND 7 |
#define GEN5_LOGICOPFUNCTION_AND 8 |
#define GEN5_LOGICOPFUNCTION_EQUIV 9 |
#define GEN5_LOGICOPFUNCTION_NOOP 10 |
#define GEN5_LOGICOPFUNCTION_OR_INVERTED 11 |
#define GEN5_LOGICOPFUNCTION_COPY 12 |
#define GEN5_LOGICOPFUNCTION_OR_REVERSE 13 |
#define GEN5_LOGICOPFUNCTION_OR 14 |
#define GEN5_LOGICOPFUNCTION_SET 15 |
#define GEN5_MAPFILTER_NEAREST 0x0 |
#define GEN5_MAPFILTER_LINEAR 0x1 |
#define GEN5_MAPFILTER_ANISOTROPIC 0x2 |
#define GEN5_MIPFILTER_NONE 0 |
#define GEN5_MIPFILTER_NEAREST 1 |
#define GEN5_MIPFILTER_LINEAR 3 |
#define GEN5_POLYGON_FRONT_FACING 0 |
#define GEN5_POLYGON_BACK_FACING 1 |
#define GEN5_PREFILTER_ALWAYS 0x0 |
#define GEN5_PREFILTER_NEVER 0x1 |
#define GEN5_PREFILTER_LESS 0x2 |
#define GEN5_PREFILTER_EQUAL 0x3 |
#define GEN5_PREFILTER_LEQUAL 0x4 |
#define GEN5_PREFILTER_GREATER 0x5 |
#define GEN5_PREFILTER_NOTEQUAL 0x6 |
#define GEN5_PREFILTER_GEQUAL 0x7 |
#define GEN5_PROVOKING_VERTEX_0 0 |
#define GEN5_PROVOKING_VERTEX_1 1 |
#define GEN5_PROVOKING_VERTEX_2 2 |
#define GEN5_RASTRULE_UPPER_LEFT 0 |
#define GEN5_RASTRULE_UPPER_RIGHT 1 |
#define GEN5_RENDERTARGET_CLAMPRANGE_UNORM 0 |
#define GEN5_RENDERTARGET_CLAMPRANGE_SNORM 1 |
#define GEN5_RENDERTARGET_CLAMPRANGE_FORMAT 2 |
#define GEN5_STENCILOP_KEEP 0 |
#define GEN5_STENCILOP_ZERO 1 |
#define GEN5_STENCILOP_REPLACE 2 |
#define GEN5_STENCILOP_INCRSAT 3 |
#define GEN5_STENCILOP_DECRSAT 4 |
#define GEN5_STENCILOP_INCR 5 |
#define GEN5_STENCILOP_DECR 6 |
#define GEN5_STENCILOP_INVERT 7 |
#define GEN5_SURFACE_MIPMAPLAYOUT_BELOW 0 |
#define GEN5_SURFACE_MIPMAPLAYOUT_RIGHT 1 |
#define GEN5_SURFACEFORMAT_R32G32B32A32_FLOAT 0x000 |
#define GEN5_SURFACEFORMAT_R32G32B32A32_SINT 0x001 |
#define GEN5_SURFACEFORMAT_R32G32B32A32_UINT 0x002 |
#define GEN5_SURFACEFORMAT_R32G32B32A32_UNORM 0x003 |
#define GEN5_SURFACEFORMAT_R32G32B32A32_SNORM 0x004 |
#define GEN5_SURFACEFORMAT_R64G64_FLOAT 0x005 |
#define GEN5_SURFACEFORMAT_R32G32B32X32_FLOAT 0x006 |
#define GEN5_SURFACEFORMAT_R32G32B32A32_SSCALED 0x007 |
#define GEN5_SURFACEFORMAT_R32G32B32A32_USCALED 0x008 |
#define GEN5_SURFACEFORMAT_R32G32B32_FLOAT 0x040 |
#define GEN5_SURFACEFORMAT_R32G32B32_SINT 0x041 |
#define GEN5_SURFACEFORMAT_R32G32B32_UINT 0x042 |
#define GEN5_SURFACEFORMAT_R32G32B32_UNORM 0x043 |
#define GEN5_SURFACEFORMAT_R32G32B32_SNORM 0x044 |
#define GEN5_SURFACEFORMAT_R32G32B32_SSCALED 0x045 |
#define GEN5_SURFACEFORMAT_R32G32B32_USCALED 0x046 |
#define GEN5_SURFACEFORMAT_R16G16B16A16_UNORM 0x080 |
#define GEN5_SURFACEFORMAT_R16G16B16A16_SNORM 0x081 |
#define GEN5_SURFACEFORMAT_R16G16B16A16_SINT 0x082 |
#define GEN5_SURFACEFORMAT_R16G16B16A16_UINT 0x083 |
#define GEN5_SURFACEFORMAT_R16G16B16A16_FLOAT 0x084 |
#define GEN5_SURFACEFORMAT_R32G32_FLOAT 0x085 |
#define GEN5_SURFACEFORMAT_R32G32_SINT 0x086 |
#define GEN5_SURFACEFORMAT_R32G32_UINT 0x087 |
#define GEN5_SURFACEFORMAT_R32_FLOAT_X8X24_TYPELESS 0x088 |
#define GEN5_SURFACEFORMAT_X32_TYPELESS_G8X24_UINT 0x089 |
#define GEN5_SURFACEFORMAT_L32A32_FLOAT 0x08A |
#define GEN5_SURFACEFORMAT_R32G32_UNORM 0x08B |
#define GEN5_SURFACEFORMAT_R32G32_SNORM 0x08C |
#define GEN5_SURFACEFORMAT_R64_FLOAT 0x08D |
#define GEN5_SURFACEFORMAT_R16G16B16X16_UNORM 0x08E |
#define GEN5_SURFACEFORMAT_R16G16B16X16_FLOAT 0x08F |
#define GEN5_SURFACEFORMAT_A32X32_FLOAT 0x090 |
#define GEN5_SURFACEFORMAT_L32X32_FLOAT 0x091 |
#define GEN5_SURFACEFORMAT_I32X32_FLOAT 0x092 |
#define GEN5_SURFACEFORMAT_R16G16B16A16_SSCALED 0x093 |
#define GEN5_SURFACEFORMAT_R16G16B16A16_USCALED 0x094 |
#define GEN5_SURFACEFORMAT_R32G32_SSCALED 0x095 |
#define GEN5_SURFACEFORMAT_R32G32_USCALED 0x096 |
#define GEN5_SURFACEFORMAT_B8G8R8A8_UNORM 0x0C0 |
#define GEN5_SURFACEFORMAT_B8G8R8A8_UNORM_SRGB 0x0C1 |
#define GEN5_SURFACEFORMAT_R10G10B10A2_UNORM 0x0C2 |
#define GEN5_SURFACEFORMAT_R10G10B10A2_UNORM_SRGB 0x0C3 |
#define GEN5_SURFACEFORMAT_R10G10B10A2_UINT 0x0C4 |
#define GEN5_SURFACEFORMAT_R10G10B10_SNORM_A2_UNORM 0x0C5 |
#define GEN5_SURFACEFORMAT_R8G8B8A8_UNORM 0x0C7 |
#define GEN5_SURFACEFORMAT_R8G8B8A8_UNORM_SRGB 0x0C8 |
#define GEN5_SURFACEFORMAT_R8G8B8A8_SNORM 0x0C9 |
#define GEN5_SURFACEFORMAT_R8G8B8A8_SINT 0x0CA |
#define GEN5_SURFACEFORMAT_R8G8B8A8_UINT 0x0CB |
#define GEN5_SURFACEFORMAT_R16G16_UNORM 0x0CC |
#define GEN5_SURFACEFORMAT_R16G16_SNORM 0x0CD |
#define GEN5_SURFACEFORMAT_R16G16_SINT 0x0CE |
#define GEN5_SURFACEFORMAT_R16G16_UINT 0x0CF |
#define GEN5_SURFACEFORMAT_R16G16_FLOAT 0x0D0 |
#define GEN5_SURFACEFORMAT_B10G10R10A2_UNORM 0x0D1 |
#define GEN5_SURFACEFORMAT_B10G10R10A2_UNORM_SRGB 0x0D2 |
#define GEN5_SURFACEFORMAT_R11G11B10_FLOAT 0x0D3 |
#define GEN5_SURFACEFORMAT_R32_SINT 0x0D6 |
#define GEN5_SURFACEFORMAT_R32_UINT 0x0D7 |
#define GEN5_SURFACEFORMAT_R32_FLOAT 0x0D8 |
#define GEN5_SURFACEFORMAT_R24_UNORM_X8_TYPELESS 0x0D9 |
#define GEN5_SURFACEFORMAT_X24_TYPELESS_G8_UINT 0x0DA |
#define GEN5_SURFACEFORMAT_L16A16_UNORM 0x0DF |
#define GEN5_SURFACEFORMAT_I24X8_UNORM 0x0E0 |
#define GEN5_SURFACEFORMAT_L24X8_UNORM 0x0E1 |
#define GEN5_SURFACEFORMAT_A24X8_UNORM 0x0E2 |
#define GEN5_SURFACEFORMAT_I32_FLOAT 0x0E3 |
#define GEN5_SURFACEFORMAT_L32_FLOAT 0x0E4 |
#define GEN5_SURFACEFORMAT_A32_FLOAT 0x0E5 |
#define GEN5_SURFACEFORMAT_B8G8R8X8_UNORM 0x0E9 |
#define GEN5_SURFACEFORMAT_B8G8R8X8_UNORM_SRGB 0x0EA |
#define GEN5_SURFACEFORMAT_R8G8B8X8_UNORM 0x0EB |
#define GEN5_SURFACEFORMAT_R8G8B8X8_UNORM_SRGB 0x0EC |
#define GEN5_SURFACEFORMAT_R9G9B9E5_SHAREDEXP 0x0ED |
#define GEN5_SURFACEFORMAT_B10G10R10X2_UNORM 0x0EE |
#define GEN5_SURFACEFORMAT_L16A16_FLOAT 0x0F0 |
#define GEN5_SURFACEFORMAT_R32_UNORM 0x0F1 |
#define GEN5_SURFACEFORMAT_R32_SNORM 0x0F2 |
#define GEN5_SURFACEFORMAT_R10G10B10X2_USCALED 0x0F3 |
#define GEN5_SURFACEFORMAT_R8G8B8A8_SSCALED 0x0F4 |
#define GEN5_SURFACEFORMAT_R8G8B8A8_USCALED 0x0F5 |
#define GEN5_SURFACEFORMAT_R16G16_SSCALED 0x0F6 |
#define GEN5_SURFACEFORMAT_R16G16_USCALED 0x0F7 |
#define GEN5_SURFACEFORMAT_R32_SSCALED 0x0F8 |
#define GEN5_SURFACEFORMAT_R32_USCALED 0x0F9 |
#define GEN5_SURFACEFORMAT_B5G6R5_UNORM 0x100 |
#define GEN5_SURFACEFORMAT_B5G6R5_UNORM_SRGB 0x101 |
#define GEN5_SURFACEFORMAT_B5G5R5A1_UNORM 0x102 |
#define GEN5_SURFACEFORMAT_B5G5R5A1_UNORM_SRGB 0x103 |
#define GEN5_SURFACEFORMAT_B4G4R4A4_UNORM 0x104 |
#define GEN5_SURFACEFORMAT_B4G4R4A4_UNORM_SRGB 0x105 |
#define GEN5_SURFACEFORMAT_R8G8_UNORM 0x106 |
#define GEN5_SURFACEFORMAT_R8G8_SNORM 0x107 |
#define GEN5_SURFACEFORMAT_R8G8_SINT 0x108 |
#define GEN5_SURFACEFORMAT_R8G8_UINT 0x109 |
#define GEN5_SURFACEFORMAT_R16_UNORM 0x10A |
#define GEN5_SURFACEFORMAT_R16_SNORM 0x10B |
#define GEN5_SURFACEFORMAT_R16_SINT 0x10C |
#define GEN5_SURFACEFORMAT_R16_UINT 0x10D |
#define GEN5_SURFACEFORMAT_R16_FLOAT 0x10E |
#define GEN5_SURFACEFORMAT_I16_UNORM 0x111 |
#define GEN5_SURFACEFORMAT_L16_UNORM 0x112 |
#define GEN5_SURFACEFORMAT_A16_UNORM 0x113 |
#define GEN5_SURFACEFORMAT_L8A8_UNORM 0x114 |
#define GEN5_SURFACEFORMAT_I16_FLOAT 0x115 |
#define GEN5_SURFACEFORMAT_L16_FLOAT 0x116 |
#define GEN5_SURFACEFORMAT_A16_FLOAT 0x117 |
#define GEN5_SURFACEFORMAT_R5G5_SNORM_B6_UNORM 0x119 |
#define GEN5_SURFACEFORMAT_B5G5R5X1_UNORM 0x11A |
#define GEN5_SURFACEFORMAT_B5G5R5X1_UNORM_SRGB 0x11B |
#define GEN5_SURFACEFORMAT_R8G8_SSCALED 0x11C |
#define GEN5_SURFACEFORMAT_R8G8_USCALED 0x11D |
#define GEN5_SURFACEFORMAT_R16_SSCALED 0x11E |
#define GEN5_SURFACEFORMAT_R16_USCALED 0x11F |
#define GEN5_SURFACEFORMAT_R8_UNORM 0x140 |
#define GEN5_SURFACEFORMAT_R8_SNORM 0x141 |
#define GEN5_SURFACEFORMAT_R8_SINT 0x142 |
#define GEN5_SURFACEFORMAT_R8_UINT 0x143 |
#define GEN5_SURFACEFORMAT_A8_UNORM 0x144 |
#define GEN5_SURFACEFORMAT_I8_UNORM 0x145 |
#define GEN5_SURFACEFORMAT_L8_UNORM 0x146 |
#define GEN5_SURFACEFORMAT_P4A4_UNORM 0x147 |
#define GEN5_SURFACEFORMAT_A4P4_UNORM 0x148 |
#define GEN5_SURFACEFORMAT_R8_SSCALED 0x149 |
#define GEN5_SURFACEFORMAT_R8_USCALED 0x14A |
#define GEN5_SURFACEFORMAT_R1_UINT 0x181 |
#define GEN5_SURFACEFORMAT_YCRCB_NORMAL 0x182 |
#define GEN5_SURFACEFORMAT_YCRCB_SWAPUVY 0x183 |
#define GEN5_SURFACEFORMAT_BC1_UNORM 0x186 |
#define GEN5_SURFACEFORMAT_BC2_UNORM 0x187 |
#define GEN5_SURFACEFORMAT_BC3_UNORM 0x188 |
#define GEN5_SURFACEFORMAT_BC4_UNORM 0x189 |
#define GEN5_SURFACEFORMAT_BC5_UNORM 0x18A |
#define GEN5_SURFACEFORMAT_BC1_UNORM_SRGB 0x18B |
#define GEN5_SURFACEFORMAT_BC2_UNORM_SRGB 0x18C |
#define GEN5_SURFACEFORMAT_BC3_UNORM_SRGB 0x18D |
#define GEN5_SURFACEFORMAT_MONO8 0x18E |
#define GEN5_SURFACEFORMAT_YCRCB_SWAPUV 0x18F |
#define GEN5_SURFACEFORMAT_YCRCB_SWAPY 0x190 |
#define GEN5_SURFACEFORMAT_DXT1_RGB 0x191 |
#define GEN5_SURFACEFORMAT_FXT1 0x192 |
#define GEN5_SURFACEFORMAT_R8G8B8_UNORM 0x193 |
#define GEN5_SURFACEFORMAT_R8G8B8_SNORM 0x194 |
#define GEN5_SURFACEFORMAT_R8G8B8_SSCALED 0x195 |
#define GEN5_SURFACEFORMAT_R8G8B8_USCALED 0x196 |
#define GEN5_SURFACEFORMAT_R64G64B64A64_FLOAT 0x197 |
#define GEN5_SURFACEFORMAT_R64G64B64_FLOAT 0x198 |
#define GEN5_SURFACEFORMAT_BC4_SNORM 0x199 |
#define GEN5_SURFACEFORMAT_BC5_SNORM 0x19A |
#define GEN5_SURFACEFORMAT_R16G16B16_UNORM 0x19C |
#define GEN5_SURFACEFORMAT_R16G16B16_SNORM 0x19D |
#define GEN5_SURFACEFORMAT_R16G16B16_SSCALED 0x19E |
#define GEN5_SURFACEFORMAT_R16G16B16_USCALED 0x19F |
#define GEN5_SURFACERETURNFORMAT_FLOAT32 0 |
#define GEN5_SURFACERETURNFORMAT_S1 1 |
#define GEN5_SURFACE_1D 0 |
#define GEN5_SURFACE_2D 1 |
#define GEN5_SURFACE_3D 2 |
#define GEN5_SURFACE_CUBE 3 |
#define GEN5_SURFACE_BUFFER 4 |
#define GEN5_SURFACE_NULL 7 |
#define GEN5_BORDER_COLOR_MODE_DEFAULT 0 |
#define GEN5_BORDER_COLOR_MODE_LEGACY 1 |
#define GEN5_TEXCOORDMODE_WRAP 0 |
#define GEN5_TEXCOORDMODE_MIRROR 1 |
#define GEN5_TEXCOORDMODE_CLAMP 2 |
#define GEN5_TEXCOORDMODE_CUBE 3 |
#define GEN5_TEXCOORDMODE_CLAMP_BORDER 4 |
#define GEN5_TEXCOORDMODE_MIRROR_ONCE 5 |
#define GEN5_THREAD_PRIORITY_NORMAL 0 |
#define GEN5_THREAD_PRIORITY_HIGH 1 |
#define GEN5_TILEWALK_XMAJOR 0 |
#define GEN5_TILEWALK_YMAJOR 1 |
#define GEN5_VERTEX_SUBPIXEL_PRECISION_8BITS 0 |
#define GEN5_VERTEX_SUBPIXEL_PRECISION_4BITS 1 |
#define GEN5_VERTEXBUFFER_ACCESS_VERTEXDATA 0 |
#define GEN5_VERTEXBUFFER_ACCESS_INSTANCEDATA 1 |
#define VFCOMPONENT_NOSTORE 0 |
#define VFCOMPONENT_STORE_SRC 1 |
#define VFCOMPONENT_STORE_0 2 |
#define VFCOMPONENT_STORE_1_FLT 3 |
#define VFCOMPONENT_STORE_1_INT 4 |
#define VFCOMPONENT_STORE_VID 5 |
#define VFCOMPONENT_STORE_IID 6 |
#define VFCOMPONENT_STORE_PID 7 |
/* Execution Unit (EU) defines |
*/ |
#define GEN5_ALIGN_1 0 |
#define GEN5_ALIGN_16 1 |
#define GEN5_ADDRESS_DIRECT 0 |
#define GEN5_ADDRESS_REGISTER_INDIRECT_REGISTER 1 |
#define GEN5_CHANNEL_X 0 |
#define GEN5_CHANNEL_Y 1 |
#define GEN5_CHANNEL_Z 2 |
#define GEN5_CHANNEL_W 3 |
#define GEN5_COMPRESSION_NONE 0 |
#define GEN5_COMPRESSION_2NDHALF 1 |
#define GEN5_COMPRESSION_COMPRESSED 2 |
#define GEN5_CONDITIONAL_NONE 0 |
#define GEN5_CONDITIONAL_Z 1 |
#define GEN5_CONDITIONAL_NZ 2 |
#define GEN5_CONDITIONAL_EQ 1 /* Z */ |
#define GEN5_CONDITIONAL_NEQ 2 /* NZ */ |
#define GEN5_CONDITIONAL_G 3 |
#define GEN5_CONDITIONAL_GE 4 |
#define GEN5_CONDITIONAL_L 5 |
#define GEN5_CONDITIONAL_LE 6 |
#define GEN5_CONDITIONAL_C 7 |
#define GEN5_CONDITIONAL_O 8 |
#define GEN5_DEBUG_NONE 0 |
#define GEN5_DEBUG_BREAKPOINT 1 |
#define GEN5_DEPENDENCY_NORMAL 0 |
#define GEN5_DEPENDENCY_NOTCLEARED 1 |
#define GEN5_DEPENDENCY_NOTCHECKED 2 |
#define GEN5_DEPENDENCY_DISABLE 3 |
#define GEN5_EXECUTE_1 0 |
#define GEN5_EXECUTE_2 1 |
#define GEN5_EXECUTE_4 2 |
#define GEN5_EXECUTE_8 3 |
#define GEN5_EXECUTE_16 4 |
#define GEN5_EXECUTE_32 5 |
#define GEN5_HORIZONTAL_STRIDE_0 0 |
#define GEN5_HORIZONTAL_STRIDE_1 1 |
#define GEN5_HORIZONTAL_STRIDE_2 2 |
#define GEN5_HORIZONTAL_STRIDE_4 3 |
#define GEN5_INSTRUCTION_NORMAL 0 |
#define GEN5_INSTRUCTION_SATURATE 1 |
#define GEN5_MASK_ENABLE 0 |
#define GEN5_MASK_DISABLE 1 |
#define GEN5_OPCODE_MOV 1 |
#define GEN5_OPCODE_SEL 2 |
#define GEN5_OPCODE_NOT 4 |
#define GEN5_OPCODE_AND 5 |
#define GEN5_OPCODE_OR 6 |
#define GEN5_OPCODE_XOR 7 |
#define GEN5_OPCODE_SHR 8 |
#define GEN5_OPCODE_SHL 9 |
#define GEN5_OPCODE_RSR 10 |
#define GEN5_OPCODE_RSL 11 |
#define GEN5_OPCODE_ASR 12 |
#define GEN5_OPCODE_CMP 16 |
#define GEN5_OPCODE_JMPI 32 |
#define GEN5_OPCODE_IF 34 |
#define GEN5_OPCODE_IFF 35 |
#define GEN5_OPCODE_ELSE 36 |
#define GEN5_OPCODE_ENDIF 37 |
#define GEN5_OPCODE_DO 38 |
#define GEN5_OPCODE_WHILE 39 |
#define GEN5_OPCODE_BREAK 40 |
#define GEN5_OPCODE_CONTINUE 41 |
#define GEN5_OPCODE_HALT 42 |
#define GEN5_OPCODE_MSAVE 44 |
#define GEN5_OPCODE_MRESTORE 45 |
#define GEN5_OPCODE_PUSH 46 |
#define GEN5_OPCODE_POP 47 |
#define GEN5_OPCODE_WAIT 48 |
#define GEN5_OPCODE_SEND 49 |
#define GEN5_OPCODE_ADD 64 |
#define GEN5_OPCODE_MUL 65 |
#define GEN5_OPCODE_AVG 66 |
#define GEN5_OPCODE_FRC 67 |
#define GEN5_OPCODE_RNDU 68 |
#define GEN5_OPCODE_RNDD 69 |
#define GEN5_OPCODE_RNDE 70 |
#define GEN5_OPCODE_RNDZ 71 |
#define GEN5_OPCODE_MAC 72 |
#define GEN5_OPCODE_MACH 73 |
#define GEN5_OPCODE_LZD 74 |
#define GEN5_OPCODE_SAD2 80 |
#define GEN5_OPCODE_SADA2 81 |
#define GEN5_OPCODE_DP4 84 |
#define GEN5_OPCODE_DPH 85 |
#define GEN5_OPCODE_DP3 86 |
#define GEN5_OPCODE_DP2 87 |
#define GEN5_OPCODE_DPA2 88 |
#define GEN5_OPCODE_LINE 89 |
#define GEN5_OPCODE_NOP 126 |
#define GEN5_PREDICATE_NONE 0 |
#define GEN5_PREDICATE_NORMAL 1 |
#define GEN5_PREDICATE_ALIGN1_ANYV 2 |
#define GEN5_PREDICATE_ALIGN1_ALLV 3 |
#define GEN5_PREDICATE_ALIGN1_ANY2H 4 |
#define GEN5_PREDICATE_ALIGN1_ALL2H 5 |
#define GEN5_PREDICATE_ALIGN1_ANY4H 6 |
#define GEN5_PREDICATE_ALIGN1_ALL4H 7 |
#define GEN5_PREDICATE_ALIGN1_ANY8H 8 |
#define GEN5_PREDICATE_ALIGN1_ALL8H 9 |
#define GEN5_PREDICATE_ALIGN1_ANY16H 10 |
#define GEN5_PREDICATE_ALIGN1_ALL16H 11 |
#define GEN5_PREDICATE_ALIGN16_REPLICATE_X 2 |
#define GEN5_PREDICATE_ALIGN16_REPLICATE_Y 3 |
#define GEN5_PREDICATE_ALIGN16_REPLICATE_Z 4 |
#define GEN5_PREDICATE_ALIGN16_REPLICATE_W 5 |
#define GEN5_PREDICATE_ALIGN16_ANY4H 6 |
#define GEN5_PREDICATE_ALIGN16_ALL4H 7 |
#define GEN5_ARCHITECTURE_REGISTER_FILE 0 |
#define GEN5_GENERAL_REGISTER_FILE 1 |
#define GEN5_MESSAGE_REGISTER_FILE 2 |
#define GEN5_IMMEDIATE_VALUE 3 |
#define GEN5_REGISTER_TYPE_UD 0 |
#define GEN5_REGISTER_TYPE_D 1 |
#define GEN5_REGISTER_TYPE_UW 2 |
#define GEN5_REGISTER_TYPE_W 3 |
#define GEN5_REGISTER_TYPE_UB 4 |
#define GEN5_REGISTER_TYPE_B 5 |
#define GEN5_REGISTER_TYPE_VF 5 /* packed float vector, immediates only? */ |
#define GEN5_REGISTER_TYPE_HF 6 |
#define GEN5_REGISTER_TYPE_V 6 /* packed int vector, immediates only, uword dest only */ |
#define GEN5_REGISTER_TYPE_F 7 |
#define GEN5_ARF_NULL 0x00 |
#define GEN5_ARF_ADDRESS 0x10 |
#define GEN5_ARF_ACCUMULATOR 0x20 |
#define GEN5_ARF_FLAG 0x30 |
#define GEN5_ARF_MASK 0x40 |
#define GEN5_ARF_MASK_STACK 0x50 |
#define GEN5_ARF_MASK_STACK_DEPTH 0x60 |
#define GEN5_ARF_STATE 0x70 |
#define GEN5_ARF_CONTROL 0x80 |
#define GEN5_ARF_NOTIFICATION_COUNT 0x90 |
#define GEN5_ARF_IP 0xA0 |
#define GEN5_AMASK 0 |
#define GEN5_IMASK 1 |
#define GEN5_LMASK 2 |
#define GEN5_CMASK 3 |
#define GEN5_THREAD_NORMAL 0 |
#define GEN5_THREAD_ATOMIC 1 |
#define GEN5_THREAD_SWITCH 2 |
#define GEN5_VERTICAL_STRIDE_0 0 |
#define GEN5_VERTICAL_STRIDE_1 1 |
#define GEN5_VERTICAL_STRIDE_2 2 |
#define GEN5_VERTICAL_STRIDE_4 3 |
#define GEN5_VERTICAL_STRIDE_8 4 |
#define GEN5_VERTICAL_STRIDE_16 5 |
#define GEN5_VERTICAL_STRIDE_32 6 |
#define GEN5_VERTICAL_STRIDE_64 7 |
#define GEN5_VERTICAL_STRIDE_128 8 |
#define GEN5_VERTICAL_STRIDE_256 9 |
#define GEN5_VERTICAL_STRIDE_ONE_DIMENSIONAL 0xF |
#define GEN5_WIDTH_1 0 |
#define GEN5_WIDTH_2 1 |
#define GEN5_WIDTH_4 2 |
#define GEN5_WIDTH_8 3 |
#define GEN5_WIDTH_16 4 |
#define GEN5_STATELESS_BUFFER_BOUNDARY_1K 0 |
#define GEN5_STATELESS_BUFFER_BOUNDARY_2K 1 |
#define GEN5_STATELESS_BUFFER_BOUNDARY_4K 2 |
#define GEN5_STATELESS_BUFFER_BOUNDARY_8K 3 |
#define GEN5_STATELESS_BUFFER_BOUNDARY_16K 4 |
#define GEN5_STATELESS_BUFFER_BOUNDARY_32K 5 |
#define GEN5_STATELESS_BUFFER_BOUNDARY_64K 6 |
#define GEN5_STATELESS_BUFFER_BOUNDARY_128K 7 |
#define GEN5_STATELESS_BUFFER_BOUNDARY_256K 8 |
#define GEN5_STATELESS_BUFFER_BOUNDARY_512K 9 |
#define GEN5_STATELESS_BUFFER_BOUNDARY_1M 10 |
#define GEN5_STATELESS_BUFFER_BOUNDARY_2M 11 |
#define GEN5_POLYGON_FACING_FRONT 0 |
#define GEN5_POLYGON_FACING_BACK 1 |
#define GEN5_MESSAGE_TARGET_NULL 0 |
#define GEN5_MESSAGE_TARGET_MATH 1 |
#define GEN5_MESSAGE_TARGET_SAMPLER 2 |
#define GEN5_MESSAGE_TARGET_GATEWAY 3 |
#define GEN5_MESSAGE_TARGET_DATAPORT_READ 4 |
#define GEN5_MESSAGE_TARGET_DATAPORT_WRITE 5 |
#define GEN5_MESSAGE_TARGET_URB 6 |
#define GEN5_MESSAGE_TARGET_THREAD_SPAWNER 7 |
#define GEN5_SAMPLER_RETURN_FORMAT_FLOAT32 0 |
#define GEN5_SAMPLER_RETURN_FORMAT_UINT32 2 |
#define GEN5_SAMPLER_RETURN_FORMAT_SINT32 3 |
#define GEN5_SAMPLER_MESSAGE_SIMD8_SAMPLE 0 |
#define GEN5_SAMPLER_MESSAGE_SIMD16_SAMPLE 0 |
#define GEN5_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS 0 |
#define GEN5_SAMPLER_MESSAGE_SIMD8_KILLPIX 1 |
#define GEN5_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_LOD 1 |
#define GEN5_SAMPLER_MESSAGE_SIMD16_SAMPLE_LOD 1 |
#define GEN5_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_GRADIENTS 2 |
#define GEN5_SAMPLER_MESSAGE_SIMD8_SAMPLE_GRADIENTS 2 |
#define GEN5_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_COMPARE 0 |
#define GEN5_SAMPLER_MESSAGE_SIMD16_SAMPLE_COMPARE 2 |
#define GEN5_SAMPLER_MESSAGE_SIMD4X2_RESINFO 2 |
#define GEN5_SAMPLER_MESSAGE_SIMD8_RESINFO 2 |
#define GEN5_SAMPLER_MESSAGE_SIMD16_RESINFO 2 |
#define GEN5_SAMPLER_MESSAGE_SIMD4X2_LD 3 |
#define GEN5_SAMPLER_MESSAGE_SIMD8_LD 3 |
#define GEN5_SAMPLER_MESSAGE_SIMD16_LD 3 |
#define GEN5_DATAPORT_OWORD_BLOCK_1_OWORDLOW 0 |
#define GEN5_DATAPORT_OWORD_BLOCK_1_OWORDHIGH 1 |
#define GEN5_DATAPORT_OWORD_BLOCK_2_OWORDS 2 |
#define GEN5_DATAPORT_OWORD_BLOCK_4_OWORDS 3 |
#define GEN5_DATAPORT_OWORD_BLOCK_8_OWORDS 4 |
#define GEN5_DATAPORT_OWORD_DUAL_BLOCK_1OWORD 0 |
#define GEN5_DATAPORT_OWORD_DUAL_BLOCK_4OWORDS 2 |
#define GEN5_DATAPORT_DWORD_SCATTERED_BLOCK_8DWORDS 2 |
#define GEN5_DATAPORT_DWORD_SCATTERED_BLOCK_16DWORDS 3 |
#define GEN5_DATAPORT_READ_MESSAGE_OWORD_BLOCK_READ 0 |
#define GEN5_DATAPORT_READ_MESSAGE_OWORD_DUAL_BLOCK_READ 1 |
#define GEN5_DATAPORT_READ_MESSAGE_DWORD_BLOCK_READ 2 |
#define GEN5_DATAPORT_READ_MESSAGE_DWORD_SCATTERED_READ 3 |
#define GEN5_DATAPORT_READ_TARGET_DATA_CACHE 0 |
#define GEN5_DATAPORT_READ_TARGET_RENDER_CACHE 1 |
#define GEN5_DATAPORT_READ_TARGET_SAMPLER_CACHE 2 |
#define GEN5_DATAPORT_RENDER_TARGET_WRITE_SIMD16_SINGLE_SOURCE 0 |
#define GEN5_DATAPORT_RENDER_TARGET_WRITE_SIMD16_SINGLE_SOURCE_REPLICATED 1 |
#define GEN5_DATAPORT_RENDER_TARGET_WRITE_SIMD8_DUAL_SOURCE_SUBSPAN01 2 |
#define GEN5_DATAPORT_RENDER_TARGET_WRITE_SIMD8_DUAL_SOURCE_SUBSPAN23 3 |
#define GEN5_DATAPORT_RENDER_TARGET_WRITE_SIMD8_SINGLE_SOURCE_SUBSPAN01 4 |
#define GEN5_DATAPORT_WRITE_MESSAGE_OWORD_BLOCK_WRITE 0 |
#define GEN5_DATAPORT_WRITE_MESSAGE_OWORD_DUAL_BLOCK_WRITE 1 |
#define GEN5_DATAPORT_WRITE_MESSAGE_DWORD_BLOCK_WRITE 2 |
#define GEN5_DATAPORT_WRITE_MESSAGE_DWORD_SCATTERED_WRITE 3 |
#define GEN5_DATAPORT_WRITE_MESSAGE_RENDER_TARGET_WRITE 4 |
#define GEN5_DATAPORT_WRITE_MESSAGE_STREAMED_VERTEX_BUFFER_WRITE 5 |
#define GEN5_DATAPORT_WRITE_MESSAGE_FLUSH_RENDER_CACHE 7 |
#define GEN5_MATH_FUNCTION_INV 1 |
#define GEN5_MATH_FUNCTION_LOG 2 |
#define GEN5_MATH_FUNCTION_EXP 3 |
#define GEN5_MATH_FUNCTION_SQRT 4 |
#define GEN5_MATH_FUNCTION_RSQ 5 |
#define GEN5_MATH_FUNCTION_SIN 6 /* was 7 */ |
#define GEN5_MATH_FUNCTION_COS 7 /* was 8 */ |
#define GEN5_MATH_FUNCTION_SINCOS 8 /* was 6 */ |
#define GEN5_MATH_FUNCTION_TAN 9 |
#define GEN5_MATH_FUNCTION_POW 10 |
#define GEN5_MATH_FUNCTION_INT_DIV_QUOTIENT_AND_REMAINDER 11 |
#define GEN5_MATH_FUNCTION_INT_DIV_QUOTIENT 12 |
#define GEN5_MATH_FUNCTION_INT_DIV_REMAINDER 13 |
#define GEN5_MATH_INTEGER_UNSIGNED 0 |
#define GEN5_MATH_INTEGER_SIGNED 1 |
#define GEN5_MATH_PRECISION_FULL 0 |
#define GEN5_MATH_PRECISION_PARTIAL 1 |
#define GEN5_MATH_SATURATE_NONE 0 |
#define GEN5_MATH_SATURATE_SATURATE 1 |
#define GEN5_MATH_DATA_VECTOR 0 |
#define GEN5_MATH_DATA_SCALAR 1 |
#define GEN5_URB_OPCODE_WRITE 0 |
#define GEN5_URB_SWIZZLE_NONE 0 |
#define GEN5_URB_SWIZZLE_INTERLEAVE 1 |
#define GEN5_URB_SWIZZLE_TRANSPOSE 2 |
#define GEN5_SCRATCH_SPACE_SIZE_1K 0 |
#define GEN5_SCRATCH_SPACE_SIZE_2K 1 |
#define GEN5_SCRATCH_SPACE_SIZE_4K 2 |
#define GEN5_SCRATCH_SPACE_SIZE_8K 3 |
#define GEN5_SCRATCH_SPACE_SIZE_16K 4 |
#define GEN5_SCRATCH_SPACE_SIZE_32K 5 |
#define GEN5_SCRATCH_SPACE_SIZE_64K 6 |
#define GEN5_SCRATCH_SPACE_SIZE_128K 7 |
#define GEN5_SCRATCH_SPACE_SIZE_256K 8 |
#define GEN5_SCRATCH_SPACE_SIZE_512K 9 |
#define GEN5_SCRATCH_SPACE_SIZE_1M 10 |
#define GEN5_SCRATCH_SPACE_SIZE_2M 11 |
#define CMD_URB_FENCE 0x6000 |
#define CMD_CONST_BUFFER_STATE 0x6001 |
#define CMD_CONST_BUFFER 0x6002 |
#define CMD_STATE_BASE_ADDRESS 0x6101 |
#define CMD_STATE_INSN_POINTER 0x6102 |
#define CMD_PIPELINE_SELECT 0x6104 |
#define CMD_PIPELINED_STATE_POINTERS 0x7800 |
#define CMD_BINDING_TABLE_PTRS 0x7801 |
#define CMD_VERTEX_BUFFER 0x7808 |
#define CMD_VERTEX_ELEMENT 0x7809 |
#define CMD_INDEX_BUFFER 0x780a |
#define CMD_VF_STATISTICS 0x780b |
#define CMD_DRAW_RECT 0x7900 |
#define CMD_BLEND_CONSTANT_COLOR 0x7901 |
#define CMD_CHROMA_KEY 0x7904 |
#define CMD_DEPTH_BUFFER 0x7905 |
#define CMD_POLY_STIPPLE_OFFSET 0x7906 |
#define CMD_POLY_STIPPLE_PATTERN 0x7907 |
#define CMD_LINE_STIPPLE_PATTERN 0x7908 |
#define CMD_GLOBAL_DEPTH_OFFSET_CLAMP 0x7908 |
#define CMD_PIPE_CONTROL 0x7a00 |
#define CMD_3D_PRIM 0x7b00 |
#define CMD_MI_FLUSH 0x0200 |
/* Various values from the R0 vertex header: |
*/ |
#define R02_PRIM_END 0x1 |
#define R02_PRIM_START 0x2 |
/* media pipeline */ |
#define GEN5_VFE_MODE_GENERIC 0x0 |
#define GEN5_VFE_MODE_VLD_MPEG2 0x1 |
#define GEN5_VFE_MODE_IS 0x2 |
#define GEN5_VFE_MODE_AVC_MC 0x4 |
#define GEN5_VFE_MODE_AVC_IT 0x7 |
#define GEN5_VFE_MODE_VC1_IT 0xB |
#define GEN5_VFE_DEBUG_COUNTER_FREE 0 |
#define GEN5_VFE_DEBUG_COUNTER_FROZEN 1 |
#define GEN5_VFE_DEBUG_COUNTER_ONCE 2 |
#define GEN5_VFE_DEBUG_COUNTER_ALWAYS 3 |
/* VLD_STATE */ |
#define GEN5_MPEG_TOP_FIELD 1 |
#define GEN5_MPEG_BOTTOM_FIELD 2 |
#define GEN5_MPEG_FRAME 3 |
#define GEN5_MPEG_QSCALE_LINEAR 0 |
#define GEN5_MPEG_QSCALE_NONLINEAR 1 |
#define GEN5_MPEG_ZIGZAG_SCAN 0 |
#define GEN5_MPEG_ALTER_VERTICAL_SCAN 1 |
#define GEN5_MPEG_I_PICTURE 1 |
#define GEN5_MPEG_P_PICTURE 2 |
#define GEN5_MPEG_B_PICTURE 3 |
/* Command packets: |
*/ |
struct header |
{ |
unsigned int length:16; |
unsigned int opcode:16; |
}; |
union header_union |
{ |
struct header bits; |
unsigned int dword; |
}; |
struct gen5_3d_control |
{ |
struct |
{ |
unsigned int length:8; |
unsigned int notify_enable:1; |
unsigned int pad:3; |
unsigned int wc_flush_enable:1; |
unsigned int depth_stall_enable:1; |
unsigned int operation:2; |
unsigned int opcode:16; |
} header; |
struct |
{ |
unsigned int pad:2; |
unsigned int dest_addr_type:1; |
unsigned int dest_addr:29; |
} dest; |
unsigned int dword2; |
unsigned int dword3; |
}; |
struct gen5_3d_primitive |
{ |
struct |
{ |
unsigned int length:8; |
unsigned int pad:2; |
unsigned int topology:5; |
unsigned int indexed:1; |
unsigned int opcode:16; |
} header; |
unsigned int verts_per_instance; |
unsigned int start_vert_location; |
unsigned int instance_count; |
unsigned int start_instance_location; |
unsigned int base_vert_location; |
}; |
/* These seem to be passed around as function args, so it works out |
* better to keep them as #defines: |
*/ |
#define GEN5_FLUSH_READ_CACHE 0x1 |
#define GEN5_FLUSH_STATE_CACHE 0x2 |
#define GEN5_INHIBIT_FLUSH_RENDER_CACHE 0x4 |
#define GEN5_FLUSH_SNAPSHOT_COUNTERS 0x8 |
struct gen5_mi_flush |
{ |
unsigned int flags:4; |
unsigned int pad:12; |
unsigned int opcode:16; |
}; |
struct gen5_vf_statistics |
{ |
unsigned int statistics_enable:1; |
unsigned int pad:15; |
unsigned int opcode:16; |
}; |
struct gen5_binding_table_pointers |
{ |
struct header header; |
unsigned int vs; |
unsigned int gs; |
unsigned int clp; |
unsigned int sf; |
unsigned int wm; |
}; |
struct gen5_blend_constant_color |
{ |
struct header header; |
float blend_constant_color[4]; |
}; |
struct gen5_depthbuffer |
{ |
union header_union header; |
union { |
struct { |
unsigned int pitch:18; |
unsigned int format:3; |
unsigned int pad:4; |
unsigned int depth_offset_disable:1; |
unsigned int tile_walk:1; |
unsigned int tiled_surface:1; |
unsigned int pad2:1; |
unsigned int surface_type:3; |
} bits; |
unsigned int dword; |
} dword1; |
unsigned int dword2_base_addr; |
union { |
struct { |
unsigned int pad:1; |
unsigned int mipmap_layout:1; |
unsigned int lod:4; |
unsigned int width:13; |
unsigned int height:13; |
} bits; |
unsigned int dword; |
} dword3; |
union { |
struct { |
unsigned int pad:12; |
unsigned int min_array_element:9; |
unsigned int depth:11; |
} bits; |
unsigned int dword; |
} dword4; |
}; |
struct gen5_drawrect |
{ |
struct header header; |
unsigned int xmin:16; |
unsigned int ymin:16; |
unsigned int xmax:16; |
unsigned int ymax:16; |
unsigned int xorg:16; |
unsigned int yorg:16; |
}; |
struct gen5_global_depth_offset_clamp |
{ |
struct header header; |
float depth_offset_clamp; |
}; |
struct gen5_indexbuffer |
{ |
union { |
struct |
{ |
unsigned int length:8; |
unsigned int index_format:2; |
unsigned int cut_index_enable:1; |
unsigned int pad:5; |
unsigned int opcode:16; |
} bits; |
unsigned int dword; |
} header; |
unsigned int buffer_start; |
unsigned int buffer_end; |
}; |
struct gen5_line_stipple |
{ |
struct header header; |
struct |
{ |
unsigned int pattern:16; |
unsigned int pad:16; |
} bits0; |
struct |
{ |
unsigned int repeat_count:9; |
unsigned int pad:7; |
unsigned int inverse_repeat_count:16; |
} bits1; |
}; |
struct gen5_pipelined_state_pointers |
{ |
struct header header; |
struct { |
unsigned int pad:5; |
unsigned int offset:27; |
} vs; |
struct |
{ |
unsigned int enable:1; |
unsigned int pad:4; |
unsigned int offset:27; |
} gs; |
struct |
{ |
unsigned int enable:1; |
unsigned int pad:4; |
unsigned int offset:27; |
} clp; |
struct |
{ |
unsigned int pad:5; |
unsigned int offset:27; |
} sf; |
struct |
{ |
unsigned int pad:5; |
unsigned int offset:27; |
} wm; |
struct |
{ |
unsigned int pad:5; |
unsigned int offset:27; /* KW: check me! */ |
} cc; |
}; |
struct gen5_polygon_stipple_offset |
{ |
struct header header; |
struct { |
unsigned int y_offset:5; |
unsigned int pad:3; |
unsigned int x_offset:5; |
unsigned int pad0:19; |
} bits0; |
}; |
struct gen5_polygon_stipple |
{ |
struct header header; |
unsigned int stipple[32]; |
}; |
struct gen5_pipeline_select |
{ |
struct |
{ |
unsigned int pipeline_select:1; |
unsigned int pad:15; |
unsigned int opcode:16; |
} header; |
}; |
struct gen5_pipe_control |
{ |
struct |
{ |
unsigned int length:8; |
unsigned int notify_enable:1; |
unsigned int pad:2; |
unsigned int instruction_state_cache_flush_enable:1; |
unsigned int write_cache_flush_enable:1; |
unsigned int depth_stall_enable:1; |
unsigned int post_sync_operation:2; |
unsigned int opcode:16; |
} header; |
struct |
{ |
unsigned int pad:2; |
unsigned int dest_addr_type:1; |
unsigned int dest_addr:29; |
} bits1; |
unsigned int data0; |
unsigned int data1; |
}; |
struct gen5_urb_fence |
{ |
struct |
{ |
unsigned int length:8; |
unsigned int vs_realloc:1; |
unsigned int gs_realloc:1; |
unsigned int clp_realloc:1; |
unsigned int sf_realloc:1; |
unsigned int vfe_realloc:1; |
unsigned int cs_realloc:1; |
unsigned int pad:2; |
unsigned int opcode:16; |
} header; |
struct |
{ |
unsigned int vs_fence:10; |
unsigned int gs_fence:10; |
unsigned int clp_fence:10; |
unsigned int pad:2; |
} bits0; |
struct |
{ |
unsigned int sf_fence:10; |
unsigned int vf_fence:10; |
unsigned int cs_fence:10; |
unsigned int pad:2; |
} bits1; |
}; |
struct gen5_constant_buffer_state /* previously gen5_command_streamer */ |
{ |
struct header header; |
struct |
{ |
unsigned int nr_urb_entries:3; |
unsigned int pad:1; |
unsigned int urb_entry_size:5; |
unsigned int pad0:23; |
} bits0; |
}; |
struct gen5_constant_buffer |
{ |
struct |
{ |
unsigned int length:8; |
unsigned int valid:1; |
unsigned int pad:7; |
unsigned int opcode:16; |
} header; |
struct |
{ |
unsigned int buffer_length:6; |
unsigned int buffer_address:26; |
} bits0; |
}; |
struct gen5_state_base_address |
{ |
struct header header; |
struct |
{ |
unsigned int modify_enable:1; |
unsigned int pad:4; |
unsigned int general_state_address:27; |
} bits0; |
struct |
{ |
unsigned int modify_enable:1; |
unsigned int pad:4; |
unsigned int surface_state_address:27; |
} bits1; |
struct |
{ |
unsigned int modify_enable:1; |
unsigned int pad:4; |
unsigned int indirect_object_state_address:27; |
} bits2; |
struct |
{ |
unsigned int modify_enable:1; |
unsigned int pad:11; |
unsigned int general_state_upper_bound:20; |
} bits3; |
struct |
{ |
unsigned int modify_enable:1; |
unsigned int pad:11; |
unsigned int indirect_object_state_upper_bound:20; |
} bits4; |
}; |
struct gen5_state_prefetch |
{ |
struct header header; |
struct |
{ |
unsigned int prefetch_count:3; |
unsigned int pad:3; |
unsigned int prefetch_pointer:26; |
} bits0; |
}; |
struct gen5_system_instruction_pointer |
{ |
struct header header; |
struct |
{ |
unsigned int pad:4; |
unsigned int system_instruction_pointer:28; |
} bits0; |
}; |
/* State structs for the various fixed function units: |
*/ |
struct thread0 |
{ |
unsigned int pad0:1; |
unsigned int grf_reg_count:3; |
unsigned int pad1:2; |
unsigned int kernel_start_pointer:26; |
}; |
struct thread1 |
{ |
unsigned int ext_halt_exception_enable:1; |
unsigned int sw_exception_enable:1; |
unsigned int mask_stack_exception_enable:1; |
unsigned int timeout_exception_enable:1; |
unsigned int illegal_op_exception_enable:1; |
unsigned int pad0:3; |
unsigned int depth_coef_urb_read_offset:6; /* WM only */ |
unsigned int pad1:2; |
unsigned int floating_point_mode:1; |
unsigned int thread_priority:1; |
unsigned int binding_table_entry_count:8; |
unsigned int pad3:5; |
unsigned int single_program_flow:1; |
}; |
struct thread2 |
{ |
unsigned int per_thread_scratch_space:4; |
unsigned int pad0:6; |
unsigned int scratch_space_base_pointer:22; |
}; |
struct thread3 |
{ |
unsigned int dispatch_grf_start_reg:4; |
unsigned int urb_entry_read_offset:6; |
unsigned int pad0:1; |
unsigned int urb_entry_read_length:6; |
unsigned int pad1:1; |
unsigned int const_urb_entry_read_offset:6; |
unsigned int pad2:1; |
unsigned int const_urb_entry_read_length:6; |
unsigned int pad3:1; |
}; |
struct gen5_clip_unit_state |
{ |
struct thread0 thread0; |
struct thread1 thread1; |
struct thread2 thread2; |
struct thread3 thread3; |
struct |
{ |
unsigned int pad0:9; |
unsigned int gs_output_stats:1; /* not always */ |
unsigned int stats_enable:1; |
unsigned int nr_urb_entries:7; |
unsigned int pad1:1; |
unsigned int urb_entry_allocation_size:5; |
unsigned int pad2:1; |
unsigned int max_threads:6; /* may be less */ |
unsigned int pad3:1; |
} thread4; |
struct |
{ |
unsigned int pad0:13; |
unsigned int clip_mode:3; |
unsigned int userclip_enable_flags:8; |
unsigned int userclip_must_clip:1; |
unsigned int pad1:1; |
unsigned int guard_band_enable:1; |
unsigned int viewport_z_clip_enable:1; |
unsigned int viewport_xy_clip_enable:1; |
unsigned int vertex_position_space:1; |
unsigned int api_mode:1; |
unsigned int pad2:1; |
} clip5; |
struct |
{ |
unsigned int pad0:5; |
unsigned int clipper_viewport_state_ptr:27; |
} clip6; |
float viewport_xmin; |
float viewport_xmax; |
float viewport_ymin; |
float viewport_ymax; |
}; |
struct gen5_cc_unit_state |
{ |
struct |
{ |
unsigned int pad0:3; |
unsigned int bf_stencil_pass_depth_pass_op:3; |
unsigned int bf_stencil_pass_depth_fail_op:3; |
unsigned int bf_stencil_fail_op:3; |
unsigned int bf_stencil_func:3; |
unsigned int bf_stencil_enable:1; |
unsigned int pad1:2; |
unsigned int stencil_write_enable:1; |
unsigned int stencil_pass_depth_pass_op:3; |
unsigned int stencil_pass_depth_fail_op:3; |
unsigned int stencil_fail_op:3; |
unsigned int stencil_func:3; |
unsigned int stencil_enable:1; |
} cc0; |
struct |
{ |
unsigned int bf_stencil_ref:8; |
unsigned int stencil_write_mask:8; |
unsigned int stencil_test_mask:8; |
unsigned int stencil_ref:8; |
} cc1; |
struct |
{ |
unsigned int logicop_enable:1; |
unsigned int pad0:10; |
unsigned int depth_write_enable:1; |
unsigned int depth_test_function:3; |
unsigned int depth_test:1; |
unsigned int bf_stencil_write_mask:8; |
unsigned int bf_stencil_test_mask:8; |
} cc2; |
struct |
{ |
unsigned int pad0:8; |
unsigned int alpha_test_func:3; |
unsigned int alpha_test:1; |
unsigned int blend_enable:1; |
unsigned int ia_blend_enable:1; |
unsigned int pad1:1; |
unsigned int alpha_test_format:1; |
unsigned int pad2:16; |
} cc3; |
struct |
{ |
unsigned int pad0:5; |
unsigned int cc_viewport_state_offset:27; |
} cc4; |
struct |
{ |
unsigned int pad0:2; |
unsigned int ia_dest_blend_factor:5; |
unsigned int ia_src_blend_factor:5; |
unsigned int ia_blend_function:3; |
unsigned int statistics_enable:1; |
unsigned int logicop_func:4; |
unsigned int pad1:11; |
unsigned int dither_enable:1; |
} cc5; |
struct |
{ |
unsigned int clamp_post_alpha_blend:1; |
unsigned int clamp_pre_alpha_blend:1; |
unsigned int clamp_range:2; |
unsigned int pad0:11; |
unsigned int y_dither_offset:2; |
unsigned int x_dither_offset:2; |
unsigned int dest_blend_factor:5; |
unsigned int src_blend_factor:5; |
unsigned int blend_function:3; |
} cc6; |
struct { |
union { |
float f; |
unsigned char ub[4]; |
} alpha_ref; |
} cc7; |
}; |
struct gen5_sf_unit_state |
{ |
struct thread0 thread0; |
struct { |
unsigned int pad0:7; |
unsigned int sw_exception_enable:1; |
unsigned int pad1:3; |
unsigned int mask_stack_exception_enable:1; |
unsigned int pad2:1; |
unsigned int illegal_op_exception_enable:1; |
unsigned int pad3:2; |
unsigned int floating_point_mode:1; |
unsigned int thread_priority:1; |
unsigned int binding_table_entry_count:8; |
unsigned int pad4:5; |
unsigned int single_program_flow:1; |
} sf1; |
struct thread2 thread2; |
struct thread3 thread3; |
struct |
{ |
unsigned int pad0:10; |
unsigned int stats_enable:1; |
unsigned int nr_urb_entries:7; |
unsigned int pad1:1; |
unsigned int urb_entry_allocation_size:5; |
unsigned int pad2:1; |
unsigned int max_threads:6; |
unsigned int pad3:1; |
} thread4; |
struct |
{ |
unsigned int front_winding:1; |
unsigned int viewport_transform:1; |
unsigned int pad0:3; |
unsigned int sf_viewport_state_offset:27; |
} sf5; |
struct |
{ |
unsigned int pad0:9; |
unsigned int dest_org_vbias:4; |
unsigned int dest_org_hbias:4; |
unsigned int scissor:1; |
unsigned int disable_2x2_trifilter:1; |
unsigned int disable_zero_pix_trifilter:1; |
unsigned int point_rast_rule:2; |
unsigned int line_endcap_aa_region_width:2; |
unsigned int line_width:4; |
unsigned int fast_scissor_disable:1; |
unsigned int cull_mode:2; |
unsigned int aa_enable:1; |
} sf6; |
struct |
{ |
unsigned int point_size:11; |
unsigned int use_point_size_state:1; |
unsigned int subpixel_precision:1; |
unsigned int sprite_point:1; |
unsigned int pad0:11; |
unsigned int trifan_pv:2; |
unsigned int linestrip_pv:2; |
unsigned int tristrip_pv:2; |
unsigned int line_last_pixel_enable:1; |
} sf7; |
}; |
struct gen5_gs_unit_state |
{ |
struct thread0 thread0; |
struct thread1 thread1; |
struct thread2 thread2; |
struct thread3 thread3; |
struct |
{ |
unsigned int pad0:10; |
unsigned int stats_enable:1; |
unsigned int nr_urb_entries:7; |
unsigned int pad1:1; |
unsigned int urb_entry_allocation_size:5; |
unsigned int pad2:1; |
unsigned int max_threads:1; |
unsigned int pad3:6; |
} thread4; |
struct |
{ |
unsigned int sampler_count:3; |
unsigned int pad0:2; |
unsigned int sampler_state_pointer:27; |
} gs5; |
struct |
{ |
unsigned int max_vp_index:4; |
unsigned int pad0:26; |
unsigned int reorder_enable:1; |
unsigned int pad1:1; |
} gs6; |
}; |
struct gen5_vs_unit_state |
{ |
struct thread0 thread0; |
struct thread1 thread1; |
struct thread2 thread2; |
struct thread3 thread3; |
struct |
{ |
unsigned int pad0:10; |
unsigned int stats_enable:1; |
unsigned int nr_urb_entries:7; |
unsigned int pad1:1; |
unsigned int urb_entry_allocation_size:5; |
unsigned int pad2:1; |
unsigned int max_threads:4; |
unsigned int pad3:3; |
} thread4; |
struct |
{ |
unsigned int sampler_count:3; |
unsigned int pad0:2; |
unsigned int sampler_state_pointer:27; |
} vs5; |
struct |
{ |
unsigned int vs_enable:1; |
unsigned int vert_cache_disable:1; |
unsigned int pad0:30; |
} vs6; |
}; |
struct gen5_wm_unit_state |
{ |
struct thread0 thread0; |
struct thread1 thread1; |
struct thread2 thread2; |
struct thread3 thread3; |
struct { |
unsigned int stats_enable:1; |
unsigned int pad0:1; |
unsigned int sampler_count:3; |
unsigned int sampler_state_pointer:27; |
} wm4; |
struct |
{ |
unsigned int enable_8_pix:1; |
unsigned int enable_16_pix:1; |
unsigned int enable_32_pix:1; |
unsigned int pad0:7; |
unsigned int legacy_global_depth_bias:1; |
unsigned int line_stipple:1; |
unsigned int depth_offset:1; |
unsigned int polygon_stipple:1; |
unsigned int line_aa_region_width:2; |
unsigned int line_endcap_aa_region_width:2; |
unsigned int early_depth_test:1; |
unsigned int thread_dispatch_enable:1; |
unsigned int program_uses_depth:1; |
unsigned int program_computes_depth:1; |
unsigned int program_uses_killpixel:1; |
unsigned int legacy_line_rast: 1; |
unsigned int transposed_urb_read:1; |
unsigned int max_threads:7; |
} wm5; |
float global_depth_offset_constant; |
float global_depth_offset_scale; |
struct { |
unsigned int pad0:1; |
unsigned int grf_reg_count_1:3; |
unsigned int pad1:2; |
unsigned int kernel_start_pointer_1:26; |
} wm8; |
struct { |
unsigned int pad0:1; |
unsigned int grf_reg_count_2:3; |
unsigned int pad1:2; |
unsigned int kernel_start_pointer_2:26; |
} wm9; |
struct { |
unsigned int pad0:1; |
unsigned int grf_reg_count_3:3; |
unsigned int pad1:2; |
unsigned int kernel_start_pointer_3:26; |
} wm10; |
}; |
struct gen5_wm_unit_state_padded { |
struct gen5_wm_unit_state state; |
char pad[64 - sizeof(struct gen5_wm_unit_state)]; |
}; |
/* The hardware supports two different modes for border color. The |
* default (OpenGL) mode uses floating-point color channels, while the |
* legacy mode uses 4 bytes. |
* |
* More significantly, the legacy mode respects the components of the |
* border color for channels not present in the source, (whereas the |
* default mode will ignore the border color's alpha channel and use |
* alpha==1 for an RGB source, for example). |
* |
* The legacy mode matches the semantics specified by the Render |
* extension. |
*/ |
struct gen5_sampler_default_border_color { |
float color[4]; |
}; |
struct gen5_sampler_legacy_border_color { |
uint8_t color[4]; |
}; |
struct gen5_sampler_state { |
struct { |
unsigned int shadow_function:3; |
unsigned int lod_bias:11; |
unsigned int min_filter:3; |
unsigned int mag_filter:3; |
unsigned int mip_filter:2; |
unsigned int base_level:5; |
unsigned int pad:1; |
unsigned int lod_preclamp:1; |
unsigned int border_color_mode:1; |
unsigned int pad0:1; |
unsigned int disable:1; |
} ss0; |
struct { |
unsigned int r_wrap_mode:3; |
unsigned int t_wrap_mode:3; |
unsigned int s_wrap_mode:3; |
unsigned int pad:3; |
unsigned int max_lod:10; |
unsigned int min_lod:10; |
} ss1; |
struct { |
unsigned int pad:5; |
unsigned int border_color_pointer:27; |
} ss2; |
struct { |
uint32_t pad:13; |
uint32_t address_round:6; |
uint32_t max_aniso:3; |
uint32_t chroma_key_mode:1; |
uint32_t chroma_key_index:2; |
uint32_t chroma_key_enable:1; |
uint32_t mbz:6; |
} ss3; |
}; |
struct gen5_clipper_viewport |
{ |
float xmin; |
float xmax; |
float ymin; |
float ymax; |
}; |
struct gen5_cc_viewport |
{ |
float min_depth; |
float max_depth; |
}; |
struct gen5_sf_viewport |
{ |
struct { |
float m00; |
float m11; |
float m22; |
float m30; |
float m31; |
float m32; |
} viewport; |
struct { |
short xmin; |
short ymin; |
short xmax; |
short ymax; |
} scissor; |
}; |
/* Documented in the subsystem/shared-functions/sampler chapter... |
*/ |
struct gen5_surface_state |
{ |
struct { |
unsigned int cube_pos_z:1; |
unsigned int cube_neg_z:1; |
unsigned int cube_pos_y:1; |
unsigned int cube_neg_y:1; |
unsigned int cube_pos_x:1; |
unsigned int cube_neg_x:1; |
unsigned int pad:3; |
unsigned int render_cache_read_mode:1; |
unsigned int mipmap_layout_mode:1; |
unsigned int vert_line_stride_ofs:1; |
unsigned int vert_line_stride:1; |
unsigned int color_blend:1; |
unsigned int writedisable_blue:1; |
unsigned int writedisable_green:1; |
unsigned int writedisable_red:1; |
unsigned int writedisable_alpha:1; |
unsigned int surface_format:9; |
unsigned int data_return_format:1; |
unsigned int pad0:1; |
unsigned int surface_type:3; |
} ss0; |
struct { |
unsigned int base_addr; |
} ss1; |
struct { |
unsigned int render_target_rotation:2; |
unsigned int mip_count:4; |
unsigned int width:13; |
unsigned int height:13; |
} ss2; |
struct { |
unsigned int tile_walk:1; |
unsigned int tiled_surface:1; |
unsigned int pad:1; |
unsigned int pitch:18; |
unsigned int depth:11; |
} ss3; |
struct { |
unsigned int pad:19; |
unsigned int min_array_elt:9; |
unsigned int min_lod:4; |
} ss4; |
struct { |
unsigned int pad:20; |
unsigned int y_offset:4; |
unsigned int pad2:1; |
unsigned int x_offset:7; |
} ss5; |
}; |
/* Surface state DW0 */ |
#define GEN5_SURFACE_RC_READ_WRITE (1 << 8) |
#define GEN5_SURFACE_MIPLAYOUT_SHIFT 10 |
#define GEN5_SURFACE_MIPMAPLAYOUT_BELOW 0 |
#define GEN5_SURFACE_MIPMAPLAYOUT_RIGHT 1 |
#define GEN5_SURFACE_CUBEFACE_ENABLES 0x3f |
#define GEN5_SURFACE_BLEND_ENABLED (1 << 13) |
#define GEN5_SURFACE_WRITEDISABLE_B_SHIFT 14 |
#define GEN5_SURFACE_WRITEDISABLE_G_SHIFT 15 |
#define GEN5_SURFACE_WRITEDISABLE_R_SHIFT 16 |
#define GEN5_SURFACE_WRITEDISABLE_A_SHIFT 17 |
#define GEN5_SURFACE_FORMAT_SHIFT 18 |
#define GEN5_SURFACE_FORMAT_MASK INTEL_MASK(26, 18) |
#define GEN5_SURFACE_TYPE_SHIFT 29 |
#define GEN5_SURFACE_TYPE_MASK GEN5_MASK(31, 29) |
#define GEN5_SURFACE_1D 0 |
#define GEN5_SURFACE_2D 1 |
#define GEN5_SURFACE_3D 2 |
#define GEN5_SURFACE_CUBE 3 |
#define GEN5_SURFACE_BUFFER 4 |
#define GEN5_SURFACE_NULL 7 |
/* Surface state DW2 */ |
#define GEN5_SURFACE_HEIGHT_SHIFT 19 |
#define GEN5_SURFACE_HEIGHT_MASK GEN5_MASK(31, 19) |
#define GEN5_SURFACE_WIDTH_SHIFT 6 |
#define GEN5_SURFACE_WIDTH_MASK GEN5_MASK(18, 6) |
#define GEN5_SURFACE_LOD_SHIFT 2 |
#define GEN5_SURFACE_LOD_MASK GEN5_MASK(5, 2) |
/* Surface state DW3 */ |
#define GEN5_SURFACE_DEPTH_SHIFT 21 |
#define GEN5_SURFACE_DEPTH_MASK GEN5_MASK(31, 21) |
#define GEN5_SURFACE_PITCH_SHIFT 3 |
#define GEN5_SURFACE_PITCH_MASK GEN5_MASK(19, 3) |
#define GEN5_SURFACE_TILED (1 << 1) |
#define GEN5_SURFACE_TILED_Y (1 << 0) |
/* Surface state DW4 */ |
#define GEN5_SURFACE_MIN_LOD_SHIFT 28 |
#define GEN5_SURFACE_MIN_LOD_MASK GEN5_MASK(31, 28) |
/* Surface state DW5 */ |
#define GEN5_SURFACE_X_OFFSET_SHIFT 25 |
#define GEN5_SURFACE_X_OFFSET_MASK GEN5_MASK(31, 25) |
#define GEN5_SURFACE_Y_OFFSET_SHIFT 20 |
#define GEN5_SURFACE_Y_OFFSET_MASK GEN5_MASK(23, 20) |
struct gen5_vertex_buffer_state |
{ |
struct { |
unsigned int pitch:11; |
unsigned int pad:15; |
unsigned int access_type:1; |
unsigned int vb_index:5; |
} vb0; |
unsigned int start_addr; |
unsigned int max_index; |
#if 1 |
unsigned int instance_data_step_rate; /* not included for sequential/random vertices? */ |
#endif |
}; |
#define GEN5_VBP_MAX 17 |
struct gen5_vb_array_state { |
struct header header; |
struct gen5_vertex_buffer_state vb[GEN5_VBP_MAX]; |
}; |
struct gen5_vertex_element_state |
{ |
struct |
{ |
unsigned int src_offset:11; |
unsigned int pad:5; |
unsigned int src_format:9; |
unsigned int pad0:1; |
unsigned int valid:1; |
unsigned int vertex_buffer_index:5; |
} ve0; |
struct |
{ |
unsigned int dst_offset:8; |
unsigned int pad:8; |
unsigned int vfcomponent3:4; |
unsigned int vfcomponent2:4; |
unsigned int vfcomponent1:4; |
unsigned int vfcomponent0:4; |
} ve1; |
}; |
#define GEN5_VEP_MAX 18 |
struct gen5_vertex_element_packet { |
struct header header; |
struct gen5_vertex_element_state ve[GEN5_VEP_MAX]; /* note: less than _TNL_ATTRIB_MAX */ |
}; |
struct gen5_urb_immediate { |
unsigned int opcode:4; |
unsigned int offset:6; |
unsigned int swizzle_control:2; |
unsigned int pad:1; |
unsigned int allocate:1; |
unsigned int used:1; |
unsigned int complete:1; |
unsigned int response_length:4; |
unsigned int msg_length:4; |
unsigned int msg_target:4; |
unsigned int pad1:3; |
unsigned int end_of_thread:1; |
}; |
/* Instruction format for the execution units: |
*/ |
struct gen5_instruction |
{ |
struct |
{ |
unsigned int opcode:7; |
unsigned int pad:1; |
unsigned int access_mode:1; |
unsigned int mask_control:1; |
unsigned int dependency_control:2; |
unsigned int compression_control:2; |
unsigned int thread_control:2; |
unsigned int predicate_control:4; |
unsigned int predicate_inverse:1; |
unsigned int execution_size:3; |
unsigned int destreg__conditonalmod:4; /* destreg - send, conditionalmod - others */ |
unsigned int pad0:2; |
unsigned int debug_control:1; |
unsigned int saturate:1; |
} header; |
union { |
struct |
{ |
unsigned int dest_reg_file:2; |
unsigned int dest_reg_type:3; |
unsigned int src0_reg_file:2; |
unsigned int src0_reg_type:3; |
unsigned int src1_reg_file:2; |
unsigned int src1_reg_type:3; |
unsigned int pad:1; |
unsigned int dest_subreg_nr:5; |
unsigned int dest_reg_nr:8; |
unsigned int dest_horiz_stride:2; |
unsigned int dest_address_mode:1; |
} da1; |
struct |
{ |
unsigned int dest_reg_file:2; |
unsigned int dest_reg_type:3; |
unsigned int src0_reg_file:2; |
unsigned int src0_reg_type:3; |
unsigned int pad:6; |
int dest_indirect_offset:10; /* offset against the deref'd address reg */ |
unsigned int dest_subreg_nr:3; /* subnr for the address reg a0.x */ |
unsigned int dest_horiz_stride:2; |
unsigned int dest_address_mode:1; |
} ia1; |
struct |
{ |
unsigned int dest_reg_file:2; |
unsigned int dest_reg_type:3; |
unsigned int src0_reg_file:2; |
unsigned int src0_reg_type:3; |
unsigned int src1_reg_file:2; |
unsigned int src1_reg_type:3; |
unsigned int pad0:1; |
unsigned int dest_writemask:4; |
unsigned int dest_subreg_nr:1; |
unsigned int dest_reg_nr:8; |
unsigned int pad1:2; |
unsigned int dest_address_mode:1; |
} da16; |
struct |
{ |
unsigned int dest_reg_file:2; |
unsigned int dest_reg_type:3; |
unsigned int src0_reg_file:2; |
unsigned int src0_reg_type:3; |
unsigned int pad0:6; |
unsigned int dest_writemask:4; |
int dest_indirect_offset:6; |
unsigned int dest_subreg_nr:3; |
unsigned int pad1:2; |
unsigned int dest_address_mode:1; |
} ia16; |
} bits1; |
union { |
struct |
{ |
unsigned int src0_subreg_nr:5; |
unsigned int src0_reg_nr:8; |
unsigned int src0_abs:1; |
unsigned int src0_negate:1; |
unsigned int src0_address_mode:1; |
unsigned int src0_horiz_stride:2; |
unsigned int src0_width:3; |
unsigned int src0_vert_stride:4; |
unsigned int flag_reg_nr:1; |
unsigned int pad:6; |
} da1; |
struct |
{ |
int src0_indirect_offset:10; |
unsigned int src0_subreg_nr:3; |
unsigned int src0_abs:1; |
unsigned int src0_negate:1; |
unsigned int src0_address_mode:1; |
unsigned int src0_horiz_stride:2; |
unsigned int src0_width:3; |
unsigned int src0_vert_stride:4; |
unsigned int flag_reg_nr:1; |
unsigned int pad:6; |
} ia1; |
struct |
{ |
unsigned int src0_swz_x:2; |
unsigned int src0_swz_y:2; |
unsigned int src0_subreg_nr:1; |
unsigned int src0_reg_nr:8; |
unsigned int src0_abs:1; |
unsigned int src0_negate:1; |
unsigned int src0_address_mode:1; |
unsigned int src0_swz_z:2; |
unsigned int src0_swz_w:2; |
unsigned int pad0:1; |
unsigned int src0_vert_stride:4; |
unsigned int flag_reg_nr:1; |
unsigned int pad1:6; |
} da16; |
struct |
{ |
unsigned int src0_swz_x:2; |
unsigned int src0_swz_y:2; |
int src0_indirect_offset:6; |
unsigned int src0_subreg_nr:3; |
unsigned int src0_abs:1; |
unsigned int src0_negate:1; |
unsigned int src0_address_mode:1; |
unsigned int src0_swz_z:2; |
unsigned int src0_swz_w:2; |
unsigned int pad0:1; |
unsigned int src0_vert_stride:4; |
unsigned int flag_reg_nr:1; |
unsigned int pad1:6; |
} ia16; |
} bits2; |
union |
{ |
struct |
{ |
unsigned int src1_subreg_nr:5; |
unsigned int src1_reg_nr:8; |
unsigned int src1_abs:1; |
unsigned int src1_negate:1; |
unsigned int pad:1; |
unsigned int src1_horiz_stride:2; |
unsigned int src1_width:3; |
unsigned int src1_vert_stride:4; |
unsigned int pad0:7; |
} da1; |
struct |
{ |
unsigned int src1_swz_x:2; |
unsigned int src1_swz_y:2; |
unsigned int src1_subreg_nr:1; |
unsigned int src1_reg_nr:8; |
unsigned int src1_abs:1; |
unsigned int src1_negate:1; |
unsigned int pad0:1; |
unsigned int src1_swz_z:2; |
unsigned int src1_swz_w:2; |
unsigned int pad1:1; |
unsigned int src1_vert_stride:4; |
unsigned int pad2:7; |
} da16; |
struct |
{ |
int src1_indirect_offset:10; |
unsigned int src1_subreg_nr:3; |
unsigned int src1_abs:1; |
unsigned int src1_negate:1; |
unsigned int pad0:1; |
unsigned int src1_horiz_stride:2; |
unsigned int src1_width:3; |
unsigned int src1_vert_stride:4; |
unsigned int flag_reg_nr:1; |
unsigned int pad1:6; |
} ia1; |
struct |
{ |
unsigned int src1_swz_x:2; |
unsigned int src1_swz_y:2; |
int src1_indirect_offset:6; |
unsigned int src1_subreg_nr:3; |
unsigned int src1_abs:1; |
unsigned int src1_negate:1; |
unsigned int pad0:1; |
unsigned int src1_swz_z:2; |
unsigned int src1_swz_w:2; |
unsigned int pad1:1; |
unsigned int src1_vert_stride:4; |
unsigned int flag_reg_nr:1; |
unsigned int pad2:6; |
} ia16; |
struct |
{ |
int jump_count:16; /* note: signed */ |
unsigned int pop_count:4; |
unsigned int pad0:12; |
} if_else; |
struct { |
unsigned int function:4; |
unsigned int int_type:1; |
unsigned int precision:1; |
unsigned int saturate:1; |
unsigned int data_type:1; |
unsigned int pad0:8; |
unsigned int response_length:4; |
unsigned int msg_length:4; |
unsigned int msg_target:4; |
unsigned int pad1:3; |
unsigned int end_of_thread:1; |
} math; |
struct { |
unsigned int binding_table_index:8; |
unsigned int sampler:4; |
unsigned int return_format:2; |
unsigned int msg_type:2; |
unsigned int response_length:4; |
unsigned int msg_length:4; |
unsigned int msg_target:4; |
unsigned int pad1:3; |
unsigned int end_of_thread:1; |
} sampler; |
struct gen5_urb_immediate urb; |
struct { |
unsigned int binding_table_index:8; |
unsigned int msg_control:4; |
unsigned int msg_type:2; |
unsigned int target_cache:2; |
unsigned int response_length:4; |
unsigned int msg_length:4; |
unsigned int msg_target:4; |
unsigned int pad1:3; |
unsigned int end_of_thread:1; |
} dp_read; |
struct { |
unsigned int binding_table_index:8; |
unsigned int msg_control:3; |
unsigned int pixel_scoreboard_clear:1; |
unsigned int msg_type:3; |
unsigned int send_commit_msg:1; |
unsigned int response_length:4; |
unsigned int msg_length:4; |
unsigned int msg_target:4; |
unsigned int pad1:3; |
unsigned int end_of_thread:1; |
} dp_write; |
struct { |
unsigned int pad:16; |
unsigned int response_length:4; |
unsigned int msg_length:4; |
unsigned int msg_target:4; |
unsigned int pad1:3; |
unsigned int end_of_thread:1; |
} generic; |
unsigned int ud; |
} bits3; |
}; |
/* media pipeline */ |
struct gen5_vfe_state { |
struct { |
unsigned int per_thread_scratch_space:4; |
unsigned int pad3:3; |
unsigned int extend_vfe_state_present:1; |
unsigned int pad2:2; |
unsigned int scratch_base:22; |
} vfe0; |
struct { |
unsigned int debug_counter_control:2; |
unsigned int children_present:1; |
unsigned int vfe_mode:4; |
unsigned int pad2:2; |
unsigned int num_urb_entries:7; |
unsigned int urb_entry_alloc_size:9; |
unsigned int max_threads:7; |
} vfe1; |
struct { |
unsigned int pad4:4; |
unsigned int interface_descriptor_base:28; |
} vfe2; |
}; |
struct gen5_vld_state { |
struct { |
unsigned int pad6:6; |
unsigned int scan_order:1; |
unsigned int intra_vlc_format:1; |
unsigned int quantizer_scale_type:1; |
unsigned int concealment_motion_vector:1; |
unsigned int frame_predict_frame_dct:1; |
unsigned int top_field_first:1; |
unsigned int picture_structure:2; |
unsigned int intra_dc_precision:2; |
unsigned int f_code_0_0:4; |
unsigned int f_code_0_1:4; |
unsigned int f_code_1_0:4; |
unsigned int f_code_1_1:4; |
} vld0; |
struct { |
unsigned int pad2:9; |
unsigned int picture_coding_type:2; |
unsigned int pad:21; |
} vld1; |
struct { |
unsigned int index_0:4; |
unsigned int index_1:4; |
unsigned int index_2:4; |
unsigned int index_3:4; |
unsigned int index_4:4; |
unsigned int index_5:4; |
unsigned int index_6:4; |
unsigned int index_7:4; |
} desc_remap_table0; |
struct { |
unsigned int index_8:4; |
unsigned int index_9:4; |
unsigned int index_10:4; |
unsigned int index_11:4; |
unsigned int index_12:4; |
unsigned int index_13:4; |
unsigned int index_14:4; |
unsigned int index_15:4; |
} desc_remap_table1; |
}; |
struct gen5_interface_descriptor { |
struct { |
unsigned int grf_reg_blocks:4; |
unsigned int pad:2; |
unsigned int kernel_start_pointer:26; |
} desc0; |
struct { |
unsigned int pad:7; |
unsigned int software_exception:1; |
unsigned int pad2:3; |
unsigned int maskstack_exception:1; |
unsigned int pad3:1; |
unsigned int illegal_opcode_exception:1; |
unsigned int pad4:2; |
unsigned int floating_point_mode:1; |
unsigned int thread_priority:1; |
unsigned int single_program_flow:1; |
unsigned int pad5:1; |
unsigned int const_urb_entry_read_offset:6; |
unsigned int const_urb_entry_read_len:6; |
} desc1; |
struct { |
unsigned int pad:2; |
unsigned int sampler_count:3; |
unsigned int sampler_state_pointer:27; |
} desc2; |
struct { |
unsigned int binding_table_entry_count:5; |
unsigned int binding_table_pointer:27; |
} desc3; |
}; |
struct gen6_blend_state |
{ |
struct { |
unsigned int dest_blend_factor:5; |
unsigned int source_blend_factor:5; |
unsigned int pad3:1; |
unsigned int blend_func:3; |
unsigned int pad2:1; |
unsigned int ia_dest_blend_factor:5; |
unsigned int ia_source_blend_factor:5; |
unsigned int pad1:1; |
unsigned int ia_blend_func:3; |
unsigned int pad0:1; |
unsigned int ia_blend_enable:1; |
unsigned int blend_enable:1; |
} blend0; |
struct { |
unsigned int post_blend_clamp_enable:1; |
unsigned int pre_blend_clamp_enable:1; |
unsigned int clamp_range:2; |
unsigned int pad0:4; |
unsigned int x_dither_offset:2; |
unsigned int y_dither_offset:2; |
unsigned int dither_enable:1; |
unsigned int alpha_test_func:3; |
unsigned int alpha_test_enable:1; |
unsigned int pad1:1; |
unsigned int logic_op_func:4; |
unsigned int logic_op_enable:1; |
unsigned int pad2:1; |
unsigned int write_disable_b:1; |
unsigned int write_disable_g:1; |
unsigned int write_disable_r:1; |
unsigned int write_disable_a:1; |
unsigned int pad3:1; |
unsigned int alpha_to_coverage_dither:1; |
unsigned int alpha_to_one:1; |
unsigned int alpha_to_coverage:1; |
} blend1; |
}; |
struct gen6_color_calc_state |
{ |
struct { |
unsigned int alpha_test_format:1; |
unsigned int pad0:14; |
unsigned int round_disable:1; |
unsigned int bf_stencil_ref:8; |
unsigned int stencil_ref:8; |
} cc0; |
union { |
float alpha_ref_f; |
struct { |
unsigned int ui:8; |
unsigned int pad0:24; |
} alpha_ref_fi; |
} cc1; |
float constant_r; |
float constant_g; |
float constant_b; |
float constant_a; |
}; |
struct gen6_depth_stencil_state |
{ |
struct { |
unsigned int pad0:3; |
unsigned int bf_stencil_pass_depth_pass_op:3; |
unsigned int bf_stencil_pass_depth_fail_op:3; |
unsigned int bf_stencil_fail_op:3; |
unsigned int bf_stencil_func:3; |
unsigned int bf_stencil_enable:1; |
unsigned int pad1:2; |
unsigned int stencil_write_enable:1; |
unsigned int stencil_pass_depth_pass_op:3; |
unsigned int stencil_pass_depth_fail_op:3; |
unsigned int stencil_fail_op:3; |
unsigned int stencil_func:3; |
unsigned int stencil_enable:1; |
} ds0; |
struct { |
unsigned int bf_stencil_write_mask:8; |
unsigned int bf_stencil_test_mask:8; |
unsigned int stencil_write_mask:8; |
unsigned int stencil_test_mask:8; |
} ds1; |
struct { |
unsigned int pad0:26; |
unsigned int depth_write_enable:1; |
unsigned int depth_test_func:3; |
unsigned int pad1:1; |
unsigned int depth_test_enable:1; |
} ds2; |
}; |
typedef enum { |
SAMPLER_FILTER_NEAREST = 0, |
SAMPLER_FILTER_BILINEAR, |
FILTER_COUNT |
} sampler_filter_t; |
typedef enum { |
SAMPLER_EXTEND_NONE = 0, |
SAMPLER_EXTEND_REPEAT, |
SAMPLER_EXTEND_PAD, |
SAMPLER_EXTEND_REFLECT, |
EXTEND_COUNT |
} sampler_extend_t; |
typedef enum { |
WM_KERNEL = 0, |
WM_KERNEL_P, |
WM_KERNEL_MASK, |
WM_KERNEL_MASK_P, |
WM_KERNEL_MASKCA, |
WM_KERNEL_MASKCA_P, |
WM_KERNEL_MASKSA, |
WM_KERNEL_MASKSA_P, |
WM_KERNEL_OPACITY, |
WM_KERNEL_OPACITY_P, |
WM_KERNEL_VIDEO_PLANAR, |
WM_KERNEL_VIDEO_PACKED, |
KERNEL_COUNT |
} wm_kernel_t; |
#endif |
/drivers/video/Intel-2D/sna/gen6_render.c |
---|
0,0 → 1,3583 |
/* |
* Copyright © 2006,2008,2011 Intel Corporation |
* Copyright © 2007 Red Hat, Inc. |
* |
* Permission is hereby granted, free of charge, to any person obtaining a |
* copy of this software and associated documentation files (the "Software"), |
* to deal in the Software without restriction, including without limitation |
* the rights to use, copy, modify, merge, publish, distribute, sublicense, |
* and/or sell copies of the Software, and to permit persons to whom the |
* Software is furnished to do so, subject to the following conditions: |
* |
* The above copyright notice and this permission notice (including the next |
* paragraph) shall be included in all copies or substantial portions of the |
* Software. |
* |
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL |
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, |
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE |
* SOFTWARE. |
* |
* Authors: |
* Wang Zhenyu <zhenyu.z.wang@sna.com> |
* Eric Anholt <eric@anholt.net> |
* Carl Worth <cworth@redhat.com> |
* Keith Packard <keithp@keithp.com> |
* Chris Wilson <chris@chris-wilson.co.uk> |
* |
*/ |
#include "sna.h" |
#include "sna_reg.h" |
#include "sna_render.h" |
#include "sna_render_inline.h" |
//#include "sna_video.h" |
#include "brw/brw.h" |
#include "gen6_render.h" |
#include "gen4_source.h" |
#include "gen4_vertex.h" |
#define NO_COMPOSITE 0 |
#define NO_COMPOSITE_SPANS 0 |
#define NO_COPY 0 |
#define NO_COPY_BOXES 0 |
#define NO_FILL 0 |
#define NO_FILL_BOXES 0 |
#define NO_FILL_ONE 0 |
#define NO_FILL_CLEAR 0 |
#define NO_RING_SWITCH 0 |
#define PREFER_RENDER 0 |
#define USE_8_PIXEL_DISPATCH 1 |
#define USE_16_PIXEL_DISPATCH 1 |
#define USE_32_PIXEL_DISPATCH 0 |
#if !USE_8_PIXEL_DISPATCH && !USE_16_PIXEL_DISPATCH && !USE_32_PIXEL_DISPATCH |
#error "Must select at least 8, 16 or 32 pixel dispatch" |
#endif |
#define GEN6_MAX_SIZE 8192 |
struct gt_info { |
const char *name; |
int max_vs_threads; |
int max_gs_threads; |
int max_wm_threads; |
struct { |
int size; |
int max_vs_entries; |
int max_gs_entries; |
} urb; |
}; |
static const struct gt_info gt1_info = { |
.name = "Sandybridge (gen6, gt1)", |
.max_vs_threads = 24, |
.max_gs_threads = 21, |
.max_wm_threads = 40, |
.urb = { 32, 256, 256 }, |
}; |
static const struct gt_info gt2_info = { |
.name = "Sandybridge (gen6, gt2)", |
.max_vs_threads = 60, |
.max_gs_threads = 60, |
.max_wm_threads = 80, |
.urb = { 64, 256, 256 }, |
}; |
static const uint32_t ps_kernel_packed[][4] = { |
#include "exa_wm_src_affine.g6b" |
#include "exa_wm_src_sample_argb.g6b" |
#include "exa_wm_yuv_rgb.g6b" |
#include "exa_wm_write.g6b" |
}; |
static const uint32_t ps_kernel_planar[][4] = { |
#include "exa_wm_src_affine.g6b" |
#include "exa_wm_src_sample_planar.g6b" |
#include "exa_wm_yuv_rgb.g6b" |
#include "exa_wm_write.g6b" |
}; |
#define NOKERNEL(kernel_enum, func, ns) \ |
[GEN6_WM_KERNEL_##kernel_enum] = {#kernel_enum, func, 0, ns} |
#define KERNEL(kernel_enum, kernel, ns) \ |
[GEN6_WM_KERNEL_##kernel_enum] = {#kernel_enum, kernel, sizeof(kernel), ns} |
static const struct wm_kernel_info { |
const char *name; |
const void *data; |
unsigned int size; |
unsigned int num_surfaces; |
} wm_kernels[] = { |
NOKERNEL(NOMASK, brw_wm_kernel__affine, 2), |
NOKERNEL(NOMASK_P, brw_wm_kernel__projective, 2), |
NOKERNEL(MASK, brw_wm_kernel__affine_mask, 3), |
NOKERNEL(MASK_P, brw_wm_kernel__projective_mask, 3), |
NOKERNEL(MASKCA, brw_wm_kernel__affine_mask_ca, 3), |
NOKERNEL(MASKCA_P, brw_wm_kernel__projective_mask_ca, 3), |
NOKERNEL(MASKSA, brw_wm_kernel__affine_mask_sa, 3), |
NOKERNEL(MASKSA_P, brw_wm_kernel__projective_mask_sa, 3), |
NOKERNEL(OPACITY, brw_wm_kernel__affine_opacity, 2), |
NOKERNEL(OPACITY_P, brw_wm_kernel__projective_opacity, 2), |
KERNEL(VIDEO_PLANAR, ps_kernel_planar, 7), |
KERNEL(VIDEO_PACKED, ps_kernel_packed, 2), |
}; |
#undef KERNEL |
static const struct blendinfo { |
bool src_alpha; |
uint32_t src_blend; |
uint32_t dst_blend; |
} gen6_blend_op[] = { |
/* Clear */ {0, GEN6_BLENDFACTOR_ZERO, GEN6_BLENDFACTOR_ZERO}, |
/* Src */ {0, GEN6_BLENDFACTOR_ONE, GEN6_BLENDFACTOR_ZERO}, |
/* Dst */ {0, GEN6_BLENDFACTOR_ZERO, GEN6_BLENDFACTOR_ONE}, |
/* Over */ {1, GEN6_BLENDFACTOR_ONE, GEN6_BLENDFACTOR_INV_SRC_ALPHA}, |
/* OverReverse */ {0, GEN6_BLENDFACTOR_INV_DST_ALPHA, GEN6_BLENDFACTOR_ONE}, |
/* In */ {0, GEN6_BLENDFACTOR_DST_ALPHA, GEN6_BLENDFACTOR_ZERO}, |
/* InReverse */ {1, GEN6_BLENDFACTOR_ZERO, GEN6_BLENDFACTOR_SRC_ALPHA}, |
/* Out */ {0, GEN6_BLENDFACTOR_INV_DST_ALPHA, GEN6_BLENDFACTOR_ZERO}, |
/* OutReverse */ {1, GEN6_BLENDFACTOR_ZERO, GEN6_BLENDFACTOR_INV_SRC_ALPHA}, |
/* Atop */ {1, GEN6_BLENDFACTOR_DST_ALPHA, GEN6_BLENDFACTOR_INV_SRC_ALPHA}, |
/* AtopReverse */ {1, GEN6_BLENDFACTOR_INV_DST_ALPHA, GEN6_BLENDFACTOR_SRC_ALPHA}, |
/* Xor */ {1, GEN6_BLENDFACTOR_INV_DST_ALPHA, GEN6_BLENDFACTOR_INV_SRC_ALPHA}, |
/* Add */ {0, GEN6_BLENDFACTOR_ONE, GEN6_BLENDFACTOR_ONE}, |
}; |
/** |
* Highest-valued BLENDFACTOR used in gen6_blend_op. |
* |
* This leaves out GEN6_BLENDFACTOR_INV_DST_COLOR, |
* GEN6_BLENDFACTOR_INV_CONST_{COLOR,ALPHA}, |
* GEN6_BLENDFACTOR_INV_SRC1_{COLOR,ALPHA} |
*/ |
#define GEN6_BLENDFACTOR_COUNT (GEN6_BLENDFACTOR_INV_DST_ALPHA + 1) |
#define GEN6_BLEND_STATE_PADDED_SIZE ALIGN(sizeof(struct gen6_blend_state), 64) |
#define BLEND_OFFSET(s, d) \ |
(((s) * GEN6_BLENDFACTOR_COUNT + (d)) * GEN6_BLEND_STATE_PADDED_SIZE) |
#define NO_BLEND BLEND_OFFSET(GEN6_BLENDFACTOR_ONE, GEN6_BLENDFACTOR_ZERO) |
#define CLEAR BLEND_OFFSET(GEN6_BLENDFACTOR_ZERO, GEN6_BLENDFACTOR_ZERO) |
#define SAMPLER_OFFSET(sf, se, mf, me) \ |
(((((sf) * EXTEND_COUNT + (se)) * FILTER_COUNT + (mf)) * EXTEND_COUNT + (me) + 2) * 2 * sizeof(struct gen6_sampler_state)) |
#define VERTEX_2s2s 0 |
#define COPY_SAMPLER 0 |
#define COPY_VERTEX VERTEX_2s2s |
#define COPY_FLAGS(a) GEN6_SET_FLAGS(COPY_SAMPLER, (a) == GXcopy ? NO_BLEND : CLEAR, GEN6_WM_KERNEL_NOMASK, COPY_VERTEX) |
#define FILL_SAMPLER (2 * sizeof(struct gen6_sampler_state)) |
#define FILL_VERTEX VERTEX_2s2s |
#define FILL_FLAGS(op, format) GEN6_SET_FLAGS(FILL_SAMPLER, gen6_get_blend((op), false, (format)), GEN6_WM_KERNEL_NOMASK, FILL_VERTEX) |
#define FILL_FLAGS_NOBLEND GEN6_SET_FLAGS(FILL_SAMPLER, NO_BLEND, GEN6_WM_KERNEL_NOMASK, FILL_VERTEX) |
#define GEN6_SAMPLER(f) (((f) >> 16) & 0xfff0) |
#define GEN6_BLEND(f) (((f) >> 0) & 0xfff0) |
#define GEN6_KERNEL(f) (((f) >> 16) & 0xf) |
#define GEN6_VERTEX(f) (((f) >> 0) & 0xf) |
#define GEN6_SET_FLAGS(S, B, K, V) (((S) | (K)) << 16 | ((B) | (V))) |
#define OUT_BATCH(v) batch_emit(sna, v) |
#define OUT_VERTEX(x,y) vertex_emit_2s(sna, x,y) |
#define OUT_VERTEX_F(v) vertex_emit(sna, v) |
static inline bool too_large(int width, int height) |
{ |
return width > GEN6_MAX_SIZE || height > GEN6_MAX_SIZE; |
} |
static uint32_t gen6_get_blend(int op, |
bool has_component_alpha, |
uint32_t dst_format) |
{ |
uint32_t src, dst; |
src = GEN6_BLENDFACTOR_ONE; //gen6_blend_op[op].src_blend; |
dst = GEN6_BLENDFACTOR_INV_SRC_ALPHA; //gen6_blend_op[op].dst_blend; |
// dst = GEN6_BLENDFACTOR_ZERO; //gen6_blend_op[op].dst_blend; |
#if 0 |
/* If there's no dst alpha channel, adjust the blend op so that |
* we'll treat it always as 1. |
*/ |
if (PICT_FORMAT_A(dst_format) == 0) { |
if (src == GEN6_BLENDFACTOR_DST_ALPHA) |
src = GEN6_BLENDFACTOR_ONE; |
else if (src == GEN6_BLENDFACTOR_INV_DST_ALPHA) |
src = GEN6_BLENDFACTOR_ZERO; |
} |
/* If the source alpha is being used, then we should only be in a |
* case where the source blend factor is 0, and the source blend |
* value is the mask channels multiplied by the source picture's alpha. |
*/ |
if (has_component_alpha && gen6_blend_op[op].src_alpha) { |
if (dst == GEN6_BLENDFACTOR_SRC_ALPHA) |
dst = GEN6_BLENDFACTOR_SRC_COLOR; |
else if (dst == GEN6_BLENDFACTOR_INV_SRC_ALPHA) |
dst = GEN6_BLENDFACTOR_INV_SRC_COLOR; |
} |
DBG(("blend op=%d, dst=%x [A=%d] => src=%d, dst=%d => offset=%x\n", |
op, dst_format, PICT_FORMAT_A(dst_format), |
src, dst, (int)BLEND_OFFSET(src, dst))); |
#endif |
return BLEND_OFFSET(src, dst); |
} |
static uint32_t gen6_get_card_format(PictFormat format) |
{ |
switch (format) { |
default: |
return -1; |
case PICT_a8r8g8b8: |
return GEN6_SURFACEFORMAT_B8G8R8A8_UNORM; |
case PICT_x8r8g8b8: |
return GEN6_SURFACEFORMAT_B8G8R8X8_UNORM; |
case PICT_a8b8g8r8: |
return GEN6_SURFACEFORMAT_R8G8B8A8_UNORM; |
case PICT_x8b8g8r8: |
return GEN6_SURFACEFORMAT_R8G8B8X8_UNORM; |
case PICT_a2r10g10b10: |
return GEN6_SURFACEFORMAT_B10G10R10A2_UNORM; |
case PICT_x2r10g10b10: |
return GEN6_SURFACEFORMAT_B10G10R10X2_UNORM; |
case PICT_r8g8b8: |
return GEN6_SURFACEFORMAT_R8G8B8_UNORM; |
case PICT_r5g6b5: |
return GEN6_SURFACEFORMAT_B5G6R5_UNORM; |
case PICT_a1r5g5b5: |
return GEN6_SURFACEFORMAT_B5G5R5A1_UNORM; |
case PICT_a8: |
return GEN6_SURFACEFORMAT_A8_UNORM; |
case PICT_a4r4g4b4: |
return GEN6_SURFACEFORMAT_B4G4R4A4_UNORM; |
} |
} |
static uint32_t gen6_get_dest_format(PictFormat format) |
{ |
return GEN6_SURFACEFORMAT_B8G8R8A8_UNORM; |
#if 0 |
switch (format) { |
default: |
return -1; |
case PICT_a8r8g8b8: |
case PICT_x8r8g8b8: |
return GEN6_SURFACEFORMAT_B8G8R8A8_UNORM; |
case PICT_a8b8g8r8: |
case PICT_x8b8g8r8: |
return GEN6_SURFACEFORMAT_R8G8B8A8_UNORM; |
case PICT_a2r10g10b10: |
case PICT_x2r10g10b10: |
return GEN6_SURFACEFORMAT_B10G10R10A2_UNORM; |
case PICT_r5g6b5: |
return GEN6_SURFACEFORMAT_B5G6R5_UNORM; |
case PICT_x1r5g5b5: |
case PICT_a1r5g5b5: |
return GEN6_SURFACEFORMAT_B5G5R5A1_UNORM; |
case PICT_a8: |
return GEN6_SURFACEFORMAT_A8_UNORM; |
case PICT_a4r4g4b4: |
case PICT_x4r4g4b4: |
return GEN6_SURFACEFORMAT_B4G4R4A4_UNORM; |
} |
#endif |
} |
#if 0 |
static bool gen6_check_dst_format(PictFormat format) |
{ |
if (gen6_get_dest_format(format) != -1) |
return true; |
DBG(("%s: unhandled format: %x\n", __FUNCTION__, (int)format)); |
return false; |
} |
static bool gen6_check_format(uint32_t format) |
{ |
if (gen6_get_card_format(format) != -1) |
return true; |
DBG(("%s: unhandled format: %x\n", __FUNCTION__, (int)format)); |
return false; |
} |
static uint32_t gen6_filter(uint32_t filter) |
{ |
switch (filter) { |
default: |
assert(0); |
case PictFilterNearest: |
return SAMPLER_FILTER_NEAREST; |
case PictFilterBilinear: |
return SAMPLER_FILTER_BILINEAR; |
} |
} |
static uint32_t gen6_check_filter(PicturePtr picture) |
{ |
switch (picture->filter) { |
case PictFilterNearest: |
case PictFilterBilinear: |
return true; |
default: |
return false; |
} |
} |
static uint32_t gen6_repeat(uint32_t repeat) |
{ |
switch (repeat) { |
default: |
assert(0); |
case RepeatNone: |
return SAMPLER_EXTEND_NONE; |
case RepeatNormal: |
return SAMPLER_EXTEND_REPEAT; |
case RepeatPad: |
return SAMPLER_EXTEND_PAD; |
case RepeatReflect: |
return SAMPLER_EXTEND_REFLECT; |
} |
} |
static bool gen6_check_repeat(PicturePtr picture) |
{ |
if (!picture->repeat) |
return true; |
switch (picture->repeatType) { |
case RepeatNone: |
case RepeatNormal: |
case RepeatPad: |
case RepeatReflect: |
return true; |
default: |
return false; |
} |
} |
#endif |
static int |
gen6_choose_composite_kernel(int op, bool has_mask, bool is_ca, bool is_affine) |
{ |
int base; |
if (has_mask) { |
if (is_ca) { |
if (gen6_blend_op[op].src_alpha) |
base = GEN6_WM_KERNEL_MASKSA; |
else |
base = GEN6_WM_KERNEL_MASKCA; |
} else |
base = GEN6_WM_KERNEL_MASK; |
} else |
base = GEN6_WM_KERNEL_NOMASK; |
return base + !is_affine; |
} |
static void |
gen6_emit_urb(struct sna *sna) |
{ |
OUT_BATCH(GEN6_3DSTATE_URB | (3 - 2)); |
OUT_BATCH(((1 - 1) << GEN6_3DSTATE_URB_VS_SIZE_SHIFT) | |
(sna->render_state.gen6.info->urb.max_vs_entries << GEN6_3DSTATE_URB_VS_ENTRIES_SHIFT)); /* at least 24 on GEN6 */ |
OUT_BATCH((0 << GEN6_3DSTATE_URB_GS_SIZE_SHIFT) | |
(0 << GEN6_3DSTATE_URB_GS_ENTRIES_SHIFT)); /* no GS thread */ |
} |
static void |
gen6_emit_state_base_address(struct sna *sna) |
{ |
OUT_BATCH(GEN6_STATE_BASE_ADDRESS | (10 - 2)); |
OUT_BATCH(0); /* general */ |
OUT_BATCH(kgem_add_reloc(&sna->kgem, /* surface */ |
sna->kgem.nbatch, |
NULL, |
I915_GEM_DOMAIN_INSTRUCTION << 16, |
BASE_ADDRESS_MODIFY)); |
OUT_BATCH(kgem_add_reloc(&sna->kgem, /* instruction */ |
sna->kgem.nbatch, |
sna->render_state.gen6.general_bo, |
I915_GEM_DOMAIN_INSTRUCTION << 16, |
BASE_ADDRESS_MODIFY)); |
OUT_BATCH(0); /* indirect */ |
OUT_BATCH(kgem_add_reloc(&sna->kgem, |
sna->kgem.nbatch, |
sna->render_state.gen6.general_bo, |
I915_GEM_DOMAIN_INSTRUCTION << 16, |
BASE_ADDRESS_MODIFY)); |
/* upper bounds, disable */ |
OUT_BATCH(0); |
OUT_BATCH(BASE_ADDRESS_MODIFY); |
OUT_BATCH(0); |
OUT_BATCH(BASE_ADDRESS_MODIFY); |
} |
static void |
gen6_emit_viewports(struct sna *sna) |
{ |
OUT_BATCH(GEN6_3DSTATE_VIEWPORT_STATE_POINTERS | |
GEN6_3DSTATE_VIEWPORT_STATE_MODIFY_CC | |
(4 - 2)); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
} |
static void |
gen6_emit_vs(struct sna *sna) |
{ |
/* disable VS constant buffer */ |
OUT_BATCH(GEN6_3DSTATE_CONSTANT_VS | (5 - 2)); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(GEN6_3DSTATE_VS | (6 - 2)); |
OUT_BATCH(0); /* no VS kernel */ |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); /* pass-through */ |
} |
static void |
gen6_emit_gs(struct sna *sna) |
{ |
/* disable GS constant buffer */ |
OUT_BATCH(GEN6_3DSTATE_CONSTANT_GS | (5 - 2)); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(GEN6_3DSTATE_GS | (7 - 2)); |
OUT_BATCH(0); /* no GS kernel */ |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); /* pass-through */ |
} |
static void |
gen6_emit_clip(struct sna *sna) |
{ |
OUT_BATCH(GEN6_3DSTATE_CLIP | (4 - 2)); |
OUT_BATCH(0); |
OUT_BATCH(0); /* pass-through */ |
OUT_BATCH(0); |
} |
static void |
gen6_emit_wm_constants(struct sna *sna) |
{ |
/* disable WM constant buffer */ |
OUT_BATCH(GEN6_3DSTATE_CONSTANT_PS | (5 - 2)); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
} |
static void |
gen6_emit_null_depth_buffer(struct sna *sna) |
{ |
OUT_BATCH(GEN6_3DSTATE_DEPTH_BUFFER | (7 - 2)); |
OUT_BATCH(GEN6_SURFACE_NULL << GEN6_3DSTATE_DEPTH_BUFFER_TYPE_SHIFT | |
GEN6_DEPTHFORMAT_D32_FLOAT << GEN6_3DSTATE_DEPTH_BUFFER_FORMAT_SHIFT); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(GEN6_3DSTATE_CLEAR_PARAMS | (2 - 2)); |
OUT_BATCH(0); |
} |
static void |
gen6_emit_invariant(struct sna *sna) |
{ |
OUT_BATCH(GEN6_PIPELINE_SELECT | PIPELINE_SELECT_3D); |
OUT_BATCH(GEN6_3DSTATE_MULTISAMPLE | (3 - 2)); |
OUT_BATCH(GEN6_3DSTATE_MULTISAMPLE_PIXEL_LOCATION_CENTER | |
GEN6_3DSTATE_MULTISAMPLE_NUMSAMPLES_1); /* 1 sample/pixel */ |
OUT_BATCH(0); |
OUT_BATCH(GEN6_3DSTATE_SAMPLE_MASK | (2 - 2)); |
OUT_BATCH(1); |
gen6_emit_urb(sna); |
gen6_emit_state_base_address(sna); |
gen6_emit_viewports(sna); |
gen6_emit_vs(sna); |
gen6_emit_gs(sna); |
gen6_emit_clip(sna); |
gen6_emit_wm_constants(sna); |
gen6_emit_null_depth_buffer(sna); |
sna->render_state.gen6.needs_invariant = false; |
} |
static bool |
gen6_emit_cc(struct sna *sna, int blend) |
{ |
struct gen6_render_state *render = &sna->render_state.gen6; |
if (render->blend == blend) |
return blend != NO_BLEND; |
DBG(("%s: blend = %x\n", __FUNCTION__, blend)); |
OUT_BATCH(GEN6_3DSTATE_CC_STATE_POINTERS | (4 - 2)); |
OUT_BATCH((render->cc_blend + blend) | 1); |
if (render->blend == (unsigned)-1) { |
OUT_BATCH(1); |
OUT_BATCH(1); |
} else { |
OUT_BATCH(0); |
OUT_BATCH(0); |
} |
render->blend = blend; |
return blend != NO_BLEND; |
} |
static void |
gen6_emit_sampler(struct sna *sna, uint32_t state) |
{ |
if (sna->render_state.gen6.samplers == state) |
return; |
sna->render_state.gen6.samplers = state; |
DBG(("%s: sampler = %x\n", __FUNCTION__, state)); |
OUT_BATCH(GEN6_3DSTATE_SAMPLER_STATE_POINTERS | |
GEN6_3DSTATE_SAMPLER_STATE_MODIFY_PS | |
(4 - 2)); |
OUT_BATCH(0); /* VS */ |
OUT_BATCH(0); /* GS */ |
OUT_BATCH(sna->render_state.gen6.wm_state + state); |
} |
static void |
gen6_emit_sf(struct sna *sna, bool has_mask) |
{ |
int num_sf_outputs = has_mask ? 2 : 1; |
if (sna->render_state.gen6.num_sf_outputs == num_sf_outputs) |
return; |
DBG(("%s: num_sf_outputs=%d, read_length=%d, read_offset=%d\n", |
__FUNCTION__, num_sf_outputs, 1, 0)); |
sna->render_state.gen6.num_sf_outputs = num_sf_outputs; |
OUT_BATCH(GEN6_3DSTATE_SF | (20 - 2)); |
OUT_BATCH(num_sf_outputs << GEN6_3DSTATE_SF_NUM_OUTPUTS_SHIFT | |
1 << GEN6_3DSTATE_SF_URB_ENTRY_READ_LENGTH_SHIFT | |
1 << GEN6_3DSTATE_SF_URB_ENTRY_READ_OFFSET_SHIFT); |
OUT_BATCH(0); |
OUT_BATCH(GEN6_3DSTATE_SF_CULL_NONE); |
OUT_BATCH(2 << GEN6_3DSTATE_SF_TRIFAN_PROVOKE_SHIFT); /* DW4 */ |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); /* DW9 */ |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); /* DW14 */ |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); /* DW19 */ |
} |
static void |
gen6_emit_wm(struct sna *sna, unsigned int kernel, bool has_mask) |
{ |
const uint32_t *kernels; |
if (sna->render_state.gen6.kernel == kernel) |
return; |
sna->render_state.gen6.kernel = kernel; |
kernels = sna->render_state.gen6.wm_kernel[kernel]; |
DBG(("%s: switching to %s, num_surfaces=%d (8-pixel? %d, 16-pixel? %d,32-pixel? %d)\n", |
__FUNCTION__, |
wm_kernels[kernel].name, wm_kernels[kernel].num_surfaces, |
kernels[0], kernels[1], kernels[2])); |
OUT_BATCH(GEN6_3DSTATE_WM | (9 - 2)); |
OUT_BATCH(kernels[0] ?: kernels[1] ?: kernels[2]); |
OUT_BATCH(1 << GEN6_3DSTATE_WM_SAMPLER_COUNT_SHIFT | |
wm_kernels[kernel].num_surfaces << GEN6_3DSTATE_WM_BINDING_TABLE_ENTRY_COUNT_SHIFT); |
OUT_BATCH(0); /* scratch space */ |
OUT_BATCH((kernels[0] ? 4 : kernels[1] ? 6 : 8) << GEN6_3DSTATE_WM_DISPATCH_0_START_GRF_SHIFT | |
8 << GEN6_3DSTATE_WM_DISPATCH_1_START_GRF_SHIFT | |
6 << GEN6_3DSTATE_WM_DISPATCH_2_START_GRF_SHIFT); |
OUT_BATCH((sna->render_state.gen6.info->max_wm_threads - 1) << GEN6_3DSTATE_WM_MAX_THREADS_SHIFT | |
(kernels[0] ? GEN6_3DSTATE_WM_8_DISPATCH_ENABLE : 0) | |
(kernels[1] ? GEN6_3DSTATE_WM_16_DISPATCH_ENABLE : 0) | |
(kernels[2] ? GEN6_3DSTATE_WM_32_DISPATCH_ENABLE : 0) | |
GEN6_3DSTATE_WM_DISPATCH_ENABLE); |
OUT_BATCH((1 + has_mask) << GEN6_3DSTATE_WM_NUM_SF_OUTPUTS_SHIFT | |
GEN6_3DSTATE_WM_PERSPECTIVE_PIXEL_BARYCENTRIC); |
OUT_BATCH(kernels[2]); |
OUT_BATCH(kernels[1]); |
} |
static bool |
gen6_emit_binding_table(struct sna *sna, uint16_t offset) |
{ |
if (sna->render_state.gen6.surface_table == offset) |
return false; |
/* Binding table pointers */ |
OUT_BATCH(GEN6_3DSTATE_BINDING_TABLE_POINTERS | |
GEN6_3DSTATE_BINDING_TABLE_MODIFY_PS | |
(4 - 2)); |
OUT_BATCH(0); /* vs */ |
OUT_BATCH(0); /* gs */ |
/* Only the PS uses the binding table */ |
OUT_BATCH(offset*4); |
sna->render_state.gen6.surface_table = offset; |
return true; |
} |
static bool |
gen6_emit_drawing_rectangle(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
uint32_t limit = (op->dst.height - 1) << 16 | (op->dst.width - 1); |
uint32_t offset = (uint16_t)op->dst.y << 16 | (uint16_t)op->dst.x; |
assert(!too_large(op->dst.x, op->dst.y)); |
assert(!too_large(op->dst.width, op->dst.height)); |
if (sna->render_state.gen6.drawrect_limit == limit && |
sna->render_state.gen6.drawrect_offset == offset) |
return false; |
/* [DevSNB-C+{W/A}] Before any depth stall flush (including those |
* produced by non-pipelined state commands), software needs to first |
* send a PIPE_CONTROL with no bits set except Post-Sync Operation != |
* 0. |
* |
* [Dev-SNB{W/A}]: Pipe-control with CS-stall bit set must be sent |
* BEFORE the pipe-control with a post-sync op and no write-cache |
* flushes. |
*/ |
if (!sna->render_state.gen6.first_state_packet) { |
OUT_BATCH(GEN6_PIPE_CONTROL | (4 - 2)); |
OUT_BATCH(GEN6_PIPE_CONTROL_CS_STALL | |
GEN6_PIPE_CONTROL_STALL_AT_SCOREBOARD); |
OUT_BATCH(0); |
OUT_BATCH(0); |
} |
OUT_BATCH(GEN6_PIPE_CONTROL | (4 - 2)); |
OUT_BATCH(GEN6_PIPE_CONTROL_WRITE_TIME); |
OUT_BATCH(kgem_add_reloc(&sna->kgem, sna->kgem.nbatch, |
sna->render_state.gen6.general_bo, |
I915_GEM_DOMAIN_INSTRUCTION << 16 | |
I915_GEM_DOMAIN_INSTRUCTION, |
64)); |
OUT_BATCH(0); |
OUT_BATCH(GEN6_3DSTATE_DRAWING_RECTANGLE | (4 - 2)); |
OUT_BATCH(0); |
OUT_BATCH(limit); |
OUT_BATCH(offset); |
sna->render_state.gen6.drawrect_offset = offset; |
sna->render_state.gen6.drawrect_limit = limit; |
return true; |
} |
static void |
gen6_emit_vertex_elements(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
/* |
* vertex data in vertex buffer |
* position: (x, y) |
* texture coordinate 0: (u0, v0) if (is_affine is true) else (u0, v0, w0) |
* texture coordinate 1 if (has_mask is true): same as above |
*/ |
struct gen6_render_state *render = &sna->render_state.gen6; |
uint32_t src_format, dw; |
int id = GEN6_VERTEX(op->u.gen6.flags); |
bool has_mask; |
DBG(("%s: setup id=%d\n", __FUNCTION__, id)); |
if (render->ve_id == id) |
return; |
render->ve_id = id; |
/* The VUE layout |
* dword 0-3: pad (0.0, 0.0, 0.0. 0.0) |
* dword 4-7: position (x, y, 1.0, 1.0), |
* dword 8-11: texture coordinate 0 (u0, v0, w0, 1.0) |
* dword 12-15: texture coordinate 1 (u1, v1, w1, 1.0) |
* |
* dword 4-15 are fetched from vertex buffer |
*/ |
has_mask = (id >> 2) != 0; |
OUT_BATCH(GEN6_3DSTATE_VERTEX_ELEMENTS | |
((2 * (3 + has_mask)) + 1 - 2)); |
OUT_BATCH(id << VE0_VERTEX_BUFFER_INDEX_SHIFT | VE0_VALID | |
GEN6_SURFACEFORMAT_R32G32B32A32_FLOAT << VE0_FORMAT_SHIFT | |
0 << VE0_OFFSET_SHIFT); |
OUT_BATCH(GEN6_VFCOMPONENT_STORE_0 << VE1_VFCOMPONENT_0_SHIFT | |
GEN6_VFCOMPONENT_STORE_0 << VE1_VFCOMPONENT_1_SHIFT | |
GEN6_VFCOMPONENT_STORE_0 << VE1_VFCOMPONENT_2_SHIFT | |
GEN6_VFCOMPONENT_STORE_0 << VE1_VFCOMPONENT_3_SHIFT); |
/* x,y */ |
OUT_BATCH(id << VE0_VERTEX_BUFFER_INDEX_SHIFT | VE0_VALID | |
GEN6_SURFACEFORMAT_R16G16_SSCALED << VE0_FORMAT_SHIFT | |
0 << VE0_OFFSET_SHIFT); |
OUT_BATCH(GEN6_VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_0_SHIFT | |
GEN6_VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_1_SHIFT | |
GEN6_VFCOMPONENT_STORE_0 << VE1_VFCOMPONENT_2_SHIFT | |
GEN6_VFCOMPONENT_STORE_1_FLT << VE1_VFCOMPONENT_3_SHIFT); |
/* u0, v0, w0 */ |
DBG(("%s: first channel %d floats, offset=4b\n", __FUNCTION__, id & 3)); |
dw = GEN6_VFCOMPONENT_STORE_1_FLT << VE1_VFCOMPONENT_3_SHIFT; |
switch (id & 3) { |
default: |
assert(0); |
case 0: |
src_format = GEN6_SURFACEFORMAT_R16G16_SSCALED; |
dw |= GEN6_VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_0_SHIFT; |
dw |= GEN6_VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_1_SHIFT; |
dw |= GEN6_VFCOMPONENT_STORE_0 << VE1_VFCOMPONENT_2_SHIFT; |
break; |
case 1: |
src_format = GEN6_SURFACEFORMAT_R32_FLOAT; |
dw |= GEN6_VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_0_SHIFT; |
dw |= GEN6_VFCOMPONENT_STORE_0 << VE1_VFCOMPONENT_1_SHIFT; |
dw |= GEN6_VFCOMPONENT_STORE_0 << VE1_VFCOMPONENT_2_SHIFT; |
break; |
case 2: |
src_format = GEN6_SURFACEFORMAT_R32G32_FLOAT; |
dw |= GEN6_VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_0_SHIFT; |
dw |= GEN6_VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_1_SHIFT; |
dw |= GEN6_VFCOMPONENT_STORE_0 << VE1_VFCOMPONENT_2_SHIFT; |
break; |
case 3: |
src_format = GEN6_SURFACEFORMAT_R32G32B32_FLOAT; |
dw |= GEN6_VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_0_SHIFT; |
dw |= GEN6_VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_1_SHIFT; |
dw |= GEN6_VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_2_SHIFT; |
break; |
} |
OUT_BATCH(id << VE0_VERTEX_BUFFER_INDEX_SHIFT | VE0_VALID | |
src_format << VE0_FORMAT_SHIFT | |
4 << VE0_OFFSET_SHIFT); |
OUT_BATCH(dw); |
/* u1, v1, w1 */ |
if (has_mask) { |
unsigned offset = 4 + ((id & 3) ?: 1) * sizeof(float); |
DBG(("%s: second channel %d floats, offset=%db\n", __FUNCTION__, id >> 2, offset)); |
dw = GEN6_VFCOMPONENT_STORE_1_FLT << VE1_VFCOMPONENT_3_SHIFT; |
switch (id >> 2) { |
case 1: |
src_format = GEN6_SURFACEFORMAT_R32_FLOAT; |
dw |= GEN6_VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_0_SHIFT; |
dw |= GEN6_VFCOMPONENT_STORE_0 << VE1_VFCOMPONENT_1_SHIFT; |
dw |= GEN6_VFCOMPONENT_STORE_0 << VE1_VFCOMPONENT_2_SHIFT; |
break; |
default: |
assert(0); |
case 2: |
src_format = GEN6_SURFACEFORMAT_R32G32_FLOAT; |
dw |= GEN6_VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_0_SHIFT; |
dw |= GEN6_VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_1_SHIFT; |
dw |= GEN6_VFCOMPONENT_STORE_0 << VE1_VFCOMPONENT_2_SHIFT; |
break; |
case 3: |
src_format = GEN6_SURFACEFORMAT_R32G32B32_FLOAT; |
dw |= GEN6_VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_0_SHIFT; |
dw |= GEN6_VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_1_SHIFT; |
dw |= GEN6_VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_2_SHIFT; |
break; |
} |
OUT_BATCH(id << VE0_VERTEX_BUFFER_INDEX_SHIFT | VE0_VALID | |
src_format << VE0_FORMAT_SHIFT | |
offset << VE0_OFFSET_SHIFT); |
OUT_BATCH(dw); |
} |
} |
static void |
gen6_emit_flush(struct sna *sna) |
{ |
OUT_BATCH(GEN6_PIPE_CONTROL | (4 - 2)); |
OUT_BATCH(GEN6_PIPE_CONTROL_WC_FLUSH | |
GEN6_PIPE_CONTROL_TC_FLUSH | |
GEN6_PIPE_CONTROL_CS_STALL); |
OUT_BATCH(0); |
OUT_BATCH(0); |
} |
static void |
gen6_emit_state(struct sna *sna, |
const struct sna_composite_op *op, |
uint16_t wm_binding_table) |
{ |
bool need_stall = wm_binding_table & 1; |
assert(op->dst.bo->exec); |
if (gen6_emit_cc(sna, GEN6_BLEND(op->u.gen6.flags))) |
need_stall = false; |
gen6_emit_sampler(sna, GEN6_SAMPLER(op->u.gen6.flags)); |
gen6_emit_sf(sna, GEN6_VERTEX(op->u.gen6.flags) >> 2); |
gen6_emit_wm(sna, GEN6_KERNEL(op->u.gen6.flags), GEN6_VERTEX(op->u.gen6.flags) >> 2); |
gen6_emit_vertex_elements(sna, op); |
need_stall |= gen6_emit_binding_table(sna, wm_binding_table & ~1); |
if (gen6_emit_drawing_rectangle(sna, op)) |
need_stall = false; |
if (kgem_bo_is_dirty(op->src.bo) || kgem_bo_is_dirty(op->mask.bo)) { |
gen6_emit_flush(sna); |
kgem_clear_dirty(&sna->kgem); |
assert(op->dst.bo->exec); |
kgem_bo_mark_dirty(op->dst.bo); |
need_stall = false; |
} |
if (need_stall) { |
OUT_BATCH(GEN6_PIPE_CONTROL | (4 - 2)); |
OUT_BATCH(GEN6_PIPE_CONTROL_CS_STALL | |
GEN6_PIPE_CONTROL_STALL_AT_SCOREBOARD); |
OUT_BATCH(0); |
OUT_BATCH(0); |
} |
sna->render_state.gen6.first_state_packet = false; |
} |
static bool gen6_magic_ca_pass(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
struct gen6_render_state *state = &sna->render_state.gen6; |
if (!op->need_magic_ca_pass) |
return false; |
DBG(("%s: CA fixup (%d -> %d)\n", __FUNCTION__, |
sna->render.vertex_start, sna->render.vertex_index)); |
gen6_emit_flush(sna); |
gen6_emit_cc(sna, gen6_get_blend(PictOpAdd, true, op->dst.format)); |
gen6_emit_wm(sna, |
gen6_choose_composite_kernel(PictOpAdd, |
true, true, |
op->is_affine), |
true); |
OUT_BATCH(GEN6_3DPRIMITIVE | |
GEN6_3DPRIMITIVE_VERTEX_SEQUENTIAL | |
_3DPRIM_RECTLIST << GEN6_3DPRIMITIVE_TOPOLOGY_SHIFT | |
0 << 9 | |
4); |
OUT_BATCH(sna->render.vertex_index - sna->render.vertex_start); |
OUT_BATCH(sna->render.vertex_start); |
OUT_BATCH(1); /* single instance */ |
OUT_BATCH(0); /* start instance location */ |
OUT_BATCH(0); /* index buffer offset, ignored */ |
state->last_primitive = sna->kgem.nbatch; |
return true; |
} |
typedef struct gen6_surface_state_padded { |
struct gen6_surface_state state; |
char pad[32 - sizeof(struct gen6_surface_state)]; |
} gen6_surface_state_padded; |
static void null_create(struct sna_static_stream *stream) |
{ |
/* A bunch of zeros useful for legacy border color and depth-stencil */ |
sna_static_stream_map(stream, 64, 64); |
} |
static void scratch_create(struct sna_static_stream *stream) |
{ |
/* 64 bytes of scratch space for random writes, such as |
* the pipe-control w/a. |
*/ |
sna_static_stream_map(stream, 64, 64); |
} |
static void |
sampler_state_init(struct gen6_sampler_state *sampler_state, |
sampler_filter_t filter, |
sampler_extend_t extend) |
{ |
sampler_state->ss0.lod_preclamp = 1; /* GL mode */ |
/* We use the legacy mode to get the semantics specified by |
* the Render extension. */ |
sampler_state->ss0.border_color_mode = GEN6_BORDER_COLOR_MODE_LEGACY; |
switch (filter) { |
default: |
case SAMPLER_FILTER_NEAREST: |
sampler_state->ss0.min_filter = GEN6_MAPFILTER_NEAREST; |
sampler_state->ss0.mag_filter = GEN6_MAPFILTER_NEAREST; |
break; |
case SAMPLER_FILTER_BILINEAR: |
sampler_state->ss0.min_filter = GEN6_MAPFILTER_LINEAR; |
sampler_state->ss0.mag_filter = GEN6_MAPFILTER_LINEAR; |
break; |
} |
switch (extend) { |
default: |
case SAMPLER_EXTEND_NONE: |
sampler_state->ss1.r_wrap_mode = GEN6_TEXCOORDMODE_CLAMP_BORDER; |
sampler_state->ss1.s_wrap_mode = GEN6_TEXCOORDMODE_CLAMP_BORDER; |
sampler_state->ss1.t_wrap_mode = GEN6_TEXCOORDMODE_CLAMP_BORDER; |
break; |
case SAMPLER_EXTEND_REPEAT: |
sampler_state->ss1.r_wrap_mode = GEN6_TEXCOORDMODE_WRAP; |
sampler_state->ss1.s_wrap_mode = GEN6_TEXCOORDMODE_WRAP; |
sampler_state->ss1.t_wrap_mode = GEN6_TEXCOORDMODE_WRAP; |
break; |
case SAMPLER_EXTEND_PAD: |
sampler_state->ss1.r_wrap_mode = GEN6_TEXCOORDMODE_CLAMP; |
sampler_state->ss1.s_wrap_mode = GEN6_TEXCOORDMODE_CLAMP; |
sampler_state->ss1.t_wrap_mode = GEN6_TEXCOORDMODE_CLAMP; |
break; |
case SAMPLER_EXTEND_REFLECT: |
sampler_state->ss1.r_wrap_mode = GEN6_TEXCOORDMODE_MIRROR; |
sampler_state->ss1.s_wrap_mode = GEN6_TEXCOORDMODE_MIRROR; |
sampler_state->ss1.t_wrap_mode = GEN6_TEXCOORDMODE_MIRROR; |
break; |
} |
} |
static void |
sampler_copy_init(struct gen6_sampler_state *ss) |
{ |
sampler_state_init(ss, SAMPLER_FILTER_NEAREST, SAMPLER_EXTEND_NONE); |
ss->ss3.non_normalized_coord = 1; |
sampler_state_init(ss+1, SAMPLER_FILTER_NEAREST, SAMPLER_EXTEND_NONE); |
} |
static void |
sampler_fill_init(struct gen6_sampler_state *ss) |
{ |
sampler_state_init(ss, SAMPLER_FILTER_NEAREST, SAMPLER_EXTEND_REPEAT); |
ss->ss3.non_normalized_coord = 1; |
sampler_state_init(ss+1, SAMPLER_FILTER_NEAREST, SAMPLER_EXTEND_NONE); |
} |
static uint32_t |
gen6_tiling_bits(uint32_t tiling) |
{ |
switch (tiling) { |
default: assert(0); |
case I915_TILING_NONE: return 0; |
case I915_TILING_X: return GEN6_SURFACE_TILED; |
case I915_TILING_Y: return GEN6_SURFACE_TILED | GEN6_SURFACE_TILED_Y; |
} |
} |
/** |
* Sets up the common fields for a surface state buffer for the given |
* picture in the given surface state buffer. |
*/ |
static int |
gen6_bind_bo(struct sna *sna, |
struct kgem_bo *bo, |
uint32_t width, |
uint32_t height, |
uint32_t format, |
bool is_dst) |
{ |
uint32_t *ss; |
uint32_t domains; |
uint16_t offset; |
uint32_t is_scanout = is_dst && bo->scanout; |
/* After the first bind, we manage the cache domains within the batch */ |
offset = kgem_bo_get_binding(bo, format | is_dst << 30 | is_scanout << 31); |
if (offset) { |
DBG(("[%x] bo(handle=%d), format=%d, reuse %s binding\n", |
offset, bo->handle, format, |
is_dst ? "render" : "sampler")); |
if (is_dst) |
kgem_bo_mark_dirty(bo); |
return offset * sizeof(uint32_t); |
} |
offset = sna->kgem.surface -= |
sizeof(struct gen6_surface_state_padded) / sizeof(uint32_t); |
ss = sna->kgem.batch + offset; |
ss[0] = (GEN6_SURFACE_2D << GEN6_SURFACE_TYPE_SHIFT | |
GEN6_SURFACE_BLEND_ENABLED | |
format << GEN6_SURFACE_FORMAT_SHIFT); |
if (is_dst) { |
ss[0] |= GEN6_SURFACE_RC_READ_WRITE; |
domains = I915_GEM_DOMAIN_RENDER << 16 |I915_GEM_DOMAIN_RENDER; |
} else |
domains = I915_GEM_DOMAIN_SAMPLER << 16; |
ss[1] = kgem_add_reloc(&sna->kgem, offset + 1, bo, domains, 0); |
ss[2] = ((width - 1) << GEN6_SURFACE_WIDTH_SHIFT | |
(height - 1) << GEN6_SURFACE_HEIGHT_SHIFT); |
assert(bo->pitch <= (1 << 18)); |
ss[3] = (gen6_tiling_bits(bo->tiling) | |
(bo->pitch - 1) << GEN6_SURFACE_PITCH_SHIFT); |
ss[4] = 0; |
ss[5] = (is_scanout || bo->io) ? 0 : 3 << 16; |
kgem_bo_set_binding(bo, format | is_dst << 30 | is_scanout << 31, offset); |
DBG(("[%x] bind bo(handle=%d, addr=%d), format=%d, width=%d, height=%d, pitch=%d, tiling=%d -> %s\n", |
offset, bo->handle, ss[1], |
format, width, height, bo->pitch, bo->tiling, |
domains & 0xffff ? "render" : "sampler")); |
return offset * sizeof(uint32_t); |
} |
static void gen6_emit_vertex_buffer(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
int id = GEN6_VERTEX(op->u.gen6.flags); |
OUT_BATCH(GEN6_3DSTATE_VERTEX_BUFFERS | 3); |
OUT_BATCH(id << VB0_BUFFER_INDEX_SHIFT | VB0_VERTEXDATA | |
4*op->floats_per_vertex << VB0_BUFFER_PITCH_SHIFT); |
sna->render.vertex_reloc[sna->render.nvertex_reloc++] = sna->kgem.nbatch; |
OUT_BATCH(0); |
OUT_BATCH(~0); /* max address: disabled */ |
OUT_BATCH(0); |
sna->render.vb_id |= 1 << id; |
} |
static void gen6_emit_primitive(struct sna *sna) |
{ |
if (sna->kgem.nbatch == sna->render_state.gen6.last_primitive) { |
DBG(("%s: continuing previous primitive, start=%d, index=%d\n", |
__FUNCTION__, |
sna->render.vertex_start, |
sna->render.vertex_index)); |
sna->render.vertex_offset = sna->kgem.nbatch - 5; |
return; |
} |
OUT_BATCH(GEN6_3DPRIMITIVE | |
GEN6_3DPRIMITIVE_VERTEX_SEQUENTIAL | |
_3DPRIM_RECTLIST << GEN6_3DPRIMITIVE_TOPOLOGY_SHIFT | |
0 << 9 | |
4); |
sna->render.vertex_offset = sna->kgem.nbatch; |
OUT_BATCH(0); /* vertex count, to be filled in later */ |
OUT_BATCH(sna->render.vertex_index); |
OUT_BATCH(1); /* single instance */ |
OUT_BATCH(0); /* start instance location */ |
OUT_BATCH(0); /* index buffer offset, ignored */ |
sna->render.vertex_start = sna->render.vertex_index; |
DBG(("%s: started new primitive: index=%d\n", |
__FUNCTION__, sna->render.vertex_start)); |
sna->render_state.gen6.last_primitive = sna->kgem.nbatch; |
} |
static bool gen6_rectangle_begin(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
int id = 1 << GEN6_VERTEX(op->u.gen6.flags); |
int ndwords; |
if (sna_vertex_wait__locked(&sna->render) && sna->render.vertex_offset) |
return true; |
ndwords = op->need_magic_ca_pass ? 60 : 6; |
if ((sna->render.vb_id & id) == 0) |
ndwords += 5; |
if (!kgem_check_batch(&sna->kgem, ndwords)) |
return false; |
if ((sna->render.vb_id & id) == 0) |
gen6_emit_vertex_buffer(sna, op); |
gen6_emit_primitive(sna); |
return true; |
} |
static int gen6_get_rectangles__flush(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
/* Preventing discarding new vbo after lock contention */ |
if (sna_vertex_wait__locked(&sna->render)) { |
int rem = vertex_space(sna); |
if (rem > op->floats_per_rect) |
return rem; |
} |
if (!kgem_check_batch(&sna->kgem, op->need_magic_ca_pass ? 65 : 5)) |
return 0; |
if (!kgem_check_reloc_and_exec(&sna->kgem, 2)) |
return 0; |
if (sna->render.vertex_offset) { |
gen4_vertex_flush(sna); |
if (gen6_magic_ca_pass(sna, op)) { |
gen6_emit_flush(sna); |
gen6_emit_cc(sna, GEN6_BLEND(op->u.gen6.flags)); |
gen6_emit_wm(sna, |
GEN6_KERNEL(op->u.gen6.flags), |
GEN6_VERTEX(op->u.gen6.flags) >> 2); |
} |
} |
return gen4_vertex_finish(sna); |
} |
inline static int gen6_get_rectangles(struct sna *sna, |
const struct sna_composite_op *op, |
int want, |
void (*emit_state)(struct sna *, const struct sna_composite_op *op)) |
{ |
int rem; |
assert(want); |
start: |
rem = vertex_space(sna); |
if (unlikely(rem < op->floats_per_rect)) { |
DBG(("flushing vbo for %s: %d < %d\n", |
__FUNCTION__, rem, op->floats_per_rect)); |
rem = gen6_get_rectangles__flush(sna, op); |
if (unlikely(rem == 0)) |
goto flush; |
} |
if (unlikely(sna->render.vertex_offset == 0)) { |
if (!gen6_rectangle_begin(sna, op)) |
goto flush; |
else |
goto start; |
} |
assert(rem <= vertex_space(sna)); |
assert(op->floats_per_rect <= rem); |
if (want > 1 && want * op->floats_per_rect > rem) |
want = rem / op->floats_per_rect; |
assert(want > 0); |
sna->render.vertex_index += 3*want; |
return want; |
flush: |
if (sna->render.vertex_offset) { |
gen4_vertex_flush(sna); |
gen6_magic_ca_pass(sna, op); |
} |
sna_vertex_wait__locked(&sna->render); |
_kgem_submit(&sna->kgem); |
emit_state(sna, op); |
goto start; |
} |
inline static uint32_t *gen6_composite_get_binding_table(struct sna *sna, |
uint16_t *offset) |
{ |
uint32_t *table; |
sna->kgem.surface -= |
sizeof(struct gen6_surface_state_padded) / sizeof(uint32_t); |
/* Clear all surplus entries to zero in case of prefetch */ |
table = memset(sna->kgem.batch + sna->kgem.surface, |
0, sizeof(struct gen6_surface_state_padded)); |
DBG(("%s(%x)\n", __FUNCTION__, 4*sna->kgem.surface)); |
*offset = sna->kgem.surface; |
return table; |
} |
static bool |
gen6_get_batch(struct sna *sna, const struct sna_composite_op *op) |
{ |
kgem_set_mode(&sna->kgem, KGEM_RENDER, op->dst.bo); |
if (!kgem_check_batch_with_surfaces(&sna->kgem, 150, 4)) { |
DBG(("%s: flushing batch: %d < %d+%d\n", |
__FUNCTION__, sna->kgem.surface - sna->kgem.nbatch, |
150, 4*8)); |
kgem_submit(&sna->kgem); |
_kgem_set_mode(&sna->kgem, KGEM_RENDER); |
} |
if (sna->render_state.gen6.needs_invariant) |
gen6_emit_invariant(sna); |
return kgem_bo_is_dirty(op->dst.bo); |
} |
static void gen6_emit_composite_state(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
uint32_t *binding_table; |
uint16_t offset; |
bool dirty; |
dirty = gen6_get_batch(sna, op); |
binding_table = gen6_composite_get_binding_table(sna, &offset); |
binding_table[0] = |
gen6_bind_bo(sna, |
op->dst.bo, op->dst.width, op->dst.height, |
gen6_get_dest_format(op->dst.format), |
true); |
binding_table[1] = |
gen6_bind_bo(sna, |
op->src.bo, op->src.width, op->src.height, |
op->src.card_format, |
false); |
if (op->mask.bo) { |
binding_table[2] = |
gen6_bind_bo(sna, |
op->mask.bo, |
op->mask.width, |
op->mask.height, |
op->mask.card_format, |
false); |
} |
if (sna->kgem.surface == offset && |
*(uint64_t *)(sna->kgem.batch + sna->render_state.gen6.surface_table) == *(uint64_t*)binding_table && |
(op->mask.bo == NULL || |
sna->kgem.batch[sna->render_state.gen6.surface_table+2] == binding_table[2])) { |
sna->kgem.surface += sizeof(struct gen6_surface_state_padded) / sizeof(uint32_t); |
offset = sna->render_state.gen6.surface_table; |
} |
gen6_emit_state(sna, op, offset | dirty); |
} |
static void |
gen6_align_vertex(struct sna *sna, const struct sna_composite_op *op) |
{ |
assert (sna->render.vertex_offset == 0); |
if (op->floats_per_vertex != sna->render_state.gen6.floats_per_vertex) { |
if (sna->render.vertex_size - sna->render.vertex_used < 2*op->floats_per_rect) |
gen4_vertex_finish(sna); |
DBG(("aligning vertex: was %d, now %d floats per vertex, %d->%d\n", |
sna->render_state.gen6.floats_per_vertex, |
op->floats_per_vertex, |
sna->render.vertex_index, |
(sna->render.vertex_used + op->floats_per_vertex - 1) / op->floats_per_vertex)); |
sna->render.vertex_index = (sna->render.vertex_used + op->floats_per_vertex - 1) / op->floats_per_vertex; |
sna->render.vertex_used = sna->render.vertex_index * op->floats_per_vertex; |
sna->render_state.gen6.floats_per_vertex = op->floats_per_vertex; |
} |
assert((sna->render.vertex_used % op->floats_per_vertex) == 0); |
} |
fastcall static void |
gen6_render_composite_blt(struct sna *sna, |
const struct sna_composite_op *op, |
const struct sna_composite_rectangles *r) |
{ |
gen6_get_rectangles(sna, op, 1, gen6_emit_composite_state); |
op->prim_emit(sna, op, r); |
} |
#if 0 |
fastcall static void |
gen6_render_composite_box(struct sna *sna, |
const struct sna_composite_op *op, |
const BoxRec *box) |
{ |
struct sna_composite_rectangles r; |
gen6_get_rectangles(sna, op, 1, gen6_emit_composite_state); |
DBG((" %s: (%d, %d), (%d, %d)\n", |
__FUNCTION__, |
box->x1, box->y1, box->x2, box->y2)); |
r.dst.x = box->x1; |
r.dst.y = box->y1; |
r.width = box->x2 - box->x1; |
r.height = box->y2 - box->y1; |
r.src = r.mask = r.dst; |
op->prim_emit(sna, op, &r); |
} |
static void |
gen6_render_composite_boxes__blt(struct sna *sna, |
const struct sna_composite_op *op, |
const BoxRec *box, int nbox) |
{ |
DBG(("composite_boxes(%d)\n", nbox)); |
do { |
int nbox_this_time; |
nbox_this_time = gen6_get_rectangles(sna, op, nbox, |
gen6_emit_composite_state); |
nbox -= nbox_this_time; |
do { |
struct sna_composite_rectangles r; |
DBG((" %s: (%d, %d), (%d, %d)\n", |
__FUNCTION__, |
box->x1, box->y1, box->x2, box->y2)); |
r.dst.x = box->x1; |
r.dst.y = box->y1; |
r.width = box->x2 - box->x1; |
r.height = box->y2 - box->y1; |
r.src = r.mask = r.dst; |
op->prim_emit(sna, op, &r); |
box++; |
} while (--nbox_this_time); |
} while (nbox); |
} |
static void |
gen6_render_composite_boxes(struct sna *sna, |
const struct sna_composite_op *op, |
const BoxRec *box, int nbox) |
{ |
DBG(("%s: nbox=%d\n", __FUNCTION__, nbox)); |
do { |
int nbox_this_time; |
float *v; |
nbox_this_time = gen6_get_rectangles(sna, op, nbox, |
gen6_emit_composite_state); |
assert(nbox_this_time); |
nbox -= nbox_this_time; |
v = sna->render.vertices + sna->render.vertex_used; |
sna->render.vertex_used += nbox_this_time * op->floats_per_rect; |
op->emit_boxes(op, box, nbox_this_time, v); |
box += nbox_this_time; |
} while (nbox); |
} |
static void |
gen6_render_composite_boxes__thread(struct sna *sna, |
const struct sna_composite_op *op, |
const BoxRec *box, int nbox) |
{ |
DBG(("%s: nbox=%d\n", __FUNCTION__, nbox)); |
sna_vertex_lock(&sna->render); |
do { |
int nbox_this_time; |
float *v; |
nbox_this_time = gen6_get_rectangles(sna, op, nbox, |
gen6_emit_composite_state); |
assert(nbox_this_time); |
nbox -= nbox_this_time; |
v = sna->render.vertices + sna->render.vertex_used; |
sna->render.vertex_used += nbox_this_time * op->floats_per_rect; |
sna_vertex_acquire__locked(&sna->render); |
sna_vertex_unlock(&sna->render); |
op->emit_boxes(op, box, nbox_this_time, v); |
box += nbox_this_time; |
sna_vertex_lock(&sna->render); |
sna_vertex_release__locked(&sna->render); |
} while (nbox); |
sna_vertex_unlock(&sna->render); |
} |
#endif |
#ifndef MAX |
#define MAX(a,b) ((a) > (b) ? (a) : (b)) |
#endif |
static uint32_t |
gen6_composite_create_blend_state(struct sna_static_stream *stream) |
{ |
char *base, *ptr; |
int src, dst; |
base = sna_static_stream_map(stream, |
GEN6_BLENDFACTOR_COUNT * GEN6_BLENDFACTOR_COUNT * GEN6_BLEND_STATE_PADDED_SIZE, |
64); |
ptr = base; |
for (src = 0; src < GEN6_BLENDFACTOR_COUNT; src++) { |
for (dst= 0; dst < GEN6_BLENDFACTOR_COUNT; dst++) { |
struct gen6_blend_state *blend = |
(struct gen6_blend_state *)ptr; |
blend->blend0.dest_blend_factor = dst; |
blend->blend0.source_blend_factor = src; |
blend->blend0.blend_func = GEN6_BLENDFUNCTION_ADD; |
blend->blend0.blend_enable = |
!(dst == GEN6_BLENDFACTOR_ZERO && src == GEN6_BLENDFACTOR_ONE); |
blend->blend1.post_blend_clamp_enable = 1; |
blend->blend1.pre_blend_clamp_enable = 1; |
ptr += GEN6_BLEND_STATE_PADDED_SIZE; |
} |
} |
return sna_static_stream_offsetof(stream, base); |
} |
#if 0 |
static uint32_t gen6_bind_video_source(struct sna *sna, |
struct kgem_bo *src_bo, |
uint32_t src_offset, |
int src_width, |
int src_height, |
int src_pitch, |
uint32_t src_surf_format) |
{ |
struct gen6_surface_state *ss; |
sna->kgem.surface -= sizeof(struct gen6_surface_state_padded) / sizeof(uint32_t); |
ss = memset(sna->kgem.batch + sna->kgem.surface, 0, sizeof(*ss)); |
ss->ss0.surface_type = GEN6_SURFACE_2D; |
ss->ss0.surface_format = src_surf_format; |
ss->ss1.base_addr = |
kgem_add_reloc(&sna->kgem, |
sna->kgem.surface + 1, |
src_bo, |
I915_GEM_DOMAIN_SAMPLER << 16, |
src_offset); |
ss->ss2.width = src_width - 1; |
ss->ss2.height = src_height - 1; |
ss->ss3.pitch = src_pitch - 1; |
return sna->kgem.surface * sizeof(uint32_t); |
} |
static void gen6_emit_video_state(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
struct sna_video_frame *frame = op->priv; |
uint32_t src_surf_format; |
uint32_t src_surf_base[6]; |
int src_width[6]; |
int src_height[6]; |
int src_pitch[6]; |
uint32_t *binding_table; |
uint16_t offset; |
bool dirty; |
int n_src, n; |
dirty = gen6_get_batch(sna, op); |
src_surf_base[0] = 0; |
src_surf_base[1] = 0; |
src_surf_base[2] = frame->VBufOffset; |
src_surf_base[3] = frame->VBufOffset; |
src_surf_base[4] = frame->UBufOffset; |
src_surf_base[5] = frame->UBufOffset; |
if (is_planar_fourcc(frame->id)) { |
src_surf_format = GEN6_SURFACEFORMAT_R8_UNORM; |
src_width[1] = src_width[0] = frame->width; |
src_height[1] = src_height[0] = frame->height; |
src_pitch[1] = src_pitch[0] = frame->pitch[1]; |
src_width[4] = src_width[5] = src_width[2] = src_width[3] = |
frame->width / 2; |
src_height[4] = src_height[5] = src_height[2] = src_height[3] = |
frame->height / 2; |
src_pitch[4] = src_pitch[5] = src_pitch[2] = src_pitch[3] = |
frame->pitch[0]; |
n_src = 6; |
} else { |
if (frame->id == FOURCC_UYVY) |
src_surf_format = GEN6_SURFACEFORMAT_YCRCB_SWAPY; |
else |
src_surf_format = GEN6_SURFACEFORMAT_YCRCB_NORMAL; |
src_width[0] = frame->width; |
src_height[0] = frame->height; |
src_pitch[0] = frame->pitch[0]; |
n_src = 1; |
} |
binding_table = gen6_composite_get_binding_table(sna, &offset); |
binding_table[0] = |
gen6_bind_bo(sna, |
op->dst.bo, op->dst.width, op->dst.height, |
gen6_get_dest_format(op->dst.format), |
true); |
for (n = 0; n < n_src; n++) { |
binding_table[1+n] = |
gen6_bind_video_source(sna, |
frame->bo, |
src_surf_base[n], |
src_width[n], |
src_height[n], |
src_pitch[n], |
src_surf_format); |
} |
gen6_emit_state(sna, op, offset | dirty); |
} |
static bool |
gen6_render_video(struct sna *sna, |
struct sna_video *video, |
struct sna_video_frame *frame, |
RegionPtr dstRegion, |
PixmapPtr pixmap) |
{ |
struct sna_composite_op tmp; |
int dst_width = dstRegion->extents.x2 - dstRegion->extents.x1; |
int dst_height = dstRegion->extents.y2 - dstRegion->extents.y1; |
int src_width = frame->src.x2 - frame->src.x1; |
int src_height = frame->src.y2 - frame->src.y1; |
float src_offset_x, src_offset_y; |
float src_scale_x, src_scale_y; |
int nbox, pix_xoff, pix_yoff; |
struct sna_pixmap *priv; |
unsigned filter; |
BoxPtr box; |
DBG(("%s: src=(%d, %d), dst=(%d, %d), %ldx[(%d, %d), (%d, %d)...]\n", |
__FUNCTION__, |
src_width, src_height, dst_width, dst_height, |
(long)REGION_NUM_RECTS(dstRegion), |
REGION_EXTENTS(NULL, dstRegion)->x1, |
REGION_EXTENTS(NULL, dstRegion)->y1, |
REGION_EXTENTS(NULL, dstRegion)->x2, |
REGION_EXTENTS(NULL, dstRegion)->y2)); |
priv = sna_pixmap_force_to_gpu(pixmap, MOVE_READ | MOVE_WRITE); |
if (priv == NULL) |
return false; |
memset(&tmp, 0, sizeof(tmp)); |
tmp.dst.pixmap = pixmap; |
tmp.dst.width = pixmap->drawable.width; |
tmp.dst.height = pixmap->drawable.height; |
tmp.dst.format = sna_render_format_for_depth(pixmap->drawable.depth); |
tmp.dst.bo = priv->gpu_bo; |
tmp.src.bo = frame->bo; |
tmp.mask.bo = NULL; |
tmp.floats_per_vertex = 3; |
tmp.floats_per_rect = 9; |
if (src_width == dst_width && src_height == dst_height) |
filter = SAMPLER_FILTER_NEAREST; |
else |
filter = SAMPLER_FILTER_BILINEAR; |
tmp.u.gen6.flags = |
GEN6_SET_FLAGS(SAMPLER_OFFSET(filter, SAMPLER_EXTEND_PAD, |
SAMPLER_FILTER_NEAREST, SAMPLER_EXTEND_NONE), |
NO_BLEND, |
is_planar_fourcc(frame->id) ? |
GEN6_WM_KERNEL_VIDEO_PLANAR : |
GEN6_WM_KERNEL_VIDEO_PACKED, |
2); |
tmp.priv = frame; |
kgem_set_mode(&sna->kgem, KGEM_RENDER, tmp.dst.bo); |
if (!kgem_check_bo(&sna->kgem, tmp.dst.bo, frame->bo, NULL)) { |
kgem_submit(&sna->kgem); |
assert(kgem_check_bo(&sna->kgem, tmp.dst.bo, frame->bo, NULL)); |
_kgem_set_mode(&sna->kgem, KGEM_RENDER); |
} |
gen6_emit_video_state(sna, &tmp); |
gen6_align_vertex(sna, &tmp); |
/* Set up the offset for translating from the given region (in screen |
* coordinates) to the backing pixmap. |
*/ |
#ifdef COMPOSITE |
pix_xoff = -pixmap->screen_x + pixmap->drawable.x; |
pix_yoff = -pixmap->screen_y + pixmap->drawable.y; |
#else |
pix_xoff = 0; |
pix_yoff = 0; |
#endif |
src_scale_x = (float)src_width / dst_width / frame->width; |
src_offset_x = (float)frame->src.x1 / frame->width - dstRegion->extents.x1 * src_scale_x; |
src_scale_y = (float)src_height / dst_height / frame->height; |
src_offset_y = (float)frame->src.y1 / frame->height - dstRegion->extents.y1 * src_scale_y; |
box = REGION_RECTS(dstRegion); |
nbox = REGION_NUM_RECTS(dstRegion); |
while (nbox--) { |
BoxRec r; |
r.x1 = box->x1 + pix_xoff; |
r.x2 = box->x2 + pix_xoff; |
r.y1 = box->y1 + pix_yoff; |
r.y2 = box->y2 + pix_yoff; |
gen6_get_rectangles(sna, &tmp, 1, gen6_emit_video_state); |
OUT_VERTEX(r.x2, r.y2); |
OUT_VERTEX_F(box->x2 * src_scale_x + src_offset_x); |
OUT_VERTEX_F(box->y2 * src_scale_y + src_offset_y); |
OUT_VERTEX(r.x1, r.y2); |
OUT_VERTEX_F(box->x1 * src_scale_x + src_offset_x); |
OUT_VERTEX_F(box->y2 * src_scale_y + src_offset_y); |
OUT_VERTEX(r.x1, r.y1); |
OUT_VERTEX_F(box->x1 * src_scale_x + src_offset_x); |
OUT_VERTEX_F(box->y1 * src_scale_y + src_offset_y); |
if (!DAMAGE_IS_ALL(priv->gpu_damage)) { |
sna_damage_add_box(&priv->gpu_damage, &r); |
sna_damage_subtract_box(&priv->cpu_damage, &r); |
} |
box++; |
} |
gen4_vertex_flush(sna); |
return true; |
} |
static int |
gen6_composite_picture(struct sna *sna, |
PicturePtr picture, |
struct sna_composite_channel *channel, |
int x, int y, |
int w, int h, |
int dst_x, int dst_y, |
bool precise) |
{ |
PixmapPtr pixmap; |
uint32_t color; |
int16_t dx, dy; |
DBG(("%s: (%d, %d)x(%d, %d), dst=(%d, %d)\n", |
__FUNCTION__, x, y, w, h, dst_x, dst_y)); |
channel->is_solid = false; |
channel->card_format = -1; |
if (sna_picture_is_solid(picture, &color)) |
return gen4_channel_init_solid(sna, channel, color); |
if (picture->pDrawable == NULL) { |
int ret; |
if (picture->pSourcePict->type == SourcePictTypeLinear) |
return gen4_channel_init_linear(sna, picture, channel, |
x, y, |
w, h, |
dst_x, dst_y); |
DBG(("%s -- fixup, gradient\n", __FUNCTION__)); |
ret = -1; |
if (!precise) |
ret = sna_render_picture_approximate_gradient(sna, picture, channel, |
x, y, w, h, dst_x, dst_y); |
if (ret == -1) |
ret = sna_render_picture_fixup(sna, picture, channel, |
x, y, w, h, dst_x, dst_y); |
return ret; |
} |
if (picture->alphaMap) { |
DBG(("%s -- fixup, alphamap\n", __FUNCTION__)); |
return sna_render_picture_fixup(sna, picture, channel, |
x, y, w, h, dst_x, dst_y); |
} |
if (!gen6_check_repeat(picture)) |
return sna_render_picture_fixup(sna, picture, channel, |
x, y, w, h, dst_x, dst_y); |
if (!gen6_check_filter(picture)) |
return sna_render_picture_fixup(sna, picture, channel, |
x, y, w, h, dst_x, dst_y); |
channel->repeat = picture->repeat ? picture->repeatType : RepeatNone; |
channel->filter = picture->filter; |
pixmap = get_drawable_pixmap(picture->pDrawable); |
get_drawable_deltas(picture->pDrawable, pixmap, &dx, &dy); |
x += dx + picture->pDrawable->x; |
y += dy + picture->pDrawable->y; |
channel->is_affine = sna_transform_is_affine(picture->transform); |
if (sna_transform_is_integer_translation(picture->transform, &dx, &dy)) { |
DBG(("%s: integer translation (%d, %d), removing\n", |
__FUNCTION__, dx, dy)); |
x += dx; |
y += dy; |
channel->transform = NULL; |
channel->filter = PictFilterNearest; |
} else |
channel->transform = picture->transform; |
channel->pict_format = picture->format; |
channel->card_format = gen6_get_card_format(picture->format); |
if (channel->card_format == (unsigned)-1) |
return sna_render_picture_convert(sna, picture, channel, pixmap, |
x, y, w, h, dst_x, dst_y, |
false); |
if (too_large(pixmap->drawable.width, pixmap->drawable.height)) { |
DBG(("%s: extracting from pixmap %dx%d\n", __FUNCTION__, |
pixmap->drawable.width, pixmap->drawable.height)); |
return sna_render_picture_extract(sna, picture, channel, |
x, y, w, h, dst_x, dst_y); |
} |
return sna_render_pixmap_bo(sna, channel, pixmap, |
x, y, w, h, dst_x, dst_y); |
} |
inline static void gen6_composite_channel_convert(struct sna_composite_channel *channel) |
{ |
channel->repeat = gen6_repeat(channel->repeat); |
channel->filter = gen6_filter(channel->filter); |
if (channel->card_format == (unsigned)-1) |
channel->card_format = gen6_get_card_format(channel->pict_format); |
assert(channel->card_format != (unsigned)-1); |
} |
#endif |
static void gen6_render_composite_done(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
DBG(("%s\n", __FUNCTION__)); |
assert(!sna->render.active); |
if (sna->render.vertex_offset) { |
gen4_vertex_flush(sna); |
gen6_magic_ca_pass(sna, op); |
} |
} |
#if 0 |
static bool |
gen6_composite_set_target(struct sna *sna, |
struct sna_composite_op *op, |
PicturePtr dst, |
int x, int y, int w, int h, |
bool partial) |
{ |
BoxRec box; |
op->dst.pixmap = get_drawable_pixmap(dst->pDrawable); |
op->dst.format = dst->format; |
op->dst.width = op->dst.pixmap->drawable.width; |
op->dst.height = op->dst.pixmap->drawable.height; |
if (w && h) { |
box.x1 = x; |
box.y1 = y; |
box.x2 = x + w; |
box.y2 = y + h; |
} else |
sna_render_picture_extents(dst, &box); |
// op->dst.bo = sna_drawable_use_bo (dst->pDrawable, |
// PREFER_GPU | FORCE_GPU | RENDER_GPU, |
// &box, &op->damage); |
if (op->dst.bo == NULL) |
return false; |
get_drawable_deltas(dst->pDrawable, op->dst.pixmap, |
&op->dst.x, &op->dst.y); |
DBG(("%s: pixmap=%p, format=%08x, size=%dx%d, pitch=%d, delta=(%d,%d),damage=%p\n", |
__FUNCTION__, |
op->dst.pixmap, (int)op->dst.format, |
op->dst.width, op->dst.height, |
op->dst.bo->pitch, |
op->dst.x, op->dst.y, |
op->damage ? *op->damage : (void *)-1)); |
assert(op->dst.bo->proxy == NULL); |
if (too_large(op->dst.width, op->dst.height) && |
!sna_render_composite_redirect(sna, op, x, y, w, h)) |
return false; |
return true; |
} |
static bool |
prefer_blt_composite(struct sna *sna, struct sna_composite_op *tmp) |
{ |
if (untiled_tlb_miss(tmp->dst.bo) || |
untiled_tlb_miss(tmp->src.bo)) |
return true; |
if (kgem_bo_is_render(tmp->dst.bo) || |
kgem_bo_is_render(tmp->src.bo)) |
return false; |
if (!prefer_blt_ring(sna, tmp->dst.bo, 0)) |
return false; |
return prefer_blt_bo(sna, tmp->dst.bo) || prefer_blt_bo(sna, tmp->src.bo); |
} |
static bool |
gen6_render_composite(struct sna *sna, |
uint8_t op, |
PicturePtr src, |
PicturePtr mask, |
PicturePtr dst, |
int16_t src_x, int16_t src_y, |
int16_t msk_x, int16_t msk_y, |
int16_t dst_x, int16_t dst_y, |
int16_t width, int16_t height, |
struct sna_composite_op *tmp) |
{ |
if (op >= ARRAY_SIZE(gen6_blend_op)) |
return false; |
DBG(("%s: %dx%d, current mode=%d\n", __FUNCTION__, |
width, height, sna->kgem.ring)); |
if (mask == NULL && |
try_blt(sna, dst, src, width, height) && |
sna_blt_composite(sna, op, |
src, dst, |
src_x, src_y, |
dst_x, dst_y, |
width, height, |
tmp, false)) |
return true; |
if (gen6_composite_fallback(sna, src, mask, dst)) |
return false; |
if (need_tiling(sna, width, height)) |
return sna_tiling_composite(op, src, mask, dst, |
src_x, src_y, |
msk_x, msk_y, |
dst_x, dst_y, |
width, height, |
tmp); |
if (op == PictOpClear) |
op = PictOpSrc; |
tmp->op = op; |
if (!gen6_composite_set_target(sna, tmp, dst, |
dst_x, dst_y, width, height, |
op > PictOpSrc || dst->pCompositeClip->data)) |
return false; |
switch (gen6_composite_picture(sna, src, &tmp->src, |
src_x, src_y, |
width, height, |
dst_x, dst_y, |
dst->polyMode == PolyModePrecise)) { |
case -1: |
goto cleanup_dst; |
case 0: |
if (!gen4_channel_init_solid(sna, &tmp->src, 0)) |
goto cleanup_dst; |
/* fall through to fixup */ |
case 1: |
/* Did we just switch rings to prepare the source? */ |
if (mask == NULL && |
prefer_blt_composite(sna, tmp) && |
sna_blt_composite__convert(sna, |
dst_x, dst_y, width, height, |
tmp)) |
return true; |
gen6_composite_channel_convert(&tmp->src); |
break; |
} |
tmp->is_affine = tmp->src.is_affine; |
tmp->has_component_alpha = false; |
tmp->need_magic_ca_pass = false; |
tmp->mask.bo = NULL; |
tmp->mask.filter = SAMPLER_FILTER_NEAREST; |
tmp->mask.repeat = SAMPLER_EXTEND_NONE; |
if (mask) { |
if (mask->componentAlpha && PICT_FORMAT_RGB(mask->format)) { |
tmp->has_component_alpha = true; |
/* Check if it's component alpha that relies on a source alpha and on |
* the source value. We can only get one of those into the single |
* source value that we get to blend with. |
*/ |
if (gen6_blend_op[op].src_alpha && |
(gen6_blend_op[op].src_blend != GEN6_BLENDFACTOR_ZERO)) { |
if (op != PictOpOver) |
goto cleanup_src; |
tmp->need_magic_ca_pass = true; |
tmp->op = PictOpOutReverse; |
} |
} |
if (!reuse_source(sna, |
src, &tmp->src, src_x, src_y, |
mask, &tmp->mask, msk_x, msk_y)) { |
switch (gen6_composite_picture(sna, mask, &tmp->mask, |
msk_x, msk_y, |
width, height, |
dst_x, dst_y, |
dst->polyMode == PolyModePrecise)) { |
case -1: |
goto cleanup_src; |
case 0: |
if (!gen4_channel_init_solid(sna, &tmp->mask, 0)) |
goto cleanup_src; |
/* fall through to fixup */ |
case 1: |
gen6_composite_channel_convert(&tmp->mask); |
break; |
} |
} |
tmp->is_affine &= tmp->mask.is_affine; |
} |
tmp->u.gen6.flags = |
GEN6_SET_FLAGS(SAMPLER_OFFSET(tmp->src.filter, |
tmp->src.repeat, |
tmp->mask.filter, |
tmp->mask.repeat), |
gen6_get_blend(tmp->op, |
tmp->has_component_alpha, |
tmp->dst.format), |
gen6_choose_composite_kernel(tmp->op, |
tmp->mask.bo != NULL, |
tmp->has_component_alpha, |
tmp->is_affine), |
gen4_choose_composite_emitter(sna, tmp)); |
tmp->blt = gen6_render_composite_blt; |
tmp->box = gen6_render_composite_box; |
tmp->boxes = gen6_render_composite_boxes__blt; |
if (tmp->emit_boxes) { |
tmp->boxes = gen6_render_composite_boxes; |
tmp->thread_boxes = gen6_render_composite_boxes__thread; |
} |
tmp->done = gen6_render_composite_done; |
kgem_set_mode(&sna->kgem, KGEM_RENDER, tmp->dst.bo); |
if (!kgem_check_bo(&sna->kgem, |
tmp->dst.bo, tmp->src.bo, tmp->mask.bo, |
NULL)) { |
kgem_submit(&sna->kgem); |
if (!kgem_check_bo(&sna->kgem, |
tmp->dst.bo, tmp->src.bo, tmp->mask.bo, |
NULL)) |
goto cleanup_mask; |
_kgem_set_mode(&sna->kgem, KGEM_RENDER); |
} |
gen6_emit_composite_state(sna, tmp); |
gen6_align_vertex(sna, tmp); |
return true; |
cleanup_mask: |
if (tmp->mask.bo) |
kgem_bo_destroy(&sna->kgem, tmp->mask.bo); |
cleanup_src: |
if (tmp->src.bo) |
kgem_bo_destroy(&sna->kgem, tmp->src.bo); |
cleanup_dst: |
if (tmp->redirect.real_bo) |
kgem_bo_destroy(&sna->kgem, tmp->dst.bo); |
return false; |
} |
#if !NO_COMPOSITE_SPANS |
fastcall static void |
gen6_render_composite_spans_box(struct sna *sna, |
const struct sna_composite_spans_op *op, |
const BoxRec *box, float opacity) |
{ |
DBG(("%s: src=+(%d, %d), opacity=%f, dst=+(%d, %d), box=(%d, %d) x (%d, %d)\n", |
__FUNCTION__, |
op->base.src.offset[0], op->base.src.offset[1], |
opacity, |
op->base.dst.x, op->base.dst.y, |
box->x1, box->y1, |
box->x2 - box->x1, |
box->y2 - box->y1)); |
gen6_get_rectangles(sna, &op->base, 1, gen6_emit_composite_state); |
op->prim_emit(sna, op, box, opacity); |
} |
static void |
gen6_render_composite_spans_boxes(struct sna *sna, |
const struct sna_composite_spans_op *op, |
const BoxRec *box, int nbox, |
float opacity) |
{ |
DBG(("%s: nbox=%d, src=+(%d, %d), opacity=%f, dst=+(%d, %d)\n", |
__FUNCTION__, nbox, |
op->base.src.offset[0], op->base.src.offset[1], |
opacity, |
op->base.dst.x, op->base.dst.y)); |
do { |
int nbox_this_time; |
nbox_this_time = gen6_get_rectangles(sna, &op->base, nbox, |
gen6_emit_composite_state); |
nbox -= nbox_this_time; |
do { |
DBG((" %s: (%d, %d) x (%d, %d)\n", __FUNCTION__, |
box->x1, box->y1, |
box->x2 - box->x1, |
box->y2 - box->y1)); |
op->prim_emit(sna, op, box++, opacity); |
} while (--nbox_this_time); |
} while (nbox); |
} |
fastcall static void |
gen6_render_composite_spans_boxes__thread(struct sna *sna, |
const struct sna_composite_spans_op *op, |
const struct sna_opacity_box *box, |
int nbox) |
{ |
DBG(("%s: nbox=%d, src=+(%d, %d), dst=+(%d, %d)\n", |
__FUNCTION__, nbox, |
op->base.src.offset[0], op->base.src.offset[1], |
op->base.dst.x, op->base.dst.y)); |
sna_vertex_lock(&sna->render); |
do { |
int nbox_this_time; |
float *v; |
nbox_this_time = gen6_get_rectangles(sna, &op->base, nbox, |
gen6_emit_composite_state); |
assert(nbox_this_time); |
nbox -= nbox_this_time; |
v = sna->render.vertices + sna->render.vertex_used; |
sna->render.vertex_used += nbox_this_time * op->base.floats_per_rect; |
sna_vertex_acquire__locked(&sna->render); |
sna_vertex_unlock(&sna->render); |
op->emit_boxes(op, box, nbox_this_time, v); |
box += nbox_this_time; |
sna_vertex_lock(&sna->render); |
sna_vertex_release__locked(&sna->render); |
} while (nbox); |
sna_vertex_unlock(&sna->render); |
} |
fastcall static void |
gen6_render_composite_spans_done(struct sna *sna, |
const struct sna_composite_spans_op *op) |
{ |
DBG(("%s()\n", __FUNCTION__)); |
assert(!sna->render.active); |
if (sna->render.vertex_offset) |
gen4_vertex_flush(sna); |
if (op->base.src.bo) |
kgem_bo_destroy(&sna->kgem, op->base.src.bo); |
sna_render_composite_redirect_done(sna, &op->base); |
} |
static bool |
gen6_check_composite_spans(struct sna *sna, |
uint8_t op, PicturePtr src, PicturePtr dst, |
int16_t width, int16_t height, |
unsigned flags) |
{ |
DBG(("%s: op=%d, width=%d, height=%d, flags=%x\n", |
__FUNCTION__, op, width, height, flags)); |
if (op >= ARRAY_SIZE(gen6_blend_op)) |
return false; |
if (gen6_composite_fallback(sna, src, NULL, dst)) { |
DBG(("%s: operation would fallback\n", __FUNCTION__)); |
return false; |
} |
if (need_tiling(sna, width, height) && |
!is_gpu(sna, dst->pDrawable, PREFER_GPU_SPANS)) { |
DBG(("%s: fallback, tiled operation not on GPU\n", |
__FUNCTION__)); |
return false; |
} |
if ((flags & COMPOSITE_SPANS_RECTILINEAR) == 0) { |
struct sna_pixmap *priv = sna_pixmap_from_drawable(dst->pDrawable); |
assert(priv); |
if (priv->cpu_bo && kgem_bo_is_busy(priv->cpu_bo)) |
return true; |
if (flags & COMPOSITE_SPANS_INPLACE_HINT) |
return false; |
return priv->gpu_bo && kgem_bo_is_busy(priv->gpu_bo); |
} |
return true; |
} |
static bool |
gen6_render_composite_spans(struct sna *sna, |
uint8_t op, |
PicturePtr src, |
PicturePtr dst, |
int16_t src_x, int16_t src_y, |
int16_t dst_x, int16_t dst_y, |
int16_t width, int16_t height, |
unsigned flags, |
struct sna_composite_spans_op *tmp) |
{ |
DBG(("%s: %dx%d with flags=%x, current mode=%d\n", __FUNCTION__, |
width, height, flags, sna->kgem.ring)); |
assert(gen6_check_composite_spans(sna, op, src, dst, width, height, flags)); |
if (need_tiling(sna, width, height)) { |
DBG(("%s: tiling, operation (%dx%d) too wide for pipeline\n", |
__FUNCTION__, width, height)); |
return sna_tiling_composite_spans(op, src, dst, |
src_x, src_y, dst_x, dst_y, |
width, height, flags, tmp); |
} |
tmp->base.op = op; |
if (!gen6_composite_set_target(sna, &tmp->base, dst, |
dst_x, dst_y, width, height, true)) |
return false; |
switch (gen6_composite_picture(sna, src, &tmp->base.src, |
src_x, src_y, |
width, height, |
dst_x, dst_y, |
dst->polyMode == PolyModePrecise)) { |
case -1: |
goto cleanup_dst; |
case 0: |
if (!gen4_channel_init_solid(sna, &tmp->base.src, 0)) |
goto cleanup_dst; |
/* fall through to fixup */ |
case 1: |
gen6_composite_channel_convert(&tmp->base.src); |
break; |
} |
tmp->base.mask.bo = NULL; |
tmp->base.is_affine = tmp->base.src.is_affine; |
tmp->base.need_magic_ca_pass = false; |
tmp->base.u.gen6.flags = |
GEN6_SET_FLAGS(SAMPLER_OFFSET(tmp->base.src.filter, |
tmp->base.src.repeat, |
SAMPLER_FILTER_NEAREST, |
SAMPLER_EXTEND_PAD), |
gen6_get_blend(tmp->base.op, false, tmp->base.dst.format), |
GEN6_WM_KERNEL_OPACITY | !tmp->base.is_affine, |
gen4_choose_spans_emitter(sna, tmp)); |
tmp->box = gen6_render_composite_spans_box; |
tmp->boxes = gen6_render_composite_spans_boxes; |
if (tmp->emit_boxes) |
tmp->thread_boxes = gen6_render_composite_spans_boxes__thread; |
tmp->done = gen6_render_composite_spans_done; |
kgem_set_mode(&sna->kgem, KGEM_RENDER, tmp->base.dst.bo); |
if (!kgem_check_bo(&sna->kgem, |
tmp->base.dst.bo, tmp->base.src.bo, |
NULL)) { |
kgem_submit(&sna->kgem); |
if (!kgem_check_bo(&sna->kgem, |
tmp->base.dst.bo, tmp->base.src.bo, |
NULL)) |
goto cleanup_src; |
_kgem_set_mode(&sna->kgem, KGEM_RENDER); |
} |
gen6_emit_composite_state(sna, &tmp->base); |
gen6_align_vertex(sna, &tmp->base); |
return true; |
cleanup_src: |
if (tmp->base.src.bo) |
kgem_bo_destroy(&sna->kgem, tmp->base.src.bo); |
cleanup_dst: |
if (tmp->base.redirect.real_bo) |
kgem_bo_destroy(&sna->kgem, tmp->base.dst.bo); |
return false; |
} |
#endif |
static void |
gen6_emit_copy_state(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
uint32_t *binding_table; |
uint16_t offset; |
bool dirty; |
dirty = gen6_get_batch(sna, op); |
binding_table = gen6_composite_get_binding_table(sna, &offset); |
binding_table[0] = |
gen6_bind_bo(sna, |
op->dst.bo, op->dst.width, op->dst.height, |
gen6_get_dest_format(op->dst.format), |
true); |
binding_table[1] = |
gen6_bind_bo(sna, |
op->src.bo, op->src.width, op->src.height, |
op->src.card_format, |
false); |
if (sna->kgem.surface == offset && |
*(uint64_t *)(sna->kgem.batch + sna->render_state.gen6.surface_table) == *(uint64_t*)binding_table) { |
sna->kgem.surface += sizeof(struct gen6_surface_state_padded) / sizeof(uint32_t); |
offset = sna->render_state.gen6.surface_table; |
} |
gen6_emit_state(sna, op, offset | dirty); |
} |
static inline bool prefer_blt_copy(struct sna *sna, |
struct kgem_bo *src_bo, |
struct kgem_bo *dst_bo, |
unsigned flags) |
{ |
if (flags & COPY_SYNC) |
return false; |
if (PREFER_RENDER) |
return PREFER_RENDER > 0; |
if (sna->kgem.ring == KGEM_BLT) |
return true; |
if (src_bo == dst_bo && can_switch_to_blt(sna, dst_bo, flags)) |
return true; |
if (untiled_tlb_miss(src_bo) || |
untiled_tlb_miss(dst_bo)) |
return true; |
if (kgem_bo_is_render(dst_bo) || |
kgem_bo_is_render(src_bo)) |
return false; |
if (!prefer_blt_ring(sna, dst_bo, flags)) |
return false; |
return prefer_blt_bo(sna, src_bo) || prefer_blt_bo(sna, dst_bo); |
} |
inline static void boxes_extents(const BoxRec *box, int n, BoxRec *extents) |
{ |
*extents = box[0]; |
while (--n) { |
box++; |
if (box->x1 < extents->x1) |
extents->x1 = box->x1; |
if (box->x2 > extents->x2) |
extents->x2 = box->x2; |
if (box->y1 < extents->y1) |
extents->y1 = box->y1; |
if (box->y2 > extents->y2) |
extents->y2 = box->y2; |
} |
} |
static inline bool |
overlaps(struct sna *sna, |
struct kgem_bo *src_bo, int16_t src_dx, int16_t src_dy, |
struct kgem_bo *dst_bo, int16_t dst_dx, int16_t dst_dy, |
const BoxRec *box, int n, BoxRec *extents) |
{ |
if (src_bo != dst_bo) |
return false; |
boxes_extents(box, n, extents); |
return (extents->x2 + src_dx > extents->x1 + dst_dx && |
extents->x1 + src_dx < extents->x2 + dst_dx && |
extents->y2 + src_dy > extents->y1 + dst_dy && |
extents->y1 + src_dy < extents->y2 + dst_dy); |
} |
static bool |
gen6_render_copy_boxes(struct sna *sna, uint8_t alu, |
PixmapPtr src, struct kgem_bo *src_bo, int16_t src_dx, int16_t src_dy, |
PixmapPtr dst, struct kgem_bo *dst_bo, int16_t dst_dx, int16_t dst_dy, |
const BoxRec *box, int n, unsigned flags) |
{ |
struct sna_composite_op tmp; |
BoxRec extents; |
DBG(("%s (%d, %d)->(%d, %d) x %d, alu=%x, self-copy=%d, overlaps? %d\n", |
__FUNCTION__, src_dx, src_dy, dst_dx, dst_dy, n, alu, |
src_bo == dst_bo, |
overlaps(sna, |
src_bo, src_dx, src_dy, |
dst_bo, dst_dx, dst_dy, |
box, n, &extents))); |
if (prefer_blt_copy(sna, src_bo, dst_bo, flags) && |
sna_blt_compare_depth(&src->drawable, &dst->drawable) && |
sna_blt_copy_boxes(sna, alu, |
src_bo, src_dx, src_dy, |
dst_bo, dst_dx, dst_dy, |
dst->drawable.bitsPerPixel, |
box, n)) |
return true; |
if (!(alu == GXcopy || alu == GXclear)) { |
fallback_blt: |
if (!sna_blt_compare_depth(&src->drawable, &dst->drawable)) |
return false; |
return sna_blt_copy_boxes_fallback(sna, alu, |
src, src_bo, src_dx, src_dy, |
dst, dst_bo, dst_dx, dst_dy, |
box, n); |
} |
if (overlaps(sna, |
src_bo, src_dx, src_dy, |
dst_bo, dst_dx, dst_dy, |
box, n, &extents)) { |
if (too_large(extents.x2-extents.x1, extents.y2-extents.y1)) |
goto fallback_blt; |
if (can_switch_to_blt(sna, dst_bo, flags) && |
sna_blt_compare_depth(&src->drawable, &dst->drawable) && |
sna_blt_copy_boxes(sna, alu, |
src_bo, src_dx, src_dy, |
dst_bo, dst_dx, dst_dy, |
dst->drawable.bitsPerPixel, |
box, n)) |
return true; |
return sna_render_copy_boxes__overlap(sna, alu, |
src, src_bo, src_dx, src_dy, |
dst, dst_bo, dst_dx, dst_dy, |
box, n, &extents); |
} |
if (dst->drawable.depth == src->drawable.depth) { |
tmp.dst.format = sna_render_format_for_depth(dst->drawable.depth); |
tmp.src.pict_format = tmp.dst.format; |
} else { |
tmp.dst.format = sna_format_for_depth(dst->drawable.depth); |
tmp.src.pict_format = sna_format_for_depth(src->drawable.depth); |
} |
if (!gen6_check_format(tmp.src.pict_format)) |
goto fallback_blt; |
tmp.dst.pixmap = dst; |
tmp.dst.width = dst->drawable.width; |
tmp.dst.height = dst->drawable.height; |
tmp.dst.bo = dst_bo; |
tmp.dst.x = tmp.dst.y = 0; |
tmp.damage = NULL; |
sna_render_composite_redirect_init(&tmp); |
if (too_large(tmp.dst.width, tmp.dst.height)) { |
int i; |
extents = box[0]; |
for (i = 1; i < n; i++) { |
if (box[i].x1 < extents.x1) |
extents.x1 = box[i].x1; |
if (box[i].y1 < extents.y1) |
extents.y1 = box[i].y1; |
if (box[i].x2 > extents.x2) |
extents.x2 = box[i].x2; |
if (box[i].y2 > extents.y2) |
extents.y2 = box[i].y2; |
} |
if (!sna_render_composite_redirect(sna, &tmp, |
extents.x1 + dst_dx, |
extents.y1 + dst_dy, |
extents.x2 - extents.x1, |
extents.y2 - extents.y1, |
n > 1)) |
goto fallback_tiled; |
dst_dx += tmp.dst.x; |
dst_dy += tmp.dst.y; |
tmp.dst.x = tmp.dst.y = 0; |
} |
tmp.src.card_format = gen6_get_card_format(tmp.src.pict_format); |
if (too_large(src->drawable.width, src->drawable.height)) { |
int i; |
extents = box[0]; |
for (i = 1; i < n; i++) { |
if (box[i].x1 < extents.x1) |
extents.x1 = box[i].x1; |
if (box[i].y1 < extents.y1) |
extents.y1 = box[i].y1; |
if (box[i].x2 > extents.x2) |
extents.x2 = box[i].x2; |
if (box[i].y2 > extents.y2) |
extents.y2 = box[i].y2; |
} |
if (!sna_render_pixmap_partial(sna, src, src_bo, &tmp.src, |
extents.x1 + src_dx, |
extents.y1 + src_dy, |
extents.x2 - extents.x1, |
extents.y2 - extents.y1)) { |
DBG(("%s: unable to extract partial pixmap\n", __FUNCTION__)); |
goto fallback_tiled_dst; |
} |
src_dx += tmp.src.offset[0]; |
src_dy += tmp.src.offset[1]; |
} else { |
tmp.src.bo = src_bo; |
tmp.src.width = src->drawable.width; |
tmp.src.height = src->drawable.height; |
} |
tmp.mask.bo = NULL; |
tmp.floats_per_vertex = 2; |
tmp.floats_per_rect = 6; |
tmp.need_magic_ca_pass = 0; |
tmp.u.gen6.flags = COPY_FLAGS(alu); |
assert(GEN6_KERNEL(tmp.u.gen6.flags) == GEN6_WM_KERNEL_NOMASK); |
assert(GEN6_SAMPLER(tmp.u.gen6.flags) == COPY_SAMPLER); |
assert(GEN6_VERTEX(tmp.u.gen6.flags) == COPY_VERTEX); |
kgem_set_mode(&sna->kgem, KGEM_RENDER, tmp.dst.bo); |
if (!kgem_check_bo(&sna->kgem, tmp.dst.bo, tmp.src.bo, NULL)) { |
kgem_submit(&sna->kgem); |
if (!kgem_check_bo(&sna->kgem, tmp.dst.bo, tmp.src.bo, NULL)) { |
DBG(("%s: too large for a single operation\n", |
__FUNCTION__)); |
goto fallback_tiled_src; |
} |
_kgem_set_mode(&sna->kgem, KGEM_RENDER); |
} |
gen6_emit_copy_state(sna, &tmp); |
gen6_align_vertex(sna, &tmp); |
do { |
int16_t *v; |
int n_this_time; |
n_this_time = gen6_get_rectangles(sna, &tmp, n, |
gen6_emit_copy_state); |
n -= n_this_time; |
v = (int16_t *)(sna->render.vertices + sna->render.vertex_used); |
sna->render.vertex_used += 6 * n_this_time; |
assert(sna->render.vertex_used <= sna->render.vertex_size); |
do { |
DBG((" (%d, %d) -> (%d, %d) + (%d, %d)\n", |
box->x1 + src_dx, box->y1 + src_dy, |
box->x1 + dst_dx, box->y1 + dst_dy, |
box->x2 - box->x1, box->y2 - box->y1)); |
v[0] = box->x2 + dst_dx; |
v[2] = box->x2 + src_dx; |
v[1] = v[5] = box->y2 + dst_dy; |
v[3] = v[7] = box->y2 + src_dy; |
v[8] = v[4] = box->x1 + dst_dx; |
v[10] = v[6] = box->x1 + src_dx; |
v[9] = box->y1 + dst_dy; |
v[11] = box->y1 + src_dy; |
v += 12; box++; |
} while (--n_this_time); |
} while (n); |
gen4_vertex_flush(sna); |
sna_render_composite_redirect_done(sna, &tmp); |
if (tmp.src.bo != src_bo) |
kgem_bo_destroy(&sna->kgem, tmp.src.bo); |
return true; |
fallback_tiled_src: |
if (tmp.src.bo != src_bo) |
kgem_bo_destroy(&sna->kgem, tmp.src.bo); |
fallback_tiled_dst: |
if (tmp.redirect.real_bo) |
kgem_bo_destroy(&sna->kgem, tmp.dst.bo); |
fallback_tiled: |
if (sna_blt_compare_depth(&src->drawable, &dst->drawable) && |
sna_blt_copy_boxes(sna, alu, |
src_bo, src_dx, src_dy, |
dst_bo, dst_dx, dst_dy, |
dst->drawable.bitsPerPixel, |
box, n)) |
return true; |
return sna_tiling_copy_boxes(sna, alu, |
src, src_bo, src_dx, src_dy, |
dst, dst_bo, dst_dx, dst_dy, |
box, n); |
} |
static void |
gen6_render_copy_blt(struct sna *sna, |
const struct sna_copy_op *op, |
int16_t sx, int16_t sy, |
int16_t w, int16_t h, |
int16_t dx, int16_t dy) |
{ |
int16_t *v; |
gen6_get_rectangles(sna, &op->base, 1, gen6_emit_copy_state); |
v = (int16_t *)&sna->render.vertices[sna->render.vertex_used]; |
sna->render.vertex_used += 6; |
assert(sna->render.vertex_used <= sna->render.vertex_size); |
v[0] = dx+w; v[1] = dy+h; |
v[2] = sx+w; v[3] = sy+h; |
v[4] = dx; v[5] = dy+h; |
v[6] = sx; v[7] = sy+h; |
v[8] = dx; v[9] = dy; |
v[10] = sx; v[11] = sy; |
} |
static void |
gen6_render_copy_done(struct sna *sna, const struct sna_copy_op *op) |
{ |
DBG(("%s()\n", __FUNCTION__)); |
assert(!sna->render.active); |
if (sna->render.vertex_offset) |
gen4_vertex_flush(sna); |
} |
static bool |
gen6_render_copy(struct sna *sna, uint8_t alu, |
PixmapPtr src, struct kgem_bo *src_bo, |
PixmapPtr dst, struct kgem_bo *dst_bo, |
struct sna_copy_op *op) |
{ |
DBG(("%s (alu=%d, src=(%dx%d), dst=(%dx%d))\n", |
__FUNCTION__, alu, |
src->drawable.width, src->drawable.height, |
dst->drawable.width, dst->drawable.height)); |
if (prefer_blt_copy(sna, src_bo, dst_bo, 0) && |
sna_blt_compare_depth(&src->drawable, &dst->drawable) && |
sna_blt_copy(sna, alu, |
src_bo, dst_bo, |
dst->drawable.bitsPerPixel, |
op)) |
return true; |
if (!(alu == GXcopy || alu == GXclear) || src_bo == dst_bo || |
too_large(src->drawable.width, src->drawable.height) || |
too_large(dst->drawable.width, dst->drawable.height)) { |
fallback: |
if (!sna_blt_compare_depth(&src->drawable, &dst->drawable)) |
return false; |
return sna_blt_copy(sna, alu, src_bo, dst_bo, |
dst->drawable.bitsPerPixel, |
op); |
} |
if (dst->drawable.depth == src->drawable.depth) { |
op->base.dst.format = sna_render_format_for_depth(dst->drawable.depth); |
op->base.src.pict_format = op->base.dst.format; |
} else { |
op->base.dst.format = sna_format_for_depth(dst->drawable.depth); |
op->base.src.pict_format = sna_format_for_depth(src->drawable.depth); |
} |
if (!gen6_check_format(op->base.src.pict_format)) |
goto fallback; |
op->base.dst.pixmap = dst; |
op->base.dst.width = dst->drawable.width; |
op->base.dst.height = dst->drawable.height; |
op->base.dst.bo = dst_bo; |
op->base.src.bo = src_bo; |
op->base.src.card_format = |
gen6_get_card_format(op->base.src.pict_format); |
op->base.src.width = src->drawable.width; |
op->base.src.height = src->drawable.height; |
op->base.mask.bo = NULL; |
op->base.floats_per_vertex = 2; |
op->base.floats_per_rect = 6; |
op->base.u.gen6.flags = COPY_FLAGS(alu); |
assert(GEN6_KERNEL(op->base.u.gen6.flags) == GEN6_WM_KERNEL_NOMASK); |
assert(GEN6_SAMPLER(op->base.u.gen6.flags) == COPY_SAMPLER); |
assert(GEN6_VERTEX(op->base.u.gen6.flags) == COPY_VERTEX); |
kgem_set_mode(&sna->kgem, KGEM_RENDER, dst_bo); |
if (!kgem_check_bo(&sna->kgem, dst_bo, src_bo, NULL)) { |
kgem_submit(&sna->kgem); |
if (!kgem_check_bo(&sna->kgem, dst_bo, src_bo, NULL)) |
goto fallback; |
_kgem_set_mode(&sna->kgem, KGEM_RENDER); |
} |
gen6_emit_copy_state(sna, &op->base); |
gen6_align_vertex(sna, &op->base); |
op->blt = gen6_render_copy_blt; |
op->done = gen6_render_copy_done; |
return true; |
} |
static void |
gen6_emit_fill_state(struct sna *sna, const struct sna_composite_op *op) |
{ |
uint32_t *binding_table; |
uint16_t offset; |
bool dirty; |
dirty = gen6_get_batch(sna, op); |
binding_table = gen6_composite_get_binding_table(sna, &offset); |
binding_table[0] = |
gen6_bind_bo(sna, |
op->dst.bo, op->dst.width, op->dst.height, |
gen6_get_dest_format(op->dst.format), |
true); |
binding_table[1] = |
gen6_bind_bo(sna, |
op->src.bo, 1, 1, |
GEN6_SURFACEFORMAT_B8G8R8A8_UNORM, |
false); |
if (sna->kgem.surface == offset && |
*(uint64_t *)(sna->kgem.batch + sna->render_state.gen6.surface_table) == *(uint64_t*)binding_table) { |
sna->kgem.surface += |
sizeof(struct gen6_surface_state_padded)/sizeof(uint32_t); |
offset = sna->render_state.gen6.surface_table; |
} |
gen6_emit_state(sna, op, offset | dirty); |
} |
static inline bool prefer_blt_fill(struct sna *sna, |
struct kgem_bo *bo) |
{ |
if (PREFER_RENDER) |
return PREFER_RENDER < 0; |
if (kgem_bo_is_render(bo)) |
return false; |
if (untiled_tlb_miss(bo)) |
return true; |
if (!prefer_blt_ring(sna, bo, 0)) |
return false; |
return prefer_blt_bo(sna, bo); |
} |
static bool |
gen6_render_fill_boxes(struct sna *sna, |
CARD8 op, |
PictFormat format, |
const xRenderColor *color, |
PixmapPtr dst, struct kgem_bo *dst_bo, |
const BoxRec *box, int n) |
{ |
struct sna_composite_op tmp; |
uint32_t pixel; |
DBG(("%s (op=%d, color=(%04x, %04x, %04x, %04x) [%08x])\n", |
__FUNCTION__, op, |
color->red, color->green, color->blue, color->alpha, (int)format)); |
if (op >= ARRAY_SIZE(gen6_blend_op)) { |
DBG(("%s: fallback due to unhandled blend op: %d\n", |
__FUNCTION__, op)); |
return false; |
} |
if (prefer_blt_fill(sna, dst_bo) || !gen6_check_dst_format(format)) { |
uint8_t alu = GXinvalid; |
if (op <= PictOpSrc) { |
pixel = 0; |
if (op == PictOpClear) |
alu = GXclear; |
else if (sna_get_pixel_from_rgba(&pixel, |
color->red, |
color->green, |
color->blue, |
color->alpha, |
format)) |
alu = GXcopy; |
} |
if (alu != GXinvalid && |
sna_blt_fill_boxes(sna, alu, |
dst_bo, dst->drawable.bitsPerPixel, |
pixel, box, n)) |
return true; |
if (!gen6_check_dst_format(format)) |
return false; |
} |
if (op == PictOpClear) { |
pixel = 0; |
op = PictOpSrc; |
} else if (!sna_get_pixel_from_rgba(&pixel, |
color->red, |
color->green, |
color->blue, |
color->alpha, |
PICT_a8r8g8b8)) |
return false; |
DBG(("%s(%08x x %d [(%d, %d), (%d, %d) ...])\n", |
__FUNCTION__, pixel, n, |
box[0].x1, box[0].y1, box[0].x2, box[0].y2)); |
tmp.dst.pixmap = dst; |
tmp.dst.width = dst->drawable.width; |
tmp.dst.height = dst->drawable.height; |
tmp.dst.format = format; |
tmp.dst.bo = dst_bo; |
tmp.dst.x = tmp.dst.y = 0; |
tmp.damage = NULL; |
sna_render_composite_redirect_init(&tmp); |
if (too_large(dst->drawable.width, dst->drawable.height)) { |
BoxRec extents; |
boxes_extents(box, n, &extents); |
if (!sna_render_composite_redirect(sna, &tmp, |
extents.x1, extents.y1, |
extents.x2 - extents.x1, |
extents.y2 - extents.y1, |
n > 1)) |
return sna_tiling_fill_boxes(sna, op, format, color, |
dst, dst_bo, box, n); |
} |
tmp.src.bo = sna_render_get_solid(sna, pixel); |
tmp.mask.bo = NULL; |
tmp.floats_per_vertex = 2; |
tmp.floats_per_rect = 6; |
tmp.need_magic_ca_pass = false; |
tmp.u.gen6.flags = FILL_FLAGS(op, format); |
assert(GEN6_KERNEL(tmp.u.gen6.flags) == GEN6_WM_KERNEL_NOMASK); |
assert(GEN6_SAMPLER(tmp.u.gen6.flags) == FILL_SAMPLER); |
assert(GEN6_VERTEX(tmp.u.gen6.flags) == FILL_VERTEX); |
if (!kgem_check_bo(&sna->kgem, dst_bo, NULL)) { |
kgem_submit(&sna->kgem); |
assert(kgem_check_bo(&sna->kgem, dst_bo, NULL)); |
} |
gen6_emit_fill_state(sna, &tmp); |
gen6_align_vertex(sna, &tmp); |
do { |
int n_this_time; |
int16_t *v; |
n_this_time = gen6_get_rectangles(sna, &tmp, n, |
gen6_emit_fill_state); |
n -= n_this_time; |
v = (int16_t *)(sna->render.vertices + sna->render.vertex_used); |
sna->render.vertex_used += 6 * n_this_time; |
assert(sna->render.vertex_used <= sna->render.vertex_size); |
do { |
DBG((" (%d, %d), (%d, %d)\n", |
box->x1, box->y1, box->x2, box->y2)); |
v[0] = box->x2; |
v[5] = v[1] = box->y2; |
v[8] = v[4] = box->x1; |
v[9] = box->y1; |
v[2] = v[3] = v[7] = 1; |
v[6] = v[10] = v[11] = 0; |
v += 12; box++; |
} while (--n_this_time); |
} while (n); |
gen4_vertex_flush(sna); |
kgem_bo_destroy(&sna->kgem, tmp.src.bo); |
sna_render_composite_redirect_done(sna, &tmp); |
return true; |
} |
static void |
gen6_render_op_fill_blt(struct sna *sna, |
const struct sna_fill_op *op, |
int16_t x, int16_t y, int16_t w, int16_t h) |
{ |
int16_t *v; |
DBG(("%s: (%d, %d)x(%d, %d)\n", __FUNCTION__, x, y, w, h)); |
gen6_get_rectangles(sna, &op->base, 1, gen6_emit_fill_state); |
v = (int16_t *)&sna->render.vertices[sna->render.vertex_used]; |
sna->render.vertex_used += 6; |
assert(sna->render.vertex_used <= sna->render.vertex_size); |
v[0] = x+w; |
v[4] = v[8] = x; |
v[1] = v[5] = y+h; |
v[9] = y; |
v[2] = v[3] = v[7] = 1; |
v[6] = v[10] = v[11] = 0; |
} |
fastcall static void |
gen6_render_op_fill_box(struct sna *sna, |
const struct sna_fill_op *op, |
const BoxRec *box) |
{ |
int16_t *v; |
DBG(("%s: (%d, %d),(%d, %d)\n", __FUNCTION__, |
box->x1, box->y1, box->x2, box->y2)); |
gen6_get_rectangles(sna, &op->base, 1, gen6_emit_fill_state); |
v = (int16_t *)&sna->render.vertices[sna->render.vertex_used]; |
sna->render.vertex_used += 6; |
assert(sna->render.vertex_used <= sna->render.vertex_size); |
v[0] = box->x2; |
v[8] = v[4] = box->x1; |
v[5] = v[1] = box->y2; |
v[9] = box->y1; |
v[7] = v[2] = v[3] = 1; |
v[6] = v[10] = v[11] = 0; |
} |
fastcall static void |
gen6_render_op_fill_boxes(struct sna *sna, |
const struct sna_fill_op *op, |
const BoxRec *box, |
int nbox) |
{ |
DBG(("%s: (%d, %d),(%d, %d)... x %d\n", __FUNCTION__, |
box->x1, box->y1, box->x2, box->y2, nbox)); |
do { |
int nbox_this_time; |
int16_t *v; |
nbox_this_time = gen6_get_rectangles(sna, &op->base, nbox, |
gen6_emit_fill_state); |
nbox -= nbox_this_time; |
v = (int16_t *)&sna->render.vertices[sna->render.vertex_used]; |
sna->render.vertex_used += 6 * nbox_this_time; |
assert(sna->render.vertex_used <= sna->render.vertex_size); |
do { |
v[0] = box->x2; |
v[8] = v[4] = box->x1; |
v[5] = v[1] = box->y2; |
v[9] = box->y1; |
v[7] = v[2] = v[3] = 1; |
v[6] = v[10] = v[11] = 0; |
box++; v += 12; |
} while (--nbox_this_time); |
} while (nbox); |
} |
static void |
gen6_render_op_fill_done(struct sna *sna, const struct sna_fill_op *op) |
{ |
DBG(("%s()\n", __FUNCTION__)); |
assert(!sna->render.active); |
if (sna->render.vertex_offset) |
gen4_vertex_flush(sna); |
kgem_bo_destroy(&sna->kgem, op->base.src.bo); |
} |
static bool |
gen6_render_fill(struct sna *sna, uint8_t alu, |
PixmapPtr dst, struct kgem_bo *dst_bo, |
uint32_t color, |
struct sna_fill_op *op) |
{ |
DBG(("%s: (alu=%d, color=%x)\n", __FUNCTION__, alu, color)); |
if (prefer_blt_fill(sna, dst_bo) && |
sna_blt_fill(sna, alu, |
dst_bo, dst->drawable.bitsPerPixel, |
color, |
op)) |
return true; |
if (!(alu == GXcopy || alu == GXclear) || |
too_large(dst->drawable.width, dst->drawable.height)) |
return sna_blt_fill(sna, alu, |
dst_bo, dst->drawable.bitsPerPixel, |
color, |
op); |
if (alu == GXclear) |
color = 0; |
op->base.dst.pixmap = dst; |
op->base.dst.width = dst->drawable.width; |
op->base.dst.height = dst->drawable.height; |
op->base.dst.format = sna_format_for_depth(dst->drawable.depth); |
op->base.dst.bo = dst_bo; |
op->base.dst.x = op->base.dst.y = 0; |
op->base.src.bo = |
sna_render_get_solid(sna, |
sna_rgba_for_color(color, |
dst->drawable.depth)); |
op->base.mask.bo = NULL; |
op->base.need_magic_ca_pass = false; |
op->base.floats_per_vertex = 2; |
op->base.floats_per_rect = 6; |
op->base.u.gen6.flags = FILL_FLAGS_NOBLEND; |
assert(GEN6_KERNEL(op->base.u.gen6.flags) == GEN6_WM_KERNEL_NOMASK); |
assert(GEN6_SAMPLER(op->base.u.gen6.flags) == FILL_SAMPLER); |
assert(GEN6_VERTEX(op->base.u.gen6.flags) == FILL_VERTEX); |
if (!kgem_check_bo(&sna->kgem, dst_bo, NULL)) { |
kgem_submit(&sna->kgem); |
assert(kgem_check_bo(&sna->kgem, dst_bo, NULL)); |
} |
gen6_emit_fill_state(sna, &op->base); |
gen6_align_vertex(sna, &op->base); |
op->blt = gen6_render_op_fill_blt; |
op->box = gen6_render_op_fill_box; |
op->boxes = gen6_render_op_fill_boxes; |
op->done = gen6_render_op_fill_done; |
return true; |
} |
static bool |
gen6_render_fill_one_try_blt(struct sna *sna, PixmapPtr dst, struct kgem_bo *bo, |
uint32_t color, |
int16_t x1, int16_t y1, int16_t x2, int16_t y2, |
uint8_t alu) |
{ |
BoxRec box; |
box.x1 = x1; |
box.y1 = y1; |
box.x2 = x2; |
box.y2 = y2; |
return sna_blt_fill_boxes(sna, alu, |
bo, dst->drawable.bitsPerPixel, |
color, &box, 1); |
} |
static bool |
gen6_render_fill_one(struct sna *sna, PixmapPtr dst, struct kgem_bo *bo, |
uint32_t color, |
int16_t x1, int16_t y1, |
int16_t x2, int16_t y2, |
uint8_t alu) |
{ |
struct sna_composite_op tmp; |
int16_t *v; |
/* Prefer to use the BLT if already engaged */ |
if (prefer_blt_fill(sna, bo) && |
gen6_render_fill_one_try_blt(sna, dst, bo, color, |
x1, y1, x2, y2, alu)) |
return true; |
/* Must use the BLT if we can't RENDER... */ |
if (!(alu == GXcopy || alu == GXclear) || |
too_large(dst->drawable.width, dst->drawable.height)) |
return gen6_render_fill_one_try_blt(sna, dst, bo, color, |
x1, y1, x2, y2, alu); |
if (alu == GXclear) |
color = 0; |
tmp.dst.pixmap = dst; |
tmp.dst.width = dst->drawable.width; |
tmp.dst.height = dst->drawable.height; |
tmp.dst.format = sna_format_for_depth(dst->drawable.depth); |
tmp.dst.bo = bo; |
tmp.dst.x = tmp.dst.y = 0; |
tmp.src.bo = |
sna_render_get_solid(sna, |
sna_rgba_for_color(color, |
dst->drawable.depth)); |
tmp.mask.bo = NULL; |
tmp.floats_per_vertex = 2; |
tmp.floats_per_rect = 6; |
tmp.need_magic_ca_pass = false; |
tmp.u.gen6.flags = FILL_FLAGS_NOBLEND; |
assert(GEN6_KERNEL(tmp.u.gen6.flags) == GEN6_WM_KERNEL_NOMASK); |
assert(GEN6_SAMPLER(tmp.u.gen6.flags) == FILL_SAMPLER); |
assert(GEN6_VERTEX(tmp.u.gen6.flags) == FILL_VERTEX); |
if (!kgem_check_bo(&sna->kgem, bo, NULL)) { |
kgem_submit(&sna->kgem); |
if (!kgem_check_bo(&sna->kgem, bo, NULL)) { |
kgem_bo_destroy(&sna->kgem, tmp.src.bo); |
return false; |
} |
} |
gen6_emit_fill_state(sna, &tmp); |
gen6_align_vertex(sna, &tmp); |
gen6_get_rectangles(sna, &tmp, 1, gen6_emit_fill_state); |
DBG((" (%d, %d), (%d, %d)\n", x1, y1, x2, y2)); |
v = (int16_t *)&sna->render.vertices[sna->render.vertex_used]; |
sna->render.vertex_used += 6; |
assert(sna->render.vertex_used <= sna->render.vertex_size); |
v[0] = x2; |
v[8] = v[4] = x1; |
v[5] = v[1] = y2; |
v[9] = y1; |
v[7] = v[2] = v[3] = 1; |
v[6] = v[10] = v[11] = 0; |
gen4_vertex_flush(sna); |
kgem_bo_destroy(&sna->kgem, tmp.src.bo); |
return true; |
} |
static bool |
gen6_render_clear_try_blt(struct sna *sna, PixmapPtr dst, struct kgem_bo *bo) |
{ |
BoxRec box; |
box.x1 = 0; |
box.y1 = 0; |
box.x2 = dst->drawable.width; |
box.y2 = dst->drawable.height; |
return sna_blt_fill_boxes(sna, GXclear, |
bo, dst->drawable.bitsPerPixel, |
0, &box, 1); |
} |
static bool |
gen6_render_clear(struct sna *sna, PixmapPtr dst, struct kgem_bo *bo) |
{ |
struct sna_composite_op tmp; |
int16_t *v; |
DBG(("%s: %dx%d\n", |
__FUNCTION__, |
dst->drawable.width, |
dst->drawable.height)); |
/* Prefer to use the BLT if, and only if, already engaged */ |
if (sna->kgem.ring == KGEM_BLT && |
gen6_render_clear_try_blt(sna, dst, bo)) |
return true; |
/* Must use the BLT if we can't RENDER... */ |
if (too_large(dst->drawable.width, dst->drawable.height)) |
return gen6_render_clear_try_blt(sna, dst, bo); |
tmp.dst.pixmap = dst; |
tmp.dst.width = dst->drawable.width; |
tmp.dst.height = dst->drawable.height; |
tmp.dst.format = sna_format_for_depth(dst->drawable.depth); |
tmp.dst.bo = bo; |
tmp.dst.x = tmp.dst.y = 0; |
tmp.src.bo = sna_render_get_solid(sna, 0); |
tmp.mask.bo = NULL; |
tmp.floats_per_vertex = 2; |
tmp.floats_per_rect = 6; |
tmp.need_magic_ca_pass = false; |
tmp.u.gen6.flags = FILL_FLAGS_NOBLEND; |
assert(GEN6_KERNEL(tmp.u.gen6.flags) == GEN6_WM_KERNEL_NOMASK); |
assert(GEN6_SAMPLER(tmp.u.gen6.flags) == FILL_SAMPLER); |
assert(GEN6_VERTEX(tmp.u.gen6.flags) == FILL_VERTEX); |
if (!kgem_check_bo(&sna->kgem, bo, NULL)) { |
kgem_submit(&sna->kgem); |
if (!kgem_check_bo(&sna->kgem, bo, NULL)) { |
kgem_bo_destroy(&sna->kgem, tmp.src.bo); |
return false; |
} |
} |
gen6_emit_fill_state(sna, &tmp); |
gen6_align_vertex(sna, &tmp); |
gen6_get_rectangles(sna, &tmp, 1, gen6_emit_fill_state); |
v = (int16_t *)&sna->render.vertices[sna->render.vertex_used]; |
sna->render.vertex_used += 6; |
assert(sna->render.vertex_used <= sna->render.vertex_size); |
v[0] = dst->drawable.width; |
v[5] = v[1] = dst->drawable.height; |
v[8] = v[4] = 0; |
v[9] = 0; |
v[7] = v[2] = v[3] = 1; |
v[6] = v[10] = v[11] = 0; |
gen4_vertex_flush(sna); |
kgem_bo_destroy(&sna->kgem, tmp.src.bo); |
return true; |
} |
#endif |
static void gen6_render_flush(struct sna *sna) |
{ |
gen4_vertex_close(sna); |
assert(sna->render.vb_id == 0); |
assert(sna->render.vertex_offset == 0); |
} |
static void |
gen6_render_context_switch(struct kgem *kgem, |
int new_mode) |
{ |
if (kgem->nbatch) { |
DBG(("%s: from %d to %d\n", __FUNCTION__, kgem->mode, new_mode)); |
_kgem_submit(kgem); |
} |
kgem->ring = new_mode; |
} |
static void |
gen6_render_retire(struct kgem *kgem) |
{ |
struct sna *sna; |
if (kgem->ring && (kgem->has_semaphores || !kgem->need_retire)) |
kgem->ring = kgem->mode; |
sna = container_of(kgem, struct sna, kgem); |
if (kgem->nbatch == 0 && sna->render.vbo && !kgem_bo_is_busy(sna->render.vbo)) { |
DBG(("%s: resetting idle vbo handle=%d\n", __FUNCTION__, sna->render.vbo->handle)); |
sna->render.vertex_used = 0; |
sna->render.vertex_index = 0; |
} |
} |
static void |
gen6_render_expire(struct kgem *kgem) |
{ |
struct sna *sna; |
sna = container_of(kgem, struct sna, kgem); |
if (sna->render.vbo && !sna->render.vertex_used) { |
DBG(("%s: discarding vbo handle=%d\n", __FUNCTION__, sna->render.vbo->handle)); |
kgem_bo_destroy(kgem, sna->render.vbo); |
assert(!sna->render.active); |
sna->render.vbo = NULL; |
sna->render.vertices = sna->render.vertex_data; |
sna->render.vertex_size = ARRAY_SIZE(sna->render.vertex_data); |
sna->render.vertex_used = 0; |
sna->render.vertex_index = 0; |
} |
} |
static void gen6_render_reset(struct sna *sna) |
{ |
sna->render_state.gen6.needs_invariant = true; |
sna->render_state.gen6.first_state_packet = true; |
sna->render_state.gen6.ve_id = 3 << 2; |
sna->render_state.gen6.last_primitive = -1; |
sna->render_state.gen6.num_sf_outputs = 0; |
sna->render_state.gen6.samplers = -1; |
sna->render_state.gen6.blend = -1; |
sna->render_state.gen6.kernel = -1; |
sna->render_state.gen6.drawrect_offset = -1; |
sna->render_state.gen6.drawrect_limit = -1; |
sna->render_state.gen6.surface_table = -1; |
sna->render.vertex_offset = 0; |
sna->render.nvertex_reloc = 0; |
sna->render.vb_id = 0; |
} |
static void gen6_render_fini(struct sna *sna) |
{ |
kgem_bo_destroy(&sna->kgem, sna->render_state.gen6.general_bo); |
} |
static bool is_gt2(struct sna *sna) |
{ |
return sna->PciInfo->device_id & 0x30; |
} |
static bool is_mobile(struct sna *sna) |
{ |
return (sna->PciInfo->device_id & 0xf) == 0x6; |
} |
static bool gen6_render_setup(struct sna *sna) |
{ |
struct gen6_render_state *state = &sna->render_state.gen6; |
struct sna_static_stream general; |
struct gen6_sampler_state *ss; |
int i, j, k, l, m; |
state->info = >1_info; |
if (is_gt2(sna)) |
state->info = >2_info; /* XXX requires GT_MODE WiZ disabled */ |
sna_static_stream_init(&general); |
/* Zero pad the start. If you see an offset of 0x0 in the batchbuffer |
* dumps, you know it points to zero. |
*/ |
null_create(&general); |
scratch_create(&general); |
for (m = 0; m < GEN6_KERNEL_COUNT; m++) { |
if (wm_kernels[m].size) { |
state->wm_kernel[m][1] = |
sna_static_stream_add(&general, |
wm_kernels[m].data, |
wm_kernels[m].size, |
64); |
} else { |
if (USE_8_PIXEL_DISPATCH) { |
state->wm_kernel[m][0] = |
sna_static_stream_compile_wm(sna, &general, |
wm_kernels[m].data, 8); |
} |
if (USE_16_PIXEL_DISPATCH) { |
state->wm_kernel[m][1] = |
sna_static_stream_compile_wm(sna, &general, |
wm_kernels[m].data, 16); |
} |
if (USE_32_PIXEL_DISPATCH) { |
state->wm_kernel[m][2] = |
sna_static_stream_compile_wm(sna, &general, |
wm_kernels[m].data, 32); |
} |
} |
if ((state->wm_kernel[m][0]|state->wm_kernel[m][1]|state->wm_kernel[m][2]) == 0) { |
state->wm_kernel[m][1] = |
sna_static_stream_compile_wm(sna, &general, |
wm_kernels[m].data, 16); |
} |
} |
ss = sna_static_stream_map(&general, |
2 * sizeof(*ss) * |
(2 + |
FILTER_COUNT * EXTEND_COUNT * |
FILTER_COUNT * EXTEND_COUNT), |
32); |
state->wm_state = sna_static_stream_offsetof(&general, ss); |
sampler_copy_init(ss); ss += 2; |
sampler_fill_init(ss); ss += 2; |
for (i = 0; i < FILTER_COUNT; i++) { |
for (j = 0; j < EXTEND_COUNT; j++) { |
for (k = 0; k < FILTER_COUNT; k++) { |
for (l = 0; l < EXTEND_COUNT; l++) { |
sampler_state_init(ss++, i, j); |
sampler_state_init(ss++, k, l); |
} |
} |
} |
} |
state->cc_blend = gen6_composite_create_blend_state(&general); |
state->general_bo = sna_static_stream_fini(sna, &general); |
return state->general_bo != NULL; |
} |
const char *gen6_render_init(struct sna *sna, const char *backend) |
{ |
if (!gen6_render_setup(sna)) |
return backend; |
sna->kgem.context_switch = gen6_render_context_switch; |
sna->kgem.retire = gen6_render_retire; |
sna->kgem.expire = gen6_render_expire; |
#if 0 |
#if !NO_COMPOSITE |
sna->render.composite = gen6_render_composite; |
sna->render.prefer_gpu |= PREFER_GPU_RENDER; |
#endif |
#if !NO_COMPOSITE_SPANS |
sna->render.check_composite_spans = gen6_check_composite_spans; |
sna->render.composite_spans = gen6_render_composite_spans; |
if (is_mobile(sna)) |
sna->render.prefer_gpu |= PREFER_GPU_SPANS; |
#endif |
sna->render.video = gen6_render_video; |
#if !NO_COPY_BOXES |
sna->render.copy_boxes = gen6_render_copy_boxes; |
#endif |
#if !NO_COPY |
sna->render.copy = gen6_render_copy; |
#endif |
#if !NO_FILL_BOXES |
sna->render.fill_boxes = gen6_render_fill_boxes; |
#endif |
#if !NO_FILL |
sna->render.fill = gen6_render_fill; |
#endif |
#if !NO_FILL_ONE |
sna->render.fill_one = gen6_render_fill_one; |
#endif |
#if !NO_FILL_CLEAR |
sna->render.clear = gen6_render_clear; |
#endif |
#endif |
sna->render.caps = HW_BIT_BLIT | HW_TEX_BLIT; |
sna->render.blit_tex = gen6_blit_tex; |
sna->render.flush = gen6_render_flush; |
sna->render.reset = gen6_render_reset; |
sna->render.fini = gen6_render_fini; |
sna->render.max_3d_size = GEN6_MAX_SIZE; |
sna->render.max_3d_pitch = 1 << 18; |
return sna->render_state.gen6.info->name; |
} |
static bool |
gen6_blit_tex(struct sna *sna, |
uint8_t op, bool scale, |
PixmapPtr src, struct kgem_bo *src_bo, |
PixmapPtr mask,struct kgem_bo *mask_bo, |
PixmapPtr dst, struct kgem_bo *dst_bo, |
int32_t src_x, int32_t src_y, |
int32_t msk_x, int32_t msk_y, |
int32_t dst_x, int32_t dst_y, |
int32_t width, int32_t height, |
struct sna_composite_op *tmp) |
{ |
DBG(("%s: %dx%d, current mode=%d\n", __FUNCTION__, |
width, height, sna->kgem.ring)); |
tmp->op = PictOpSrc; |
tmp->dst.pixmap = dst; |
tmp->dst.bo = dst_bo; |
tmp->dst.width = dst->drawable.width; |
tmp->dst.height = dst->drawable.height; |
tmp->dst.format = PICT_a8r8g8b8; |
tmp->src.repeat = SAMPLER_EXTEND_NONE; |
tmp->src.is_affine = true; |
tmp->src.bo = src_bo; |
tmp->src.pict_format = PICT_x8r8g8b8; |
tmp->src.card_format = gen6_get_card_format(tmp->src.pict_format); |
tmp->src.width = src->drawable.width; |
tmp->src.height = src->drawable.height; |
if ( (tmp->src.width == width) && |
(tmp->src.height == height) ) |
tmp->src.filter = SAMPLER_FILTER_NEAREST; |
else |
tmp->src.filter = SAMPLER_FILTER_BILINEAR; |
tmp->is_affine = tmp->src.is_affine; |
tmp->has_component_alpha = false; |
tmp->need_magic_ca_pass = false; |
tmp->mask.repeat = SAMPLER_EXTEND_NONE; |
tmp->mask.filter = SAMPLER_FILTER_NEAREST; |
tmp->mask.is_affine = true; |
tmp->mask.bo = mask_bo; |
tmp->mask.pict_format = PIXMAN_a8; |
tmp->mask.card_format = gen6_get_card_format(tmp->mask.pict_format); |
tmp->mask.width = mask->drawable.width; |
tmp->mask.height = mask->drawable.height; |
if( scale ) |
{ |
tmp->src.scale[0] = 1.f/width; |
tmp->src.scale[1] = 1.f/height; |
} |
else |
{ |
tmp->src.scale[0] = 1.f/src->drawable.width; |
tmp->src.scale[1] = 1.f/src->drawable.height; |
} |
// tmp->src.offset[0] = -dst_x; |
// tmp->src.offset[1] = -dst_y; |
tmp->mask.scale[0] = 1.f/mask->drawable.width; |
tmp->mask.scale[1] = 1.f/mask->drawable.height; |
// tmp->mask.offset[0] = -dst_x; |
// tmp->mask.offset[1] = -dst_y; |
tmp->u.gen6.flags = |
GEN6_SET_FLAGS(SAMPLER_OFFSET(tmp->src.filter, |
tmp->src.repeat, |
tmp->mask.filter, |
tmp->mask.repeat), |
gen6_get_blend(tmp->op, |
tmp->has_component_alpha, |
tmp->dst.format), |
/* gen6_choose_composite_kernel(tmp->op, |
tmp->mask.bo != NULL, |
tmp->has_component_alpha, |
tmp->is_affine), |
*/ |
GEN6_WM_KERNEL_MASK, |
gen4_choose_composite_emitter(sna, tmp)); |
tmp->blt = gen6_render_composite_blt; |
// tmp->box = gen6_render_composite_box; |
tmp->done = gen6_render_composite_done; |
kgem_set_mode(&sna->kgem, KGEM_RENDER, tmp->dst.bo); |
if (!kgem_check_bo(&sna->kgem, |
tmp->dst.bo, tmp->src.bo, tmp->mask.bo, |
NULL)) { |
kgem_submit(&sna->kgem); |
_kgem_set_mode(&sna->kgem, KGEM_RENDER); |
} |
gen6_emit_composite_state(sna, tmp); |
gen6_align_vertex(sna, tmp); |
return true; |
} |
/drivers/video/Intel-2D/sna/gen6_render.h |
---|
0,0 → 1,1563 |
#ifndef GEN6_RENDER_H |
#define GEN6_RENDER_H |
#define GEN6_MASK(high, low) (((1 << ((high) - (low) + 1)) - 1) << (low)) |
#define GEN6_3D(Pipeline,Opcode,Subopcode) ((3 << 29) | \ |
((Pipeline) << 27) | \ |
((Opcode) << 24) | \ |
((Subopcode) << 16)) |
#define GEN6_STATE_BASE_ADDRESS GEN6_3D(0, 1, 1) |
#define GEN6_STATE_SIP GEN6_3D(0, 1, 2) |
#define GEN6_PIPELINE_SELECT GEN6_3D(1, 1, 4) |
#define GEN6_MEDIA_STATE_POINTERS GEN6_3D(2, 0, 0) |
#define GEN6_MEDIA_OBJECT GEN6_3D(2, 1, 0) |
#define GEN6_3DSTATE_BINDING_TABLE_POINTERS GEN6_3D(3, 0, 1) |
# define GEN6_3DSTATE_BINDING_TABLE_MODIFY_PS (1 << 12)/* for GEN6 */ |
# define GEN6_3DSTATE_BINDING_TABLE_MODIFY_GS (1 << 9) /* for GEN6 */ |
# define GEN6_3DSTATE_BINDING_TABLE_MODIFY_VS (1 << 8) /* for GEN6 */ |
#define GEN6_3DSTATE_VERTEX_BUFFERS GEN6_3D(3, 0, 8) |
#define GEN6_3DSTATE_VERTEX_ELEMENTS GEN6_3D(3, 0, 9) |
#define GEN6_3DSTATE_INDEX_BUFFER GEN6_3D(3, 0, 0xa) |
#define GEN6_3DSTATE_VF_STATISTICS GEN6_3D(3, 0, 0xb) |
#define GEN6_3DSTATE_DRAWING_RECTANGLE GEN6_3D(3, 1, 0) |
#define GEN6_3DSTATE_CONSTANT_COLOR GEN6_3D(3, 1, 1) |
#define GEN6_3DSTATE_SAMPLER_PALETTE_LOAD GEN6_3D(3, 1, 2) |
#define GEN6_3DSTATE_CHROMA_KEY GEN6_3D(3, 1, 4) |
#define GEN6_3DSTATE_DEPTH_BUFFER GEN6_3D(3, 1, 5) |
# define GEN6_3DSTATE_DEPTH_BUFFER_TYPE_SHIFT 29 |
# define GEN6_3DSTATE_DEPTH_BUFFER_FORMAT_SHIFT 18 |
#define GEN6_3DSTATE_POLY_STIPPLE_OFFSET GEN6_3D(3, 1, 6) |
#define GEN6_3DSTATE_POLY_STIPPLE_PATTERN GEN6_3D(3, 1, 7) |
#define GEN6_3DSTATE_LINE_STIPPLE GEN6_3D(3, 1, 8) |
#define GEN6_3DSTATE_GLOBAL_DEPTH_OFFSET_CLAMP GEN6_3D(3, 1, 9) |
/* These two are BLC and CTG only, not BW or CL */ |
#define GEN6_3DSTATE_AA_LINE_PARAMS GEN6_3D(3, 1, 0xa) |
#define GEN6_3DSTATE_GS_SVB_INDEX GEN6_3D(3, 1, 0xb) |
#define GEN6_3DPRIMITIVE GEN6_3D(3, 3, 0) |
#define GEN6_3DSTATE_CLEAR_PARAMS GEN6_3D(3, 1, 0x10) |
/* DW1 */ |
# define GEN6_3DSTATE_DEPTH_CLEAR_VALID (1 << 15) |
#define GEN6_3DSTATE_SAMPLER_STATE_POINTERS GEN6_3D(3, 0, 0x02) |
# define GEN6_3DSTATE_SAMPLER_STATE_MODIFY_PS (1 << 12) |
# define GEN6_3DSTATE_SAMPLER_STATE_MODIFY_GS (1 << 9) |
# define GEN6_3DSTATE_SAMPLER_STATE_MODIFY_VS (1 << 8) |
#define GEN6_3DSTATE_URB GEN6_3D(3, 0, 0x05) |
/* DW1 */ |
# define GEN6_3DSTATE_URB_VS_SIZE_SHIFT 16 |
# define GEN6_3DSTATE_URB_VS_ENTRIES_SHIFT 0 |
/* DW2 */ |
# define GEN6_3DSTATE_URB_GS_ENTRIES_SHIFT 8 |
# define GEN6_3DSTATE_URB_GS_SIZE_SHIFT 0 |
#define GEN6_3DSTATE_VIEWPORT_STATE_POINTERS GEN6_3D(3, 0, 0x0d) |
# define GEN6_3DSTATE_VIEWPORT_STATE_MODIFY_CC (1 << 12) |
# define GEN6_3DSTATE_VIEWPORT_STATE_MODIFY_SF (1 << 11) |
# define GEN6_3DSTATE_VIEWPORT_STATE_MODIFY_CLIP (1 << 10) |
#define GEN6_3DSTATE_CC_STATE_POINTERS GEN6_3D(3, 0, 0x0e) |
#define GEN6_3DSTATE_VS GEN6_3D(3, 0, 0x10) |
#define GEN6_3DSTATE_GS GEN6_3D(3, 0, 0x11) |
/* DW4 */ |
# define GEN6_3DSTATE_GS_DISPATCH_START_GRF_SHIFT 0 |
#define GEN6_3DSTATE_CLIP GEN6_3D(3, 0, 0x12) |
#define GEN6_3DSTATE_SF GEN6_3D(3, 0, 0x13) |
/* DW1 */ |
# define GEN6_3DSTATE_SF_NUM_OUTPUTS_SHIFT 22 |
# define GEN6_3DSTATE_SF_URB_ENTRY_READ_LENGTH_SHIFT 11 |
# define GEN6_3DSTATE_SF_URB_ENTRY_READ_OFFSET_SHIFT 4 |
/* DW2 */ |
/* DW3 */ |
# define GEN6_3DSTATE_SF_CULL_BOTH (0 << 29) |
# define GEN6_3DSTATE_SF_CULL_NONE (1 << 29) |
# define GEN6_3DSTATE_SF_CULL_FRONT (2 << 29) |
# define GEN6_3DSTATE_SF_CULL_BACK (3 << 29) |
/* DW4 */ |
# define GEN6_3DSTATE_SF_TRI_PROVOKE_SHIFT 29 |
# define GEN6_3DSTATE_SF_LINE_PROVOKE_SHIFT 27 |
# define GEN6_3DSTATE_SF_TRIFAN_PROVOKE_SHIFT 25 |
#define GEN6_3DSTATE_WM GEN6_3D(3, 0, 0x14) |
/* DW2 */ |
# define GEN6_3DSTATE_WM_SAMPLER_COUNT_SHIFT 27 |
# define GEN6_3DSTATE_WM_BINDING_TABLE_ENTRY_COUNT_SHIFT 18 |
/* DW4 */ |
# define GEN6_3DSTATE_WM_DISPATCH_0_START_GRF_SHIFT 16 |
# define GEN6_3DSTATE_WM_DISPATCH_1_START_GRF_SHIFT 8 |
# define GEN6_3DSTATE_WM_DISPATCH_2_START_GRF_SHIFT 0 |
/* DW5 */ |
# define GEN6_3DSTATE_WM_MAX_THREADS_SHIFT 25 |
# define GEN6_3DSTATE_WM_DISPATCH_ENABLE (1 << 19) |
# define GEN6_3DSTATE_WM_32_DISPATCH_ENABLE (1 << 2) |
# define GEN6_3DSTATE_WM_16_DISPATCH_ENABLE (1 << 1) |
# define GEN6_3DSTATE_WM_8_DISPATCH_ENABLE (1 << 0) |
/* DW6 */ |
# define GEN6_3DSTATE_WM_NUM_SF_OUTPUTS_SHIFT 20 |
# define GEN6_3DSTATE_WM_NONPERSPECTIVE_SAMPLE_BARYCENTRIC (1 << 15) |
# define GEN6_3DSTATE_WM_NONPERSPECTIVE_CENTROID_BARYCENTRIC (1 << 14) |
# define GEN6_3DSTATE_WM_NONPERSPECTIVE_PIXEL_BARYCENTRIC (1 << 13) |
# define GEN6_3DSTATE_WM_PERSPECTIVE_SAMPLE_BARYCENTRIC (1 << 12) |
# define GEN6_3DSTATE_WM_PERSPECTIVE_CENTROID_BARYCENTRIC (1 << 11) |
# define GEN6_3DSTATE_WM_PERSPECTIVE_PIXEL_BARYCENTRIC (1 << 10) |
#define GEN6_3DSTATE_CONSTANT_VS GEN6_3D(3, 0, 0x15) |
#define GEN6_3DSTATE_CONSTANT_GS GEN6_3D(3, 0, 0x16) |
#define GEN6_3DSTATE_CONSTANT_PS GEN6_3D(3, 0, 0x17) |
#define GEN6_3DSTATE_SAMPLE_MASK GEN6_3D(3, 0, 0x18) |
#define GEN6_3DSTATE_MULTISAMPLE GEN6_3D(3, 1, 0x0d) |
/* DW1 */ |
# define GEN6_3DSTATE_MULTISAMPLE_PIXEL_LOCATION_CENTER (0 << 4) |
# define GEN6_3DSTATE_MULTISAMPLE_PIXEL_LOCATION_UPPER_LEFT (1 << 4) |
# define GEN6_3DSTATE_MULTISAMPLE_NUMSAMPLES_1 (0 << 1) |
# define GEN6_3DSTATE_MULTISAMPLE_NUMSAMPLES_4 (2 << 1) |
# define GEN6_3DSTATE_MULTISAMPLE_NUMSAMPLES_8 (3 << 1) |
#define PIPELINE_SELECT_3D 0 |
#define PIPELINE_SELECT_MEDIA 1 |
/* for GEN6_STATE_BASE_ADDRESS */ |
#define BASE_ADDRESS_MODIFY (1 << 0) |
/* VERTEX_BUFFER_STATE Structure */ |
#define VB0_BUFFER_INDEX_SHIFT 26 |
#define VB0_VERTEXDATA (0 << 20) |
#define VB0_INSTANCEDATA (1 << 20) |
#define VB0_BUFFER_PITCH_SHIFT 0 |
/* VERTEX_ELEMENT_STATE Structure */ |
#define VE0_VERTEX_BUFFER_INDEX_SHIFT 26 /* for GEN6 */ |
#define VE0_VALID (1 << 25) /* for GEN6 */ |
#define VE0_FORMAT_SHIFT 16 |
#define VE0_OFFSET_SHIFT 0 |
#define VE1_VFCOMPONENT_0_SHIFT 28 |
#define VE1_VFCOMPONENT_1_SHIFT 24 |
#define VE1_VFCOMPONENT_2_SHIFT 20 |
#define VE1_VFCOMPONENT_3_SHIFT 16 |
#define VE1_DESTINATION_ELEMENT_OFFSET_SHIFT 0 |
/* 3DPRIMITIVE bits */ |
#define GEN6_3DPRIMITIVE_VERTEX_SEQUENTIAL (0 << 15) |
#define GEN6_3DPRIMITIVE_VERTEX_RANDOM (1 << 15) |
/* Primitive types are in gen6_defines.h */ |
#define GEN6_3DPRIMITIVE_TOPOLOGY_SHIFT 10 |
#define GEN6_SVG_CTL 0x7400 |
#define GEN6_SVG_CTL_GS_BA (0 << 8) |
#define GEN6_SVG_CTL_SS_BA (1 << 8) |
#define GEN6_SVG_CTL_IO_BA (2 << 8) |
#define GEN6_SVG_CTL_GS_AUB (3 << 8) |
#define GEN6_SVG_CTL_IO_AUB (4 << 8) |
#define GEN6_SVG_CTL_SIP (5 << 8) |
#define GEN6_SVG_RDATA 0x7404 |
#define GEN6_SVG_WORK_CTL 0x7408 |
#define GEN6_VF_CTL 0x7500 |
#define GEN6_VF_CTL_SNAPSHOT_COMPLETE (1 << 31) |
#define GEN6_VF_CTL_SNAPSHOT_MUX_SELECT_THREADID (0 << 8) |
#define GEN6_VF_CTL_SNAPSHOT_MUX_SELECT_VF_DEBUG (1 << 8) |
#define GEN6_VF_CTL_SNAPSHOT_TYPE_VERTEX_SEQUENCE (0 << 4) |
#define GEN6_VF_CTL_SNAPSHOT_TYPE_VERTEX_INDEX (1 << 4) |
#define GEN6_VF_CTL_SKIP_INITIAL_PRIMITIVES (1 << 3) |
#define GEN6_VF_CTL_MAX_PRIMITIVES_LIMIT_ENABLE (1 << 2) |
#define GEN6_VF_CTL_VERTEX_RANGE_LIMIT_ENABLE (1 << 1) |
#define GEN6_VF_CTL_SNAPSHOT_ENABLE (1 << 0) |
#define GEN6_VF_STRG_VAL 0x7504 |
#define GEN6_VF_STR_VL_OVR 0x7508 |
#define GEN6_VF_VC_OVR 0x750c |
#define GEN6_VF_STR_PSKIP 0x7510 |
#define GEN6_VF_MAX_PRIM 0x7514 |
#define GEN6_VF_RDATA 0x7518 |
#define GEN6_VS_CTL 0x7600 |
#define GEN6_VS_CTL_SNAPSHOT_COMPLETE (1 << 31) |
#define GEN6_VS_CTL_SNAPSHOT_MUX_VERTEX_0 (0 << 8) |
#define GEN6_VS_CTL_SNAPSHOT_MUX_VERTEX_1 (1 << 8) |
#define GEN6_VS_CTL_SNAPSHOT_MUX_VALID_COUNT (2 << 8) |
#define GEN6_VS_CTL_SNAPSHOT_MUX_VS_KERNEL_POINTER (3 << 8) |
#define GEN6_VS_CTL_SNAPSHOT_ALL_THREADS (1 << 2) |
#define GEN6_VS_CTL_THREAD_SNAPSHOT_ENABLE (1 << 1) |
#define GEN6_VS_CTL_SNAPSHOT_ENABLE (1 << 0) |
#define GEN6_VS_STRG_VAL 0x7604 |
#define GEN6_VS_RDATA 0x7608 |
#define GEN6_SF_CTL 0x7b00 |
#define GEN6_SF_CTL_SNAPSHOT_COMPLETE (1 << 31) |
#define GEN6_SF_CTL_SNAPSHOT_MUX_VERTEX_0_FF_ID (0 << 8) |
#define GEN6_SF_CTL_SNAPSHOT_MUX_VERTEX_0_REL_COUNT (1 << 8) |
#define GEN6_SF_CTL_SNAPSHOT_MUX_VERTEX_1_FF_ID (2 << 8) |
#define GEN6_SF_CTL_SNAPSHOT_MUX_VERTEX_1_REL_COUNT (3 << 8) |
#define GEN6_SF_CTL_SNAPSHOT_MUX_VERTEX_2_FF_ID (4 << 8) |
#define GEN6_SF_CTL_SNAPSHOT_MUX_VERTEX_2_REL_COUNT (5 << 8) |
#define GEN6_SF_CTL_SNAPSHOT_MUX_VERTEX_COUNT (6 << 8) |
#define GEN6_SF_CTL_SNAPSHOT_MUX_SF_KERNEL_POINTER (7 << 8) |
#define GEN6_SF_CTL_MIN_MAX_PRIMITIVE_RANGE_ENABLE (1 << 4) |
#define GEN6_SF_CTL_DEBUG_CLIP_RECTANGLE_ENABLE (1 << 3) |
#define GEN6_SF_CTL_SNAPSHOT_ALL_THREADS (1 << 2) |
#define GEN6_SF_CTL_THREAD_SNAPSHOT_ENABLE (1 << 1) |
#define GEN6_SF_CTL_SNAPSHOT_ENABLE (1 << 0) |
#define GEN6_SF_STRG_VAL 0x7b04 |
#define GEN6_SF_RDATA 0x7b18 |
#define GEN6_WIZ_CTL 0x7c00 |
#define GEN6_WIZ_CTL_SNAPSHOT_COMPLETE (1 << 31) |
#define GEN6_WIZ_CTL_SUBSPAN_INSTANCE_SHIFT 16 |
#define GEN6_WIZ_CTL_SNAPSHOT_MUX_WIZ_KERNEL_POINTER (0 << 8) |
#define GEN6_WIZ_CTL_SNAPSHOT_MUX_SUBSPAN_INSTANCE (1 << 8) |
#define GEN6_WIZ_CTL_SNAPSHOT_MUX_PRIMITIVE_SEQUENCE (2 << 8) |
#define GEN6_WIZ_CTL_SINGLE_SUBSPAN_DISPATCH (1 << 6) |
#define GEN6_WIZ_CTL_IGNORE_COLOR_SCOREBOARD_STALLS (1 << 5) |
#define GEN6_WIZ_CTL_ENABLE_SUBSPAN_INSTANCE_COMPARE (1 << 4) |
#define GEN6_WIZ_CTL_USE_UPSTREAM_SNAPSHOT_FLAG (1 << 3) |
#define GEN6_WIZ_CTL_SNAPSHOT_ALL_THREADS (1 << 2) |
#define GEN6_WIZ_CTL_THREAD_SNAPSHOT_ENABLE (1 << 1) |
#define GEN6_WIZ_CTL_SNAPSHOT_ENABLE (1 << 0) |
#define GEN6_WIZ_STRG_VAL 0x7c04 |
#define GEN6_WIZ_RDATA 0x7c18 |
#define GEN6_TS_CTL 0x7e00 |
#define GEN6_TS_CTL_SNAPSHOT_COMPLETE (1 << 31) |
#define GEN6_TS_CTL_SNAPSHOT_MESSAGE_ERROR (0 << 8) |
#define GEN6_TS_CTL_SNAPSHOT_INTERFACE_DESCRIPTOR (3 << 8) |
#define GEN6_TS_CTL_SNAPSHOT_ALL_CHILD_THREADS (1 << 2) |
#define GEN6_TS_CTL_SNAPSHOT_ALL_ROOT_THREADS (1 << 1) |
#define GEN6_TS_CTL_SNAPSHOT_ENABLE (1 << 0) |
#define GEN6_TS_STRG_VAL 0x7e04 |
#define GEN6_TS_RDATA 0x7e08 |
#define GEN6_TD_CTL 0x8000 |
#define GEN6_TD_CTL_MUX_SHIFT 8 |
#define GEN6_TD_CTL_EXTERNAL_HALT_R0_DEBUG_MATCH (1 << 7) |
#define GEN6_TD_CTL_FORCE_EXTERNAL_HALT (1 << 6) |
#define GEN6_TD_CTL_EXCEPTION_MASK_OVERRIDE (1 << 5) |
#define GEN6_TD_CTL_FORCE_THREAD_BREAKPOINT_ENABLE (1 << 4) |
#define GEN6_TD_CTL_BREAKPOINT_ENABLE (1 << 2) |
#define GEN6_TD_CTL2 0x8004 |
#define GEN6_TD_CTL2_ILLEGAL_OPCODE_EXCEPTION_OVERRIDE (1 << 28) |
#define GEN6_TD_CTL2_MASKSTACK_EXCEPTION_OVERRIDE (1 << 26) |
#define GEN6_TD_CTL2_SOFTWARE_EXCEPTION_OVERRIDE (1 << 25) |
#define GEN6_TD_CTL2_ACTIVE_THREAD_LIMIT_SHIFT 16 |
#define GEN6_TD_CTL2_ACTIVE_THREAD_LIMIT_ENABLE (1 << 8) |
#define GEN6_TD_CTL2_THREAD_SPAWNER_EXECUTION_MASK_ENABLE (1 << 7) |
#define GEN6_TD_CTL2_WIZ_EXECUTION_MASK_ENABLE (1 << 6) |
#define GEN6_TD_CTL2_SF_EXECUTION_MASK_ENABLE (1 << 5) |
#define GEN6_TD_CTL2_CLIPPER_EXECUTION_MASK_ENABLE (1 << 4) |
#define GEN6_TD_CTL2_GS_EXECUTION_MASK_ENABLE (1 << 3) |
#define GEN6_TD_CTL2_VS_EXECUTION_MASK_ENABLE (1 << 0) |
#define GEN6_TD_VF_VS_EMSK 0x8008 |
#define GEN6_TD_GS_EMSK 0x800c |
#define GEN6_TD_CLIP_EMSK 0x8010 |
#define GEN6_TD_SF_EMSK 0x8014 |
#define GEN6_TD_WIZ_EMSK 0x8018 |
#define GEN6_TD_0_6_EHTRG_VAL 0x801c |
#define GEN6_TD_0_7_EHTRG_VAL 0x8020 |
#define GEN6_TD_0_6_EHTRG_MSK 0x8024 |
#define GEN6_TD_0_7_EHTRG_MSK 0x8028 |
#define GEN6_TD_RDATA 0x802c |
#define GEN6_TD_TS_EMSK 0x8030 |
#define GEN6_EU_CTL 0x8800 |
#define GEN6_EU_CTL_SELECT_SHIFT 16 |
#define GEN6_EU_CTL_DATA_MUX_SHIFT 8 |
#define GEN6_EU_ATT_0 0x8810 |
#define GEN6_EU_ATT_1 0x8814 |
#define GEN6_EU_ATT_DATA_0 0x8820 |
#define GEN6_EU_ATT_DATA_1 0x8824 |
#define GEN6_EU_ATT_CLR_0 0x8830 |
#define GEN6_EU_ATT_CLR_1 0x8834 |
#define GEN6_EU_RDATA 0x8840 |
#define GEN6_3D(Pipeline,Opcode,Subopcode) ((3 << 29) | \ |
((Pipeline) << 27) | \ |
((Opcode) << 24) | \ |
((Subopcode) << 16)) |
#define GEN6_STATE_BASE_ADDRESS GEN6_3D(0, 1, 1) |
#define GEN6_STATE_SIP GEN6_3D(0, 1, 2) |
#define GEN6_PIPELINE_SELECT GEN6_3D(1, 1, 4) |
#define GEN6_MEDIA_STATE_POINTERS GEN6_3D(2, 0, 0) |
#define GEN6_MEDIA_OBJECT GEN6_3D(2, 1, 0) |
#define GEN6_3DSTATE_BINDING_TABLE_POINTERS GEN6_3D(3, 0, 1) |
# define GEN6_3DSTATE_BINDING_TABLE_MODIFY_PS (1 << 12)/* for GEN6 */ |
# define GEN6_3DSTATE_BINDING_TABLE_MODIFY_GS (1 << 9) /* for GEN6 */ |
# define GEN6_3DSTATE_BINDING_TABLE_MODIFY_VS (1 << 8) /* for GEN6 */ |
#define GEN6_3DSTATE_VERTEX_BUFFERS GEN6_3D(3, 0, 8) |
#define GEN6_3DSTATE_VERTEX_ELEMENTS GEN6_3D(3, 0, 9) |
#define GEN6_3DSTATE_INDEX_BUFFER GEN6_3D(3, 0, 0xa) |
#define GEN6_3DSTATE_VF_STATISTICS GEN6_3D(3, 0, 0xb) |
#define GEN6_3DSTATE_DRAWING_RECTANGLE GEN6_3D(3, 1, 0) |
#define GEN6_3DSTATE_CONSTANT_COLOR GEN6_3D(3, 1, 1) |
#define GEN6_3DSTATE_SAMPLER_PALETTE_LOAD GEN6_3D(3, 1, 2) |
#define GEN6_3DSTATE_CHROMA_KEY GEN6_3D(3, 1, 4) |
#define GEN6_3DSTATE_DEPTH_BUFFER GEN6_3D(3, 1, 5) |
# define GEN6_3DSTATE_DEPTH_BUFFER_TYPE_SHIFT 29 |
# define GEN6_3DSTATE_DEPTH_BUFFER_FORMAT_SHIFT 18 |
#define GEN6_3DSTATE_POLY_STIPPLE_OFFSET GEN6_3D(3, 1, 6) |
#define GEN6_3DSTATE_POLY_STIPPLE_PATTERN GEN6_3D(3, 1, 7) |
#define GEN6_3DSTATE_LINE_STIPPLE GEN6_3D(3, 1, 8) |
#define GEN6_3DSTATE_GLOBAL_DEPTH_OFFSET_CLAMP GEN6_3D(3, 1, 9) |
/* These two are BLC and CTG only, not BW or CL */ |
#define GEN6_3DSTATE_AA_LINE_PARAMS GEN6_3D(3, 1, 0xa) |
#define GEN6_3DSTATE_GS_SVB_INDEX GEN6_3D(3, 1, 0xb) |
#define GEN6_3DPRIMITIVE GEN6_3D(3, 3, 0) |
#define GEN6_3DSTATE_CLEAR_PARAMS GEN6_3D(3, 1, 0x10) |
/* DW1 */ |
# define GEN6_3DSTATE_DEPTH_CLEAR_VALID (1 << 15) |
/* for GEN6+ */ |
#define GEN6_3DSTATE_SAMPLER_STATE_POINTERS GEN6_3D(3, 0, 0x02) |
# define GEN6_3DSTATE_SAMPLER_STATE_MODIFY_PS (1 << 12) |
# define GEN6_3DSTATE_SAMPLER_STATE_MODIFY_GS (1 << 9) |
# define GEN6_3DSTATE_SAMPLER_STATE_MODIFY_VS (1 << 8) |
#define GEN6_3DSTATE_URB GEN6_3D(3, 0, 0x05) |
/* DW1 */ |
# define GEN6_3DSTATE_URB_VS_SIZE_SHIFT 16 |
# define GEN6_3DSTATE_URB_VS_ENTRIES_SHIFT 0 |
/* DW2 */ |
# define GEN6_3DSTATE_URB_GS_ENTRIES_SHIFT 8 |
# define GEN6_3DSTATE_URB_GS_SIZE_SHIFT 0 |
#define GEN6_3DSTATE_VIEWPORT_STATE_POINTERS GEN6_3D(3, 0, 0x0d) |
# define GEN6_3DSTATE_VIEWPORT_STATE_MODIFY_CC (1 << 12) |
# define GEN6_3DSTATE_VIEWPORT_STATE_MODIFY_SF (1 << 11) |
# define GEN6_3DSTATE_VIEWPORT_STATE_MODIFY_CLIP (1 << 10) |
#define GEN6_3DSTATE_CC_STATE_POINTERS GEN6_3D(3, 0, 0x0e) |
#define GEN6_3DSTATE_VS GEN6_3D(3, 0, 0x10) |
#define GEN6_3DSTATE_GS GEN6_3D(3, 0, 0x11) |
/* DW4 */ |
# define GEN6_3DSTATE_GS_DISPATCH_START_GRF_SHIFT 0 |
#define GEN6_3DSTATE_CLIP GEN6_3D(3, 0, 0x12) |
#define GEN6_3DSTATE_SF GEN6_3D(3, 0, 0x13) |
/* DW1 */ |
# define GEN6_3DSTATE_SF_NUM_OUTPUTS_SHIFT 22 |
# define GEN6_3DSTATE_SF_URB_ENTRY_READ_LENGTH_SHIFT 11 |
# define GEN6_3DSTATE_SF_URB_ENTRY_READ_OFFSET_SHIFT 4 |
/* DW2 */ |
/* DW3 */ |
# define GEN6_3DSTATE_SF_CULL_BOTH (0 << 29) |
# define GEN6_3DSTATE_SF_CULL_NONE (1 << 29) |
# define GEN6_3DSTATE_SF_CULL_FRONT (2 << 29) |
# define GEN6_3DSTATE_SF_CULL_BACK (3 << 29) |
/* DW4 */ |
# define GEN6_3DSTATE_SF_TRI_PROVOKE_SHIFT 29 |
# define GEN6_3DSTATE_SF_LINE_PROVOKE_SHIFT 27 |
# define GEN6_3DSTATE_SF_TRIFAN_PROVOKE_SHIFT 25 |
#define GEN6_3DSTATE_WM GEN6_3D(3, 0, 0x14) |
/* DW2 */ |
# define GEN6_3DSTATE_WM_SAMPLER_COUNT_SHITF 27 |
# define GEN6_3DSTATE_WM_BINDING_TABLE_ENTRY_COUNT_SHIFT 18 |
/* DW4 */ |
# define GEN6_3DSTATE_WM_DISPATCH_START_GRF_0_SHIFT 16 |
/* DW5 */ |
# define GEN6_3DSTATE_WM_MAX_THREADS_SHIFT 25 |
# define GEN6_3DSTATE_WM_DISPATCH_ENABLE (1 << 19) |
# define GEN6_3DSTATE_WM_16_DISPATCH_ENABLE (1 << 1) |
# define GEN6_3DSTATE_WM_8_DISPATCH_ENABLE (1 << 0) |
/* DW6 */ |
# define GEN6_3DSTATE_WM_NUM_SF_OUTPUTS_SHIFT 20 |
# define GEN6_3DSTATE_WM_NONPERSPECTIVE_SAMPLE_BARYCENTRIC (1 << 15) |
# define GEN6_3DSTATE_WM_NONPERSPECTIVE_CENTROID_BARYCENTRIC (1 << 14) |
# define GEN6_3DSTATE_WM_NONPERSPECTIVE_PIXEL_BARYCENTRIC (1 << 13) |
# define GEN6_3DSTATE_WM_PERSPECTIVE_SAMPLE_BARYCENTRIC (1 << 12) |
# define GEN6_3DSTATE_WM_PERSPECTIVE_CENTROID_BARYCENTRIC (1 << 11) |
# define GEN6_3DSTATE_WM_PERSPECTIVE_PIXEL_BARYCENTRIC (1 << 10) |
#define GEN6_3DSTATE_CONSTANT_VS GEN6_3D(3, 0, 0x15) |
#define GEN6_3DSTATE_CONSTANT_GS GEN6_3D(3, 0, 0x16) |
#define GEN6_3DSTATE_CONSTANT_PS GEN6_3D(3, 0, 0x17) |
#define GEN6_3DSTATE_SAMPLE_MASK GEN6_3D(3, 0, 0x18) |
#define GEN6_3DSTATE_MULTISAMPLE GEN6_3D(3, 1, 0x0d) |
/* DW1 */ |
# define GEN6_3DSTATE_MULTISAMPLE_PIXEL_LOCATION_CENTER (0 << 4) |
# define GEN6_3DSTATE_MULTISAMPLE_PIXEL_LOCATION_UPPER_LEFT (1 << 4) |
# define GEN6_3DSTATE_MULTISAMPLE_NUMSAMPLES_1 (0 << 1) |
# define GEN6_3DSTATE_MULTISAMPLE_NUMSAMPLES_4 (2 << 1) |
# define GEN6_3DSTATE_MULTISAMPLE_NUMSAMPLES_8 (3 << 1) |
#define PIPELINE_SELECT_3D 0 |
#define PIPELINE_SELECT_MEDIA 1 |
#define UF0_CS_REALLOC (1 << 13) |
#define UF0_VFE_REALLOC (1 << 12) |
#define UF0_SF_REALLOC (1 << 11) |
#define UF0_CLIP_REALLOC (1 << 10) |
#define UF0_GS_REALLOC (1 << 9) |
#define UF0_VS_REALLOC (1 << 8) |
#define UF1_CLIP_FENCE_SHIFT 20 |
#define UF1_GS_FENCE_SHIFT 10 |
#define UF1_VS_FENCE_SHIFT 0 |
#define UF2_CS_FENCE_SHIFT 20 |
#define UF2_VFE_FENCE_SHIFT 10 |
#define UF2_SF_FENCE_SHIFT 0 |
/* for GEN6_STATE_BASE_ADDRESS */ |
#define BASE_ADDRESS_MODIFY (1 << 0) |
/* for GEN6_3DSTATE_PIPELINED_POINTERS */ |
#define GEN6_GS_DISABLE 0 |
#define GEN6_GS_ENABLE 1 |
#define GEN6_CLIP_DISABLE 0 |
#define GEN6_CLIP_ENABLE 1 |
/* for GEN6_PIPE_CONTROL */ |
#define GEN6_PIPE_CONTROL GEN6_3D(3, 2, 0) |
#define GEN6_PIPE_CONTROL_CS_STALL (1 << 20) |
#define GEN6_PIPE_CONTROL_NOWRITE (0 << 14) |
#define GEN6_PIPE_CONTROL_WRITE_QWORD (1 << 14) |
#define GEN6_PIPE_CONTROL_WRITE_DEPTH (2 << 14) |
#define GEN6_PIPE_CONTROL_WRITE_TIME (3 << 14) |
#define GEN6_PIPE_CONTROL_DEPTH_STALL (1 << 13) |
#define GEN6_PIPE_CONTROL_WC_FLUSH (1 << 12) |
#define GEN6_PIPE_CONTROL_IS_FLUSH (1 << 11) |
#define GEN6_PIPE_CONTROL_TC_FLUSH (1 << 10) |
#define GEN6_PIPE_CONTROL_NOTIFY_ENABLE (1 << 8) |
#define GEN6_PIPE_CONTROL_GLOBAL_GTT (1 << 2) |
#define GEN6_PIPE_CONTROL_LOCAL_PGTT (0 << 2) |
#define GEN6_PIPE_CONTROL_STALL_AT_SCOREBOARD (1 << 1) |
#define GEN6_PIPE_CONTROL_DEPTH_CACHE_FLUSH (1 << 0) |
/* 3DPRIMITIVE bits */ |
#define GEN6_3DPRIMITIVE_VERTEX_SEQUENTIAL (0 << 15) |
#define GEN6_3DPRIMITIVE_VERTEX_RANDOM (1 << 15) |
/* Primitive types are in gen6_defines.h */ |
#define GEN6_3DPRIMITIVE_TOPOLOGY_SHIFT 10 |
#define GEN6_SVG_CTL 0x7400 |
#define GEN6_SVG_CTL_GS_BA (0 << 8) |
#define GEN6_SVG_CTL_SS_BA (1 << 8) |
#define GEN6_SVG_CTL_IO_BA (2 << 8) |
#define GEN6_SVG_CTL_GS_AUB (3 << 8) |
#define GEN6_SVG_CTL_IO_AUB (4 << 8) |
#define GEN6_SVG_CTL_SIP (5 << 8) |
#define GEN6_SVG_RDATA 0x7404 |
#define GEN6_SVG_WORK_CTL 0x7408 |
#define GEN6_VF_CTL 0x7500 |
#define GEN6_VF_CTL_SNAPSHOT_COMPLETE (1 << 31) |
#define GEN6_VF_CTL_SNAPSHOT_MUX_SELECT_THREADID (0 << 8) |
#define GEN6_VF_CTL_SNAPSHOT_MUX_SELECT_VF_DEBUG (1 << 8) |
#define GEN6_VF_CTL_SNAPSHOT_TYPE_VERTEX_SEQUENCE (0 << 4) |
#define GEN6_VF_CTL_SNAPSHOT_TYPE_VERTEX_INDEX (1 << 4) |
#define GEN6_VF_CTL_SKIP_INITIAL_PRIMITIVES (1 << 3) |
#define GEN6_VF_CTL_MAX_PRIMITIVES_LIMIT_ENABLE (1 << 2) |
#define GEN6_VF_CTL_VERTEX_RANGE_LIMIT_ENABLE (1 << 1) |
#define GEN6_VF_CTL_SNAPSHOT_ENABLE (1 << 0) |
#define GEN6_VF_STRG_VAL 0x7504 |
#define GEN6_VF_STR_VL_OVR 0x7508 |
#define GEN6_VF_VC_OVR 0x750c |
#define GEN6_VF_STR_PSKIP 0x7510 |
#define GEN6_VF_MAX_PRIM 0x7514 |
#define GEN6_VF_RDATA 0x7518 |
#define GEN6_VS_CTL 0x7600 |
#define GEN6_VS_CTL_SNAPSHOT_COMPLETE (1 << 31) |
#define GEN6_VS_CTL_SNAPSHOT_MUX_VERTEX_0 (0 << 8) |
#define GEN6_VS_CTL_SNAPSHOT_MUX_VERTEX_1 (1 << 8) |
#define GEN6_VS_CTL_SNAPSHOT_MUX_VALID_COUNT (2 << 8) |
#define GEN6_VS_CTL_SNAPSHOT_MUX_VS_KERNEL_POINTER (3 << 8) |
#define GEN6_VS_CTL_SNAPSHOT_ALL_THREADS (1 << 2) |
#define GEN6_VS_CTL_THREAD_SNAPSHOT_ENABLE (1 << 1) |
#define GEN6_VS_CTL_SNAPSHOT_ENABLE (1 << 0) |
#define GEN6_VS_STRG_VAL 0x7604 |
#define GEN6_VS_RDATA 0x7608 |
#define GEN6_SF_CTL 0x7b00 |
#define GEN6_SF_CTL_SNAPSHOT_COMPLETE (1 << 31) |
#define GEN6_SF_CTL_SNAPSHOT_MUX_VERTEX_0_FF_ID (0 << 8) |
#define GEN6_SF_CTL_SNAPSHOT_MUX_VERTEX_0_REL_COUNT (1 << 8) |
#define GEN6_SF_CTL_SNAPSHOT_MUX_VERTEX_1_FF_ID (2 << 8) |
#define GEN6_SF_CTL_SNAPSHOT_MUX_VERTEX_1_REL_COUNT (3 << 8) |
#define GEN6_SF_CTL_SNAPSHOT_MUX_VERTEX_2_FF_ID (4 << 8) |
#define GEN6_SF_CTL_SNAPSHOT_MUX_VERTEX_2_REL_COUNT (5 << 8) |
#define GEN6_SF_CTL_SNAPSHOT_MUX_VERTEX_COUNT (6 << 8) |
#define GEN6_SF_CTL_SNAPSHOT_MUX_SF_KERNEL_POINTER (7 << 8) |
#define GEN6_SF_CTL_MIN_MAX_PRIMITIVE_RANGE_ENABLE (1 << 4) |
#define GEN6_SF_CTL_DEBUG_CLIP_RECTANGLE_ENABLE (1 << 3) |
#define GEN6_SF_CTL_SNAPSHOT_ALL_THREADS (1 << 2) |
#define GEN6_SF_CTL_THREAD_SNAPSHOT_ENABLE (1 << 1) |
#define GEN6_SF_CTL_SNAPSHOT_ENABLE (1 << 0) |
#define GEN6_SF_STRG_VAL 0x7b04 |
#define GEN6_SF_RDATA 0x7b18 |
#define GEN6_WIZ_CTL 0x7c00 |
#define GEN6_WIZ_CTL_SNAPSHOT_COMPLETE (1 << 31) |
#define GEN6_WIZ_CTL_SUBSPAN_INSTANCE_SHIFT 16 |
#define GEN6_WIZ_CTL_SNAPSHOT_MUX_WIZ_KERNEL_POINTER (0 << 8) |
#define GEN6_WIZ_CTL_SNAPSHOT_MUX_SUBSPAN_INSTANCE (1 << 8) |
#define GEN6_WIZ_CTL_SNAPSHOT_MUX_PRIMITIVE_SEQUENCE (2 << 8) |
#define GEN6_WIZ_CTL_SINGLE_SUBSPAN_DISPATCH (1 << 6) |
#define GEN6_WIZ_CTL_IGNORE_COLOR_SCOREBOARD_STALLS (1 << 5) |
#define GEN6_WIZ_CTL_ENABLE_SUBSPAN_INSTANCE_COMPARE (1 << 4) |
#define GEN6_WIZ_CTL_USE_UPSTREAM_SNAPSHOT_FLAG (1 << 3) |
#define GEN6_WIZ_CTL_SNAPSHOT_ALL_THREADS (1 << 2) |
#define GEN6_WIZ_CTL_THREAD_SNAPSHOT_ENABLE (1 << 1) |
#define GEN6_WIZ_CTL_SNAPSHOT_ENABLE (1 << 0) |
#define GEN6_WIZ_STRG_VAL 0x7c04 |
#define GEN6_WIZ_RDATA 0x7c18 |
#define GEN6_TS_CTL 0x7e00 |
#define GEN6_TS_CTL_SNAPSHOT_COMPLETE (1 << 31) |
#define GEN6_TS_CTL_SNAPSHOT_MESSAGE_ERROR (0 << 8) |
#define GEN6_TS_CTL_SNAPSHOT_INTERFACE_DESCRIPTOR (3 << 8) |
#define GEN6_TS_CTL_SNAPSHOT_ALL_CHILD_THREADS (1 << 2) |
#define GEN6_TS_CTL_SNAPSHOT_ALL_ROOT_THREADS (1 << 1) |
#define GEN6_TS_CTL_SNAPSHOT_ENABLE (1 << 0) |
#define GEN6_TS_STRG_VAL 0x7e04 |
#define GEN6_TS_RDATA 0x7e08 |
#define GEN6_TD_CTL 0x8000 |
#define GEN6_TD_CTL_MUX_SHIFT 8 |
#define GEN6_TD_CTL_EXTERNAL_HALT_R0_DEBUG_MATCH (1 << 7) |
#define GEN6_TD_CTL_FORCE_EXTERNAL_HALT (1 << 6) |
#define GEN6_TD_CTL_EXCEPTION_MASK_OVERRIDE (1 << 5) |
#define GEN6_TD_CTL_FORCE_THREAD_BREAKPOINT_ENABLE (1 << 4) |
#define GEN6_TD_CTL_BREAKPOINT_ENABLE (1 << 2) |
#define GEN6_TD_CTL2 0x8004 |
#define GEN6_TD_CTL2_ILLEGAL_OPCODE_EXCEPTION_OVERRIDE (1 << 28) |
#define GEN6_TD_CTL2_MASKSTACK_EXCEPTION_OVERRIDE (1 << 26) |
#define GEN6_TD_CTL2_SOFTWARE_EXCEPTION_OVERRIDE (1 << 25) |
#define GEN6_TD_CTL2_ACTIVE_THREAD_LIMIT_SHIFT 16 |
#define GEN6_TD_CTL2_ACTIVE_THREAD_LIMIT_ENABLE (1 << 8) |
#define GEN6_TD_CTL2_THREAD_SPAWNER_EXECUTION_MASK_ENABLE (1 << 7) |
#define GEN6_TD_CTL2_WIZ_EXECUTION_MASK_ENABLE (1 << 6) |
#define GEN6_TD_CTL2_SF_EXECUTION_MASK_ENABLE (1 << 5) |
#define GEN6_TD_CTL2_CLIPPER_EXECUTION_MASK_ENABLE (1 << 4) |
#define GEN6_TD_CTL2_GS_EXECUTION_MASK_ENABLE (1 << 3) |
#define GEN6_TD_CTL2_VS_EXECUTION_MASK_ENABLE (1 << 0) |
#define GEN6_TD_VF_VS_EMSK 0x8008 |
#define GEN6_TD_GS_EMSK 0x800c |
#define GEN6_TD_CLIP_EMSK 0x8010 |
#define GEN6_TD_SF_EMSK 0x8014 |
#define GEN6_TD_WIZ_EMSK 0x8018 |
#define GEN6_TD_0_6_EHTRG_VAL 0x801c |
#define GEN6_TD_0_7_EHTRG_VAL 0x8020 |
#define GEN6_TD_0_6_EHTRG_MSK 0x8024 |
#define GEN6_TD_0_7_EHTRG_MSK 0x8028 |
#define GEN6_TD_RDATA 0x802c |
#define GEN6_TD_TS_EMSK 0x8030 |
#define GEN6_EU_CTL 0x8800 |
#define GEN6_EU_CTL_SELECT_SHIFT 16 |
#define GEN6_EU_CTL_DATA_MUX_SHIFT 8 |
#define GEN6_EU_ATT_0 0x8810 |
#define GEN6_EU_ATT_1 0x8814 |
#define GEN6_EU_ATT_DATA_0 0x8820 |
#define GEN6_EU_ATT_DATA_1 0x8824 |
#define GEN6_EU_ATT_CLR_0 0x8830 |
#define GEN6_EU_ATT_CLR_1 0x8834 |
#define GEN6_EU_RDATA 0x8840 |
/* 3D state: |
*/ |
#define _3DOP_3DSTATE_PIPELINED 0x0 |
#define _3DOP_3DSTATE_NONPIPELINED 0x1 |
#define _3DOP_3DCONTROL 0x2 |
#define _3DOP_3DPRIMITIVE 0x3 |
#define _3DSTATE_PIPELINED_POINTERS 0x00 |
#define _3DSTATE_BINDING_TABLE_POINTERS 0x01 |
#define _3DSTATE_VERTEX_BUFFERS 0x08 |
#define _3DSTATE_VERTEX_ELEMENTS 0x09 |
#define _3DSTATE_INDEX_BUFFER 0x0A |
#define _3DSTATE_VF_STATISTICS 0x0B |
#define _3DSTATE_DRAWING_RECTANGLE 0x00 |
#define _3DSTATE_CONSTANT_COLOR 0x01 |
#define _3DSTATE_SAMPLER_PALETTE_LOAD 0x02 |
#define _3DSTATE_CHROMA_KEY 0x04 |
#define _3DSTATE_DEPTH_BUFFER 0x05 |
#define _3DSTATE_POLY_STIPPLE_OFFSET 0x06 |
#define _3DSTATE_POLY_STIPPLE_PATTERN 0x07 |
#define _3DSTATE_LINE_STIPPLE 0x08 |
#define _3DSTATE_GLOBAL_DEPTH_OFFSET_CLAMP 0x09 |
#define _3DCONTROL 0x00 |
#define _3DPRIMITIVE 0x00 |
#define _3DPRIM_POINTLIST 0x01 |
#define _3DPRIM_LINELIST 0x02 |
#define _3DPRIM_LINESTRIP 0x03 |
#define _3DPRIM_TRILIST 0x04 |
#define _3DPRIM_TRISTRIP 0x05 |
#define _3DPRIM_TRIFAN 0x06 |
#define _3DPRIM_QUADLIST 0x07 |
#define _3DPRIM_QUADSTRIP 0x08 |
#define _3DPRIM_LINELIST_ADJ 0x09 |
#define _3DPRIM_LINESTRIP_ADJ 0x0A |
#define _3DPRIM_TRILIST_ADJ 0x0B |
#define _3DPRIM_TRISTRIP_ADJ 0x0C |
#define _3DPRIM_TRISTRIP_REVERSE 0x0D |
#define _3DPRIM_POLYGON 0x0E |
#define _3DPRIM_RECTLIST 0x0F |
#define _3DPRIM_LINELOOP 0x10 |
#define _3DPRIM_POINTLIST_BF 0x11 |
#define _3DPRIM_LINESTRIP_CONT 0x12 |
#define _3DPRIM_LINESTRIP_BF 0x13 |
#define _3DPRIM_LINESTRIP_CONT_BF 0x14 |
#define _3DPRIM_TRIFAN_NOSTIPPLE 0x15 |
#define _3DPRIM_VERTEXBUFFER_ACCESS_SEQUENTIAL 0 |
#define _3DPRIM_VERTEXBUFFER_ACCESS_RANDOM 1 |
#define GEN6_ANISORATIO_2 0 |
#define GEN6_ANISORATIO_4 1 |
#define GEN6_ANISORATIO_6 2 |
#define GEN6_ANISORATIO_8 3 |
#define GEN6_ANISORATIO_10 4 |
#define GEN6_ANISORATIO_12 5 |
#define GEN6_ANISORATIO_14 6 |
#define GEN6_ANISORATIO_16 7 |
#define GEN6_BLENDFACTOR_ONE 0x1 |
#define GEN6_BLENDFACTOR_SRC_COLOR 0x2 |
#define GEN6_BLENDFACTOR_SRC_ALPHA 0x3 |
#define GEN6_BLENDFACTOR_DST_ALPHA 0x4 |
#define GEN6_BLENDFACTOR_DST_COLOR 0x5 |
#define GEN6_BLENDFACTOR_SRC_ALPHA_SATURATE 0x6 |
#define GEN6_BLENDFACTOR_CONST_COLOR 0x7 |
#define GEN6_BLENDFACTOR_CONST_ALPHA 0x8 |
#define GEN6_BLENDFACTOR_SRC1_COLOR 0x9 |
#define GEN6_BLENDFACTOR_SRC1_ALPHA 0x0A |
#define GEN6_BLENDFACTOR_ZERO 0x11 |
#define GEN6_BLENDFACTOR_INV_SRC_COLOR 0x12 |
#define GEN6_BLENDFACTOR_INV_SRC_ALPHA 0x13 |
#define GEN6_BLENDFACTOR_INV_DST_ALPHA 0x14 |
#define GEN6_BLENDFACTOR_INV_DST_COLOR 0x15 |
#define GEN6_BLENDFACTOR_INV_CONST_COLOR 0x17 |
#define GEN6_BLENDFACTOR_INV_CONST_ALPHA 0x18 |
#define GEN6_BLENDFACTOR_INV_SRC1_COLOR 0x19 |
#define GEN6_BLENDFACTOR_INV_SRC1_ALPHA 0x1A |
#define GEN6_BLENDFUNCTION_ADD 0 |
#define GEN6_BLENDFUNCTION_SUBTRACT 1 |
#define GEN6_BLENDFUNCTION_REVERSE_SUBTRACT 2 |
#define GEN6_BLENDFUNCTION_MIN 3 |
#define GEN6_BLENDFUNCTION_MAX 4 |
#define GEN6_ALPHATEST_FORMAT_UNORM8 0 |
#define GEN6_ALPHATEST_FORMAT_FLOAT32 1 |
#define GEN6_CHROMAKEY_KILL_ON_ANY_MATCH 0 |
#define GEN6_CHROMAKEY_REPLACE_BLACK 1 |
#define GEN6_CLIP_API_OGL 0 |
#define GEN6_CLIP_API_DX 1 |
#define GEN6_CLIPMODE_NORMAL 0 |
#define GEN6_CLIPMODE_CLIP_ALL 1 |
#define GEN6_CLIPMODE_CLIP_NON_REJECTED 2 |
#define GEN6_CLIPMODE_REJECT_ALL 3 |
#define GEN6_CLIPMODE_ACCEPT_ALL 4 |
#define GEN6_CLIP_NDCSPACE 0 |
#define GEN6_CLIP_SCREENSPACE 1 |
#define GEN6_COMPAREFUNCTION_ALWAYS 0 |
#define GEN6_COMPAREFUNCTION_NEVER 1 |
#define GEN6_COMPAREFUNCTION_LESS 2 |
#define GEN6_COMPAREFUNCTION_EQUAL 3 |
#define GEN6_COMPAREFUNCTION_LEQUAL 4 |
#define GEN6_COMPAREFUNCTION_GREATER 5 |
#define GEN6_COMPAREFUNCTION_NOTEQUAL 6 |
#define GEN6_COMPAREFUNCTION_GEQUAL 7 |
#define GEN6_COVERAGE_PIXELS_HALF 0 |
#define GEN6_COVERAGE_PIXELS_1 1 |
#define GEN6_COVERAGE_PIXELS_2 2 |
#define GEN6_COVERAGE_PIXELS_4 3 |
#define GEN6_CULLMODE_BOTH 0 |
#define GEN6_CULLMODE_NONE 1 |
#define GEN6_CULLMODE_FRONT 2 |
#define GEN6_CULLMODE_BACK 3 |
#define GEN6_DEFAULTCOLOR_R8G8B8A8_UNORM 0 |
#define GEN6_DEFAULTCOLOR_R32G32B32A32_FLOAT 1 |
#define GEN6_DEPTHFORMAT_D32_FLOAT_S8X24_UINT 0 |
#define GEN6_DEPTHFORMAT_D32_FLOAT 1 |
#define GEN6_DEPTHFORMAT_D24_UNORM_S8_UINT 2 |
#define GEN6_DEPTHFORMAT_D16_UNORM 5 |
#define GEN6_FLOATING_POINT_IEEE_754 0 |
#define GEN6_FLOATING_POINT_NON_IEEE_754 1 |
#define GEN6_FRONTWINDING_CW 0 |
#define GEN6_FRONTWINDING_CCW 1 |
#define GEN6_INDEX_BYTE 0 |
#define GEN6_INDEX_WORD 1 |
#define GEN6_INDEX_DWORD 2 |
#define GEN6_LOGICOPFUNCTION_CLEAR 0 |
#define GEN6_LOGICOPFUNCTION_NOR 1 |
#define GEN6_LOGICOPFUNCTION_AND_INVERTED 2 |
#define GEN6_LOGICOPFUNCTION_COPY_INVERTED 3 |
#define GEN6_LOGICOPFUNCTION_AND_REVERSE 4 |
#define GEN6_LOGICOPFUNCTION_INVERT 5 |
#define GEN6_LOGICOPFUNCTION_XOR 6 |
#define GEN6_LOGICOPFUNCTION_NAND 7 |
#define GEN6_LOGICOPFUNCTION_AND 8 |
#define GEN6_LOGICOPFUNCTION_EQUIV 9 |
#define GEN6_LOGICOPFUNCTION_NOOP 10 |
#define GEN6_LOGICOPFUNCTION_OR_INVERTED 11 |
#define GEN6_LOGICOPFUNCTION_COPY 12 |
#define GEN6_LOGICOPFUNCTION_OR_REVERSE 13 |
#define GEN6_LOGICOPFUNCTION_OR 14 |
#define GEN6_LOGICOPFUNCTION_SET 15 |
#define GEN6_MAPFILTER_NEAREST 0x0 |
#define GEN6_MAPFILTER_LINEAR 0x1 |
#define GEN6_MAPFILTER_ANISOTROPIC 0x2 |
#define GEN6_MIPFILTER_NONE 0 |
#define GEN6_MIPFILTER_NEAREST 1 |
#define GEN6_MIPFILTER_LINEAR 3 |
#define GEN6_POLYGON_FRONT_FACING 0 |
#define GEN6_POLYGON_BACK_FACING 1 |
#define GEN6_PREFILTER_ALWAYS 0x0 |
#define GEN6_PREFILTER_NEVER 0x1 |
#define GEN6_PREFILTER_LESS 0x2 |
#define GEN6_PREFILTER_EQUAL 0x3 |
#define GEN6_PREFILTER_LEQUAL 0x4 |
#define GEN6_PREFILTER_GREATER 0x5 |
#define GEN6_PREFILTER_NOTEQUAL 0x6 |
#define GEN6_PREFILTER_GEQUAL 0x7 |
#define GEN6_PROVOKING_VERTEX_0 0 |
#define GEN6_PROVOKING_VERTEX_1 1 |
#define GEN6_PROVOKING_VERTEX_2 2 |
#define GEN6_RASTRULE_UPPER_LEFT 0 |
#define GEN6_RASTRULE_UPPER_RIGHT 1 |
#define GEN6_RENDERTARGET_CLAMPRANGE_UNORM 0 |
#define GEN6_RENDERTARGET_CLAMPRANGE_SNORM 1 |
#define GEN6_RENDERTARGET_CLAMPRANGE_FORMAT 2 |
#define GEN6_STENCILOP_KEEP 0 |
#define GEN6_STENCILOP_ZERO 1 |
#define GEN6_STENCILOP_REPLACE 2 |
#define GEN6_STENCILOP_INCRSAT 3 |
#define GEN6_STENCILOP_DECRSAT 4 |
#define GEN6_STENCILOP_INCR 5 |
#define GEN6_STENCILOP_DECR 6 |
#define GEN6_STENCILOP_INVERT 7 |
#define GEN6_SURFACE_MIPMAPLAYOUT_BELOW 0 |
#define GEN6_SURFACE_MIPMAPLAYOUT_RIGHT 1 |
#define GEN6_SURFACEFORMAT_R32G32B32A32_FLOAT 0x000 |
#define GEN6_SURFACEFORMAT_R32G32B32A32_SINT 0x001 |
#define GEN6_SURFACEFORMAT_R32G32B32A32_UINT 0x002 |
#define GEN6_SURFACEFORMAT_R32G32B32A32_UNORM 0x003 |
#define GEN6_SURFACEFORMAT_R32G32B32A32_SNORM 0x004 |
#define GEN6_SURFACEFORMAT_R64G64_FLOAT 0x005 |
#define GEN6_SURFACEFORMAT_R32G32B32X32_FLOAT 0x006 |
#define GEN6_SURFACEFORMAT_R32G32B32A32_SSCALED 0x007 |
#define GEN6_SURFACEFORMAT_R32G32B32A32_USCALED 0x008 |
#define GEN6_SURFACEFORMAT_R32G32B32_FLOAT 0x040 |
#define GEN6_SURFACEFORMAT_R32G32B32_SINT 0x041 |
#define GEN6_SURFACEFORMAT_R32G32B32_UINT 0x042 |
#define GEN6_SURFACEFORMAT_R32G32B32_UNORM 0x043 |
#define GEN6_SURFACEFORMAT_R32G32B32_SNORM 0x044 |
#define GEN6_SURFACEFORMAT_R32G32B32_SSCALED 0x045 |
#define GEN6_SURFACEFORMAT_R32G32B32_USCALED 0x046 |
#define GEN6_SURFACEFORMAT_R16G16B16A16_UNORM 0x080 |
#define GEN6_SURFACEFORMAT_R16G16B16A16_SNORM 0x081 |
#define GEN6_SURFACEFORMAT_R16G16B16A16_SINT 0x082 |
#define GEN6_SURFACEFORMAT_R16G16B16A16_UINT 0x083 |
#define GEN6_SURFACEFORMAT_R16G16B16A16_FLOAT 0x084 |
#define GEN6_SURFACEFORMAT_R32G32_FLOAT 0x085 |
#define GEN6_SURFACEFORMAT_R32G32_SINT 0x086 |
#define GEN6_SURFACEFORMAT_R32G32_UINT 0x087 |
#define GEN6_SURFACEFORMAT_R32_FLOAT_X8X24_TYPELESS 0x088 |
#define GEN6_SURFACEFORMAT_X32_TYPELESS_G8X24_UINT 0x089 |
#define GEN6_SURFACEFORMAT_L32A32_FLOAT 0x08A |
#define GEN6_SURFACEFORMAT_R32G32_UNORM 0x08B |
#define GEN6_SURFACEFORMAT_R32G32_SNORM 0x08C |
#define GEN6_SURFACEFORMAT_R64_FLOAT 0x08D |
#define GEN6_SURFACEFORMAT_R16G16B16X16_UNORM 0x08E |
#define GEN6_SURFACEFORMAT_R16G16B16X16_FLOAT 0x08F |
#define GEN6_SURFACEFORMAT_A32X32_FLOAT 0x090 |
#define GEN6_SURFACEFORMAT_L32X32_FLOAT 0x091 |
#define GEN6_SURFACEFORMAT_I32X32_FLOAT 0x092 |
#define GEN6_SURFACEFORMAT_R16G16B16A16_SSCALED 0x093 |
#define GEN6_SURFACEFORMAT_R16G16B16A16_USCALED 0x094 |
#define GEN6_SURFACEFORMAT_R32G32_SSCALED 0x095 |
#define GEN6_SURFACEFORMAT_R32G32_USCALED 0x096 |
#define GEN6_SURFACEFORMAT_B8G8R8A8_UNORM 0x0C0 |
#define GEN6_SURFACEFORMAT_B8G8R8A8_UNORM_SRGB 0x0C1 |
#define GEN6_SURFACEFORMAT_R10G10B10A2_UNORM 0x0C2 |
#define GEN6_SURFACEFORMAT_R10G10B10A2_UNORM_SRGB 0x0C3 |
#define GEN6_SURFACEFORMAT_R10G10B10A2_UINT 0x0C4 |
#define GEN6_SURFACEFORMAT_R10G10B10_SNORM_A2_UNORM 0x0C5 |
#define GEN6_SURFACEFORMAT_R8G8B8A8_UNORM 0x0C7 |
#define GEN6_SURFACEFORMAT_R8G8B8A8_UNORM_SRGB 0x0C8 |
#define GEN6_SURFACEFORMAT_R8G8B8A8_SNORM 0x0C9 |
#define GEN6_SURFACEFORMAT_R8G8B8A8_SINT 0x0CA |
#define GEN6_SURFACEFORMAT_R8G8B8A8_UINT 0x0CB |
#define GEN6_SURFACEFORMAT_R16G16_UNORM 0x0CC |
#define GEN6_SURFACEFORMAT_R16G16_SNORM 0x0CD |
#define GEN6_SURFACEFORMAT_R16G16_SINT 0x0CE |
#define GEN6_SURFACEFORMAT_R16G16_UINT 0x0CF |
#define GEN6_SURFACEFORMAT_R16G16_FLOAT 0x0D0 |
#define GEN6_SURFACEFORMAT_B10G10R10A2_UNORM 0x0D1 |
#define GEN6_SURFACEFORMAT_B10G10R10A2_UNORM_SRGB 0x0D2 |
#define GEN6_SURFACEFORMAT_R11G11B10_FLOAT 0x0D3 |
#define GEN6_SURFACEFORMAT_R32_SINT 0x0D6 |
#define GEN6_SURFACEFORMAT_R32_UINT 0x0D7 |
#define GEN6_SURFACEFORMAT_R32_FLOAT 0x0D8 |
#define GEN6_SURFACEFORMAT_R24_UNORM_X8_TYPELESS 0x0D9 |
#define GEN6_SURFACEFORMAT_X24_TYPELESS_G8_UINT 0x0DA |
#define GEN6_SURFACEFORMAT_L16A16_UNORM 0x0DF |
#define GEN6_SURFACEFORMAT_I24X8_UNORM 0x0E0 |
#define GEN6_SURFACEFORMAT_L24X8_UNORM 0x0E1 |
#define GEN6_SURFACEFORMAT_A24X8_UNORM 0x0E2 |
#define GEN6_SURFACEFORMAT_I32_FLOAT 0x0E3 |
#define GEN6_SURFACEFORMAT_L32_FLOAT 0x0E4 |
#define GEN6_SURFACEFORMAT_A32_FLOAT 0x0E5 |
#define GEN6_SURFACEFORMAT_B8G8R8X8_UNORM 0x0E9 |
#define GEN6_SURFACEFORMAT_B8G8R8X8_UNORM_SRGB 0x0EA |
#define GEN6_SURFACEFORMAT_R8G8B8X8_UNORM 0x0EB |
#define GEN6_SURFACEFORMAT_R8G8B8X8_UNORM_SRGB 0x0EC |
#define GEN6_SURFACEFORMAT_R9G9B9E5_SHAREDEXP 0x0ED |
#define GEN6_SURFACEFORMAT_B10G10R10X2_UNORM 0x0EE |
#define GEN6_SURFACEFORMAT_L16A16_FLOAT 0x0F0 |
#define GEN6_SURFACEFORMAT_R32_UNORM 0x0F1 |
#define GEN6_SURFACEFORMAT_R32_SNORM 0x0F2 |
#define GEN6_SURFACEFORMAT_R10G10B10X2_USCALED 0x0F3 |
#define GEN6_SURFACEFORMAT_R8G8B8A8_SSCALED 0x0F4 |
#define GEN6_SURFACEFORMAT_R8G8B8A8_USCALED 0x0F5 |
#define GEN6_SURFACEFORMAT_R16G16_SSCALED 0x0F6 |
#define GEN6_SURFACEFORMAT_R16G16_USCALED 0x0F7 |
#define GEN6_SURFACEFORMAT_R32_SSCALED 0x0F8 |
#define GEN6_SURFACEFORMAT_R32_USCALED 0x0F9 |
#define GEN6_SURFACEFORMAT_B5G6R5_UNORM 0x100 |
#define GEN6_SURFACEFORMAT_B5G6R5_UNORM_SRGB 0x101 |
#define GEN6_SURFACEFORMAT_B5G5R5A1_UNORM 0x102 |
#define GEN6_SURFACEFORMAT_B5G5R5A1_UNORM_SRGB 0x103 |
#define GEN6_SURFACEFORMAT_B4G4R4A4_UNORM 0x104 |
#define GEN6_SURFACEFORMAT_B4G4R4A4_UNORM_SRGB 0x105 |
#define GEN6_SURFACEFORMAT_R8G8_UNORM 0x106 |
#define GEN6_SURFACEFORMAT_R8G8_SNORM 0x107 |
#define GEN6_SURFACEFORMAT_R8G8_SINT 0x108 |
#define GEN6_SURFACEFORMAT_R8G8_UINT 0x109 |
#define GEN6_SURFACEFORMAT_R16_UNORM 0x10A |
#define GEN6_SURFACEFORMAT_R16_SNORM 0x10B |
#define GEN6_SURFACEFORMAT_R16_SINT 0x10C |
#define GEN6_SURFACEFORMAT_R16_UINT 0x10D |
#define GEN6_SURFACEFORMAT_R16_FLOAT 0x10E |
#define GEN6_SURFACEFORMAT_I16_UNORM 0x111 |
#define GEN6_SURFACEFORMAT_L16_UNORM 0x112 |
#define GEN6_SURFACEFORMAT_A16_UNORM 0x113 |
#define GEN6_SURFACEFORMAT_L8A8_UNORM 0x114 |
#define GEN6_SURFACEFORMAT_I16_FLOAT 0x115 |
#define GEN6_SURFACEFORMAT_L16_FLOAT 0x116 |
#define GEN6_SURFACEFORMAT_A16_FLOAT 0x117 |
#define GEN6_SURFACEFORMAT_R5G5_SNORM_B6_UNORM 0x119 |
#define GEN6_SURFACEFORMAT_B5G5R5X1_UNORM 0x11A |
#define GEN6_SURFACEFORMAT_B5G5R5X1_UNORM_SRGB 0x11B |
#define GEN6_SURFACEFORMAT_R8G8_SSCALED 0x11C |
#define GEN6_SURFACEFORMAT_R8G8_USCALED 0x11D |
#define GEN6_SURFACEFORMAT_R16_SSCALED 0x11E |
#define GEN6_SURFACEFORMAT_R16_USCALED 0x11F |
#define GEN6_SURFACEFORMAT_R8_UNORM 0x140 |
#define GEN6_SURFACEFORMAT_R8_SNORM 0x141 |
#define GEN6_SURFACEFORMAT_R8_SINT 0x142 |
#define GEN6_SURFACEFORMAT_R8_UINT 0x143 |
#define GEN6_SURFACEFORMAT_A8_UNORM 0x144 |
#define GEN6_SURFACEFORMAT_I8_UNORM 0x145 |
#define GEN6_SURFACEFORMAT_L8_UNORM 0x146 |
#define GEN6_SURFACEFORMAT_P4A4_UNORM 0x147 |
#define GEN6_SURFACEFORMAT_A4P4_UNORM 0x148 |
#define GEN6_SURFACEFORMAT_R8_SSCALED 0x149 |
#define GEN6_SURFACEFORMAT_R8_USCALED 0x14A |
#define GEN6_SURFACEFORMAT_R1_UINT 0x181 |
#define GEN6_SURFACEFORMAT_YCRCB_NORMAL 0x182 |
#define GEN6_SURFACEFORMAT_YCRCB_SWAPUVY 0x183 |
#define GEN6_SURFACEFORMAT_BC1_UNORM 0x186 |
#define GEN6_SURFACEFORMAT_BC2_UNORM 0x187 |
#define GEN6_SURFACEFORMAT_BC3_UNORM 0x188 |
#define GEN6_SURFACEFORMAT_BC4_UNORM 0x189 |
#define GEN6_SURFACEFORMAT_BC5_UNORM 0x18A |
#define GEN6_SURFACEFORMAT_BC1_UNORM_SRGB 0x18B |
#define GEN6_SURFACEFORMAT_BC2_UNORM_SRGB 0x18C |
#define GEN6_SURFACEFORMAT_BC3_UNORM_SRGB 0x18D |
#define GEN6_SURFACEFORMAT_MONO8 0x18E |
#define GEN6_SURFACEFORMAT_YCRCB_SWAPUV 0x18F |
#define GEN6_SURFACEFORMAT_YCRCB_SWAPY 0x190 |
#define GEN6_SURFACEFORMAT_DXT1_RGB 0x191 |
#define GEN6_SURFACEFORMAT_FXT1 0x192 |
#define GEN6_SURFACEFORMAT_R8G8B8_UNORM 0x193 |
#define GEN6_SURFACEFORMAT_R8G8B8_SNORM 0x194 |
#define GEN6_SURFACEFORMAT_R8G8B8_SSCALED 0x195 |
#define GEN6_SURFACEFORMAT_R8G8B8_USCALED 0x196 |
#define GEN6_SURFACEFORMAT_R64G64B64A64_FLOAT 0x197 |
#define GEN6_SURFACEFORMAT_R64G64B64_FLOAT 0x198 |
#define GEN6_SURFACEFORMAT_BC4_SNORM 0x199 |
#define GEN6_SURFACEFORMAT_BC5_SNORM 0x19A |
#define GEN6_SURFACEFORMAT_R16G16B16_UNORM 0x19C |
#define GEN6_SURFACEFORMAT_R16G16B16_SNORM 0x19D |
#define GEN6_SURFACEFORMAT_R16G16B16_SSCALED 0x19E |
#define GEN6_SURFACEFORMAT_R16G16B16_USCALED 0x19F |
#define GEN6_SURFACERETURNFORMAT_FLOAT32 0 |
#define GEN6_SURFACERETURNFORMAT_S1 1 |
#define GEN6_SURFACE_1D 0 |
#define GEN6_SURFACE_2D 1 |
#define GEN6_SURFACE_3D 2 |
#define GEN6_SURFACE_CUBE 3 |
#define GEN6_SURFACE_BUFFER 4 |
#define GEN6_SURFACE_NULL 7 |
#define GEN6_BORDER_COLOR_MODE_DEFAULT 0 |
#define GEN6_BORDER_COLOR_MODE_LEGACY 1 |
#define GEN6_TEXCOORDMODE_WRAP 0 |
#define GEN6_TEXCOORDMODE_MIRROR 1 |
#define GEN6_TEXCOORDMODE_CLAMP 2 |
#define GEN6_TEXCOORDMODE_CUBE 3 |
#define GEN6_TEXCOORDMODE_CLAMP_BORDER 4 |
#define GEN6_TEXCOORDMODE_MIRROR_ONCE 5 |
#define GEN6_THREAD_PRIORITY_NORMAL 0 |
#define GEN6_THREAD_PRIORITY_HIGH 1 |
#define GEN6_TILEWALK_XMAJOR 0 |
#define GEN6_TILEWALK_YMAJOR 1 |
#define GEN6_VERTEX_SUBPIXEL_PRECISION_8BITS 0 |
#define GEN6_VERTEX_SUBPIXEL_PRECISION_4BITS 1 |
#define GEN6_VERTEXBUFFER_ACCESS_VERTEXDATA 0 |
#define GEN6_VERTEXBUFFER_ACCESS_INSTANCEDATA 1 |
#define GEN6_VFCOMPONENT_NOSTORE 0 |
#define GEN6_VFCOMPONENT_STORE_SRC 1 |
#define GEN6_VFCOMPONENT_STORE_0 2 |
#define GEN6_VFCOMPONENT_STORE_1_FLT 3 |
#define GEN6_VFCOMPONENT_STORE_1_INT 4 |
#define GEN6_VFCOMPONENT_STORE_VID 5 |
#define GEN6_VFCOMPONENT_STORE_IID 6 |
#define GEN6_VFCOMPONENT_STORE_PID 7 |
/* Execution Unit (EU) defines |
*/ |
#define GEN6_ALIGN_1 0 |
#define GEN6_ALIGN_16 1 |
#define GEN6_ADDRESS_DIRECT 0 |
#define GEN6_ADDRESS_REGISTER_INDIRECT_REGISTER 1 |
#define GEN6_CHANNEL_X 0 |
#define GEN6_CHANNEL_Y 1 |
#define GEN6_CHANNEL_Z 2 |
#define GEN6_CHANNEL_W 3 |
#define GEN6_COMPRESSION_NONE 0 |
#define GEN6_COMPRESSION_2NDHALF 1 |
#define GEN6_COMPRESSION_COMPRESSED 2 |
#define GEN6_CONDITIONAL_NONE 0 |
#define GEN6_CONDITIONAL_Z 1 |
#define GEN6_CONDITIONAL_NZ 2 |
#define GEN6_CONDITIONAL_EQ 1 /* Z */ |
#define GEN6_CONDITIONAL_NEQ 2 /* NZ */ |
#define GEN6_CONDITIONAL_G 3 |
#define GEN6_CONDITIONAL_GE 4 |
#define GEN6_CONDITIONAL_L 5 |
#define GEN6_CONDITIONAL_LE 6 |
#define GEN6_CONDITIONAL_C 7 |
#define GEN6_CONDITIONAL_O 8 |
#define GEN6_DEBUG_NONE 0 |
#define GEN6_DEBUG_BREAKPOINT 1 |
#define GEN6_DEPENDENCY_NORMAL 0 |
#define GEN6_DEPENDENCY_NOTCLEARED 1 |
#define GEN6_DEPENDENCY_NOTCHECKED 2 |
#define GEN6_DEPENDENCY_DISABLE 3 |
#define GEN6_EXECUTE_1 0 |
#define GEN6_EXECUTE_2 1 |
#define GEN6_EXECUTE_4 2 |
#define GEN6_EXECUTE_8 3 |
#define GEN6_EXECUTE_16 4 |
#define GEN6_EXECUTE_32 5 |
#define GEN6_HORIZONTAL_STRIDE_0 0 |
#define GEN6_HORIZONTAL_STRIDE_1 1 |
#define GEN6_HORIZONTAL_STRIDE_2 2 |
#define GEN6_HORIZONTAL_STRIDE_4 3 |
#define GEN6_INSTRUCTION_NORMAL 0 |
#define GEN6_INSTRUCTION_SATURATE 1 |
#define GEN6_MASK_ENABLE 0 |
#define GEN6_MASK_DISABLE 1 |
#define GEN6_OPCODE_MOV 1 |
#define GEN6_OPCODE_SEL 2 |
#define GEN6_OPCODE_NOT 4 |
#define GEN6_OPCODE_AND 5 |
#define GEN6_OPCODE_OR 6 |
#define GEN6_OPCODE_XOR 7 |
#define GEN6_OPCODE_SHR 8 |
#define GEN6_OPCODE_SHL 9 |
#define GEN6_OPCODE_RSR 10 |
#define GEN6_OPCODE_RSL 11 |
#define GEN6_OPCODE_ASR 12 |
#define GEN6_OPCODE_CMP 16 |
#define GEN6_OPCODE_JMPI 32 |
#define GEN6_OPCODE_IF 34 |
#define GEN6_OPCODE_IFF 35 |
#define GEN6_OPCODE_ELSE 36 |
#define GEN6_OPCODE_ENDIF 37 |
#define GEN6_OPCODE_DO 38 |
#define GEN6_OPCODE_WHILE 39 |
#define GEN6_OPCODE_BREAK 40 |
#define GEN6_OPCODE_CONTINUE 41 |
#define GEN6_OPCODE_HALT 42 |
#define GEN6_OPCODE_MSAVE 44 |
#define GEN6_OPCODE_MRESTORE 45 |
#define GEN6_OPCODE_PUSH 46 |
#define GEN6_OPCODE_POP 47 |
#define GEN6_OPCODE_WAIT 48 |
#define GEN6_OPCODE_SEND 49 |
#define GEN6_OPCODE_ADD 64 |
#define GEN6_OPCODE_MUL 65 |
#define GEN6_OPCODE_AVG 66 |
#define GEN6_OPCODE_FRC 67 |
#define GEN6_OPCODE_RNDU 68 |
#define GEN6_OPCODE_RNDD 69 |
#define GEN6_OPCODE_RNDE 70 |
#define GEN6_OPCODE_RNDZ 71 |
#define GEN6_OPCODE_MAC 72 |
#define GEN6_OPCODE_MACH 73 |
#define GEN6_OPCODE_LZD 74 |
#define GEN6_OPCODE_SAD2 80 |
#define GEN6_OPCODE_SADA2 81 |
#define GEN6_OPCODE_DP4 84 |
#define GEN6_OPCODE_DPH 85 |
#define GEN6_OPCODE_DP3 86 |
#define GEN6_OPCODE_DP2 87 |
#define GEN6_OPCODE_DPA2 88 |
#define GEN6_OPCODE_LINE 89 |
#define GEN6_OPCODE_NOP 126 |
#define GEN6_PREDICATE_NONE 0 |
#define GEN6_PREDICATE_NORMAL 1 |
#define GEN6_PREDICATE_ALIGN1_ANYV 2 |
#define GEN6_PREDICATE_ALIGN1_ALLV 3 |
#define GEN6_PREDICATE_ALIGN1_ANY2H 4 |
#define GEN6_PREDICATE_ALIGN1_ALL2H 5 |
#define GEN6_PREDICATE_ALIGN1_ANY4H 6 |
#define GEN6_PREDICATE_ALIGN1_ALL4H 7 |
#define GEN6_PREDICATE_ALIGN1_ANY8H 8 |
#define GEN6_PREDICATE_ALIGN1_ALL8H 9 |
#define GEN6_PREDICATE_ALIGN1_ANY16H 10 |
#define GEN6_PREDICATE_ALIGN1_ALL16H 11 |
#define GEN6_PREDICATE_ALIGN16_REPLICATE_X 2 |
#define GEN6_PREDICATE_ALIGN16_REPLICATE_Y 3 |
#define GEN6_PREDICATE_ALIGN16_REPLICATE_Z 4 |
#define GEN6_PREDICATE_ALIGN16_REPLICATE_W 5 |
#define GEN6_PREDICATE_ALIGN16_ANY4H 6 |
#define GEN6_PREDICATE_ALIGN16_ALL4H 7 |
#define GEN6_ARCHITECTURE_REGISTER_FILE 0 |
#define GEN6_GENERAL_REGISTER_FILE 1 |
#define GEN6_MESSAGE_REGISTER_FILE 2 |
#define GEN6_IMMEDIATE_VALUE 3 |
#define GEN6_REGISTER_TYPE_UD 0 |
#define GEN6_REGISTER_TYPE_D 1 |
#define GEN6_REGISTER_TYPE_UW 2 |
#define GEN6_REGISTER_TYPE_W 3 |
#define GEN6_REGISTER_TYPE_UB 4 |
#define GEN6_REGISTER_TYPE_B 5 |
#define GEN6_REGISTER_TYPE_VF 5 /* packed float vector, immediates only? */ |
#define GEN6_REGISTER_TYPE_HF 6 |
#define GEN6_REGISTER_TYPE_V 6 /* packed int vector, immediates only, uword dest only */ |
#define GEN6_REGISTER_TYPE_F 7 |
#define GEN6_ARF_NULL 0x00 |
#define GEN6_ARF_ADDRESS 0x10 |
#define GEN6_ARF_ACCUMULATOR 0x20 |
#define GEN6_ARF_FLAG 0x30 |
#define GEN6_ARF_MASK 0x40 |
#define GEN6_ARF_MASK_STACK 0x50 |
#define GEN6_ARF_MASK_STACK_DEPTH 0x60 |
#define GEN6_ARF_STATE 0x70 |
#define GEN6_ARF_CONTROL 0x80 |
#define GEN6_ARF_NOTIFICATION_COUNT 0x90 |
#define GEN6_ARF_IP 0xA0 |
#define GEN6_AMASK 0 |
#define GEN6_IMASK 1 |
#define GEN6_LMASK 2 |
#define GEN6_CMASK 3 |
#define GEN6_THREAD_NORMAL 0 |
#define GEN6_THREAD_ATOMIC 1 |
#define GEN6_THREAD_SWITCH 2 |
#define GEN6_VERTICAL_STRIDE_0 0 |
#define GEN6_VERTICAL_STRIDE_1 1 |
#define GEN6_VERTICAL_STRIDE_2 2 |
#define GEN6_VERTICAL_STRIDE_4 3 |
#define GEN6_VERTICAL_STRIDE_8 4 |
#define GEN6_VERTICAL_STRIDE_16 5 |
#define GEN6_VERTICAL_STRIDE_32 6 |
#define GEN6_VERTICAL_STRIDE_64 7 |
#define GEN6_VERTICAL_STRIDE_128 8 |
#define GEN6_VERTICAL_STRIDE_256 9 |
#define GEN6_VERTICAL_STRIDE_ONE_DIMENSIONAL 0xF |
#define GEN6_WIDTH_1 0 |
#define GEN6_WIDTH_2 1 |
#define GEN6_WIDTH_4 2 |
#define GEN6_WIDTH_8 3 |
#define GEN6_WIDTH_16 4 |
#define GEN6_STATELESS_BUFFER_BOUNDARY_1K 0 |
#define GEN6_STATELESS_BUFFER_BOUNDARY_2K 1 |
#define GEN6_STATELESS_BUFFER_BOUNDARY_4K 2 |
#define GEN6_STATELESS_BUFFER_BOUNDARY_8K 3 |
#define GEN6_STATELESS_BUFFER_BOUNDARY_16K 4 |
#define GEN6_STATELESS_BUFFER_BOUNDARY_32K 5 |
#define GEN6_STATELESS_BUFFER_BOUNDARY_64K 6 |
#define GEN6_STATELESS_BUFFER_BOUNDARY_128K 7 |
#define GEN6_STATELESS_BUFFER_BOUNDARY_256K 8 |
#define GEN6_STATELESS_BUFFER_BOUNDARY_512K 9 |
#define GEN6_STATELESS_BUFFER_BOUNDARY_1M 10 |
#define GEN6_STATELESS_BUFFER_BOUNDARY_2M 11 |
#define GEN6_POLYGON_FACING_FRONT 0 |
#define GEN6_POLYGON_FACING_BACK 1 |
#define GEN6_MESSAGE_TARGET_NULL 0 |
#define GEN6_MESSAGE_TARGET_MATH 1 |
#define GEN6_MESSAGE_TARGET_SAMPLER 2 |
#define GEN6_MESSAGE_TARGET_GATEWAY 3 |
#define GEN6_MESSAGE_TARGET_DATAPORT_READ 4 |
#define GEN6_MESSAGE_TARGET_DATAPORT_WRITE 5 |
#define GEN6_MESSAGE_TARGET_URB 6 |
#define GEN6_MESSAGE_TARGET_THREAD_SPAWNER 7 |
#define GEN6_SAMPLER_RETURN_FORMAT_FLOAT32 0 |
#define GEN6_SAMPLER_RETURN_FORMAT_UINT32 2 |
#define GEN6_SAMPLER_RETURN_FORMAT_SINT32 3 |
#define GEN6_SAMPLER_MESSAGE_SIMD8_SAMPLE 0 |
#define GEN6_SAMPLER_MESSAGE_SIMD16_SAMPLE 0 |
#define GEN6_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS 0 |
#define GEN6_SAMPLER_MESSAGE_SIMD8_KILLPIX 1 |
#define GEN6_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_LOD 1 |
#define GEN6_SAMPLER_MESSAGE_SIMD16_SAMPLE_LOD 1 |
#define GEN6_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_GRADIENTS 2 |
#define GEN6_SAMPLER_MESSAGE_SIMD8_SAMPLE_GRADIENTS 2 |
#define GEN6_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_COMPARE 0 |
#define GEN6_SAMPLER_MESSAGE_SIMD16_SAMPLE_COMPARE 2 |
#define GEN6_SAMPLER_MESSAGE_SIMD4X2_RESINFO 2 |
#define GEN6_SAMPLER_MESSAGE_SIMD8_RESINFO 2 |
#define GEN6_SAMPLER_MESSAGE_SIMD16_RESINFO 2 |
#define GEN6_SAMPLER_MESSAGE_SIMD4X2_LD 3 |
#define GEN6_SAMPLER_MESSAGE_SIMD8_LD 3 |
#define GEN6_SAMPLER_MESSAGE_SIMD16_LD 3 |
#define GEN6_DATAPORT_OWORD_BLOCK_1_OWORDLOW 0 |
#define GEN6_DATAPORT_OWORD_BLOCK_1_OWORDHIGH 1 |
#define GEN6_DATAPORT_OWORD_BLOCK_2_OWORDS 2 |
#define GEN6_DATAPORT_OWORD_BLOCK_4_OWORDS 3 |
#define GEN6_DATAPORT_OWORD_BLOCK_8_OWORDS 4 |
#define GEN6_DATAPORT_OWORD_DUAL_BLOCK_1OWORD 0 |
#define GEN6_DATAPORT_OWORD_DUAL_BLOCK_4OWORDS 2 |
#define GEN6_DATAPORT_DWORD_SCATTERED_BLOCK_8DWORDS 2 |
#define GEN6_DATAPORT_DWORD_SCATTERED_BLOCK_16DWORDS 3 |
#define GEN6_DATAPORT_READ_TARGET_DATA_CACHE 0 |
#define GEN6_DATAPORT_READ_TARGET_RENDER_CACHE 1 |
#define GEN6_DATAPORT_READ_TARGET_SAMPLER_CACHE 2 |
#define GEN6_DATAPORT_RENDER_TARGET_WRITE_SIMD16_SINGLE_SOURCE 0 |
#define GEN6_DATAPORT_RENDER_TARGET_WRITE_SIMD16_SINGLE_SOURCE_REPLICATED 1 |
#define GEN6_DATAPORT_RENDER_TARGET_WRITE_SIMD8_DUAL_SOURCE_SUBSPAN01 2 |
#define GEN6_DATAPORT_RENDER_TARGET_WRITE_SIMD8_DUAL_SOURCE_SUBSPAN23 3 |
#define GEN6_DATAPORT_RENDER_TARGET_WRITE_SIMD8_SINGLE_SOURCE_SUBSPAN01 4 |
#define GEN6_MATH_INTEGER_UNSIGNED 0 |
#define GEN6_MATH_INTEGER_SIGNED 1 |
#define GEN6_MATH_PRECISION_FULL 0 |
#define GEN6_MATH_PRECISION_PARTIAL 1 |
#define GEN6_MATH_SATURATE_NONE 0 |
#define GEN6_MATH_SATURATE_SATURATE 1 |
#define GEN6_MATH_DATA_VECTOR 0 |
#define GEN6_MATH_DATA_SCALAR 1 |
#define GEN6_URB_OPCODE_WRITE 0 |
#define GEN6_URB_SWIZZLE_NONE 0 |
#define GEN6_URB_SWIZZLE_INTERLEAVE 1 |
#define GEN6_URB_SWIZZLE_TRANSPOSE 2 |
#define GEN6_SCRATCH_SPACE_SIZE_1K 0 |
#define GEN6_SCRATCH_SPACE_SIZE_2K 1 |
#define GEN6_SCRATCH_SPACE_SIZE_4K 2 |
#define GEN6_SCRATCH_SPACE_SIZE_8K 3 |
#define GEN6_SCRATCH_SPACE_SIZE_16K 4 |
#define GEN6_SCRATCH_SPACE_SIZE_32K 5 |
#define GEN6_SCRATCH_SPACE_SIZE_64K 6 |
#define GEN6_SCRATCH_SPACE_SIZE_128K 7 |
#define GEN6_SCRATCH_SPACE_SIZE_256K 8 |
#define GEN6_SCRATCH_SPACE_SIZE_512K 9 |
#define GEN6_SCRATCH_SPACE_SIZE_1M 10 |
#define GEN6_SCRATCH_SPACE_SIZE_2M 11 |
/* The hardware supports two different modes for border color. The |
* default (OpenGL) mode uses floating-point color channels, while the |
* legacy mode uses 4 bytes. |
* |
* More significantly, the legacy mode respects the components of the |
* border color for channels not present in the source, (whereas the |
* default mode will ignore the border color's alpha channel and use |
* alpha==1 for an RGB source, for example). |
* |
* The legacy mode matches the semantics specified by the Render |
* extension. |
*/ |
struct gen6_sampler_default_border_color { |
float color[4]; |
}; |
struct gen6_sampler_legacy_border_color { |
uint8_t color[4]; |
}; |
struct gen6_sampler_state { |
struct { |
uint32_t shadow_function:3; |
uint32_t lod_bias:11; |
uint32_t min_filter:3; |
uint32_t mag_filter:3; |
uint32_t mip_filter:2; |
uint32_t base_level:5; |
uint32_t pad:1; |
uint32_t lod_preclamp:1; |
uint32_t border_color_mode:1; |
uint32_t pad0:1; |
uint32_t disable:1; |
} ss0; |
struct { |
uint32_t r_wrap_mode:3; |
uint32_t t_wrap_mode:3; |
uint32_t s_wrap_mode:3; |
uint32_t pad:3; |
uint32_t max_lod:10; |
uint32_t min_lod:10; |
} ss1; |
struct { |
uint32_t border_color; |
} ss2; |
struct { |
uint32_t non_normalized_coord:1; |
uint32_t pad:12; |
uint32_t address_round:6; |
uint32_t max_aniso:3; |
uint32_t chroma_key_mode:1; |
uint32_t chroma_key_index:2; |
uint32_t chroma_key_enable:1; |
uint32_t monochrome_filter_width:3; |
uint32_t monochrome_filter_height:3; |
} ss3; |
}; |
struct gen6_blend_state { |
struct { |
uint32_t dest_blend_factor:5; |
uint32_t source_blend_factor:5; |
uint32_t pad3:1; |
uint32_t blend_func:3; |
uint32_t pad2:1; |
uint32_t ia_dest_blend_factor:5; |
uint32_t ia_source_blend_factor:5; |
uint32_t pad1:1; |
uint32_t ia_blend_func:3; |
uint32_t pad0:1; |
uint32_t ia_blend_enable:1; |
uint32_t blend_enable:1; |
} blend0; |
struct { |
uint32_t post_blend_clamp_enable:1; |
uint32_t pre_blend_clamp_enable:1; |
uint32_t clamp_range:2; |
uint32_t pad0:4; |
uint32_t x_dither_offset:2; |
uint32_t y_dither_offset:2; |
uint32_t dither_enable:1; |
uint32_t alpha_test_func:3; |
uint32_t alpha_test_enable:1; |
uint32_t pad1:1; |
uint32_t logic_op_func:4; |
uint32_t logic_op_enable:1; |
uint32_t pad2:1; |
uint32_t write_disable_b:1; |
uint32_t write_disable_g:1; |
uint32_t write_disable_r:1; |
uint32_t write_disable_a:1; |
uint32_t pad3:1; |
uint32_t alpha_to_coverage_dither:1; |
uint32_t alpha_to_one:1; |
uint32_t alpha_to_coverage:1; |
} blend1; |
}; |
struct gen6_color_calc_state { |
struct { |
uint32_t alpha_test_format:1; |
uint32_t pad0:14; |
uint32_t round_disable:1; |
uint32_t bf_stencil_ref:8; |
uint32_t stencil_ref:8; |
} cc0; |
union { |
float alpha_ref_f; |
struct { |
uint32_t ui:8; |
uint32_t pad0:24; |
} alpha_ref_fi; |
} cc1; |
float constant_r; |
float constant_g; |
float constant_b; |
float constant_a; |
}; |
struct gen6_depth_stencil_state { |
struct { |
uint32_t pad0:3; |
uint32_t bf_stencil_pass_depth_pass_op:3; |
uint32_t bf_stencil_pass_depth_fail_op:3; |
uint32_t bf_stencil_fail_op:3; |
uint32_t bf_stencil_func:3; |
uint32_t bf_stencil_enable:1; |
uint32_t pad1:2; |
uint32_t stencil_write_enable:1; |
uint32_t stencil_pass_depth_pass_op:3; |
uint32_t stencil_pass_depth_fail_op:3; |
uint32_t stencil_fail_op:3; |
uint32_t stencil_func:3; |
uint32_t stencil_enable:1; |
} ds0; |
struct { |
uint32_t bf_stencil_write_mask:8; |
uint32_t bf_stencil_test_mask:8; |
uint32_t stencil_write_mask:8; |
uint32_t stencil_test_mask:8; |
} ds1; |
struct { |
uint32_t pad0:26; |
uint32_t depth_write_enable:1; |
uint32_t depth_test_func:3; |
uint32_t pad1:1; |
uint32_t depth_test_enable:1; |
} ds2; |
}; |
struct gen6_surface_state { |
struct { |
uint32_t cube_pos_z:1; |
uint32_t cube_neg_z:1; |
uint32_t cube_pos_y:1; |
uint32_t cube_neg_y:1; |
uint32_t cube_pos_x:1; |
uint32_t cube_neg_x:1; |
uint32_t pad:3; |
uint32_t render_cache_read_mode:1; |
uint32_t mipmap_layout_mode:1; |
uint32_t vert_line_stride_ofs:1; |
uint32_t vert_line_stride:1; |
uint32_t color_blend:1; |
uint32_t writedisable_blue:1; |
uint32_t writedisable_green:1; |
uint32_t writedisable_red:1; |
uint32_t writedisable_alpha:1; |
uint32_t surface_format:9; |
uint32_t data_return_format:1; |
uint32_t pad0:1; |
uint32_t surface_type:3; |
} ss0; |
struct { |
uint32_t base_addr; |
} ss1; |
struct { |
uint32_t render_target_rotation:2; |
uint32_t mip_count:4; |
uint32_t width:13; |
uint32_t height:13; |
} ss2; |
struct { |
uint32_t tile_walk:1; |
uint32_t tiled_surface:1; |
uint32_t pad:1; |
uint32_t pitch:18; |
uint32_t depth:11; |
} ss3; |
struct { |
uint32_t pad:19; |
uint32_t min_array_elt:9; |
uint32_t min_lod:4; |
} ss4; |
struct { |
uint32_t pad:20; |
uint32_t y_offset:4; |
uint32_t pad2:1; |
uint32_t x_offset:7; |
} ss5; |
}; |
/* Surface state DW0 */ |
#define GEN6_SURFACE_RC_READ_WRITE (1 << 8) |
#define GEN6_SURFACE_MIPLAYOUT_SHIFT 10 |
#define GEN6_SURFACE_MIPMAPLAYOUT_BELOW 0 |
#define GEN6_SURFACE_MIPMAPLAYOUT_RIGHT 1 |
#define GEN6_SURFACE_CUBEFACE_ENABLES 0x3f |
#define GEN6_SURFACE_BLEND_ENABLED (1 << 13) |
#define GEN6_SURFACE_WRITEDISABLE_B_SHIFT 14 |
#define GEN6_SURFACE_WRITEDISABLE_G_SHIFT 15 |
#define GEN6_SURFACE_WRITEDISABLE_R_SHIFT 16 |
#define GEN6_SURFACE_WRITEDISABLE_A_SHIFT 17 |
#define GEN6_SURFACE_FORMAT_SHIFT 18 |
#define GEN6_SURFACE_FORMAT_MASK INTEL_MASK(26, 18) |
#define GEN6_SURFACE_TYPE_SHIFT 29 |
#define GEN6_SURFACE_TYPE_MASK GEN6_MASK(31, 29) |
#define GEN6_SURFACE_1D 0 |
#define GEN6_SURFACE_2D 1 |
#define GEN6_SURFACE_3D 2 |
#define GEN6_SURFACE_CUBE 3 |
#define GEN6_SURFACE_BUFFER 4 |
#define GEN6_SURFACE_NULL 7 |
/* Surface state DW2 */ |
#define GEN6_SURFACE_HEIGHT_SHIFT 19 |
#define GEN6_SURFACE_HEIGHT_MASK GEN6_MASK(31, 19) |
#define GEN6_SURFACE_WIDTH_SHIFT 6 |
#define GEN6_SURFACE_WIDTH_MASK GEN6_MASK(18, 6) |
#define GEN6_SURFACE_LOD_SHIFT 2 |
#define GEN6_SURFACE_LOD_MASK GEN6_MASK(5, 2) |
/* Surface state DW3 */ |
#define GEN6_SURFACE_DEPTH_SHIFT 21 |
#define GEN6_SURFACE_DEPTH_MASK GEN6_MASK(31, 21) |
#define GEN6_SURFACE_PITCH_SHIFT 3 |
#define GEN6_SURFACE_PITCH_MASK GEN6_MASK(19, 3) |
#define GEN6_SURFACE_TILED (1 << 1) |
#define GEN6_SURFACE_TILED_Y (1 << 0) |
/* Surface state DW4 */ |
#define GEN6_SURFACE_MIN_LOD_SHIFT 28 |
#define GEN6_SURFACE_MIN_LOD_MASK GEN6_MASK(31, 28) |
/* Surface state DW5 */ |
#define GEN6_SURFACE_X_OFFSET_SHIFT 25 |
#define GEN6_SURFACE_X_OFFSET_MASK GEN6_MASK(31, 25) |
#define GEN6_SURFACE_Y_OFFSET_SHIFT 20 |
#define GEN6_SURFACE_Y_OFFSET_MASK GEN6_MASK(23, 20) |
struct gen6_cc_viewport { |
float min_depth; |
float max_depth; |
}; |
typedef enum { |
SAMPLER_FILTER_NEAREST = 0, |
SAMPLER_FILTER_BILINEAR, |
FILTER_COUNT |
} sampler_filter_t; |
typedef enum { |
SAMPLER_EXTEND_NONE = 0, |
SAMPLER_EXTEND_REPEAT, |
SAMPLER_EXTEND_PAD, |
SAMPLER_EXTEND_REFLECT, |
EXTEND_COUNT |
} sampler_extend_t; |
#endif |
/drivers/video/Intel-2D/sna/gen7_render.c |
---|
0,0 → 1,2326 |
/* |
* Copyright © 2006,2008,2011 Intel Corporation |
* Copyright © 2007 Red Hat, Inc. |
* |
* Permission is hereby granted, free of charge, to any person obtaining a |
* copy of this software and associated documentation files (the "Software"), |
* to deal in the Software without restriction, including without limitation |
* the rights to use, copy, modify, merge, publish, distribute, sublicense, |
* and/or sell copies of the Software, and to permit persons to whom the |
* Software is furnished to do so, subject to the following conditions: |
* |
* The above copyright notice and this permission notice (including the next |
* paragraph) shall be included in all copies or substantial portions of the |
* Software. |
* |
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL |
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, |
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE |
* SOFTWARE. |
* |
* Authors: |
* Wang Zhenyu <zhenyu.z.wang@sna.com> |
* Eric Anholt <eric@anholt.net> |
* Carl Worth <cworth@redhat.com> |
* Keith Packard <keithp@keithp.com> |
* Chris Wilson <chris@chris-wilson.co.uk> |
* |
*/ |
#ifdef HAVE_CONFIG_H |
#include "config.h" |
#endif |
#include "sna.h" |
#include "sna_reg.h" |
#include "sna_render.h" |
#include "sna_render_inline.h" |
//#include "sna_video.h" |
#include "brw/brw.h" |
#include "gen7_render.h" |
#include "gen4_source.h" |
#include "gen4_vertex.h" |
#define ALWAYS_FLUSH 0 |
#define NO_COMPOSITE 0 |
#define NO_COMPOSITE_SPANS 0 |
#define NO_COPY 0 |
#define NO_COPY_BOXES 0 |
#define NO_FILL 0 |
#define NO_FILL_BOXES 0 |
#define NO_FILL_ONE 0 |
#define NO_FILL_CLEAR 0 |
#define NO_RING_SWITCH 0 |
#define USE_8_PIXEL_DISPATCH 1 |
#define USE_16_PIXEL_DISPATCH 1 |
#define USE_32_PIXEL_DISPATCH 0 |
#if !USE_8_PIXEL_DISPATCH && !USE_16_PIXEL_DISPATCH && !USE_32_PIXEL_DISPATCH |
#error "Must select at least 8, 16 or 32 pixel dispatch" |
#endif |
#define GEN7_MAX_SIZE 16384 |
/* XXX Todo |
* |
* STR (software tiled rendering) mode. No, really. |
* 64x32 pixel blocks align with the rendering cache. Worth considering. |
*/ |
#define is_aligned(x, y) (((x) & ((y) - 1)) == 0) |
struct gt_info { |
const char *name; |
uint32_t max_vs_threads; |
uint32_t max_gs_threads; |
uint32_t max_wm_threads; |
struct { |
int size; |
int max_vs_entries; |
int max_gs_entries; |
int push_ps_size; /* in 1KBs */ |
} urb; |
int gt; |
}; |
static const struct gt_info ivb_gt_info = { |
.name = "Ivybridge (gen7)", |
.max_vs_threads = 16, |
.max_gs_threads = 16, |
.max_wm_threads = (16-1) << IVB_PS_MAX_THREADS_SHIFT, |
.urb = { 128, 64, 64, 8 }, |
.gt = 0, |
}; |
static const struct gt_info ivb_gt1_info = { |
.name = "Ivybridge (gen7, gt1)", |
.max_vs_threads = 36, |
.max_gs_threads = 36, |
.max_wm_threads = (48-1) << IVB_PS_MAX_THREADS_SHIFT, |
.urb = { 128, 512, 192, 8 }, |
.gt = 1, |
}; |
static const struct gt_info ivb_gt2_info = { |
.name = "Ivybridge (gen7, gt2)", |
.max_vs_threads = 128, |
.max_gs_threads = 128, |
.max_wm_threads = (172-1) << IVB_PS_MAX_THREADS_SHIFT, |
.urb = { 256, 704, 320, 8 }, |
.gt = 2, |
}; |
static const struct gt_info byt_gt_info = { |
.name = "Baytrail (gen7)", |
.urb = { 128, 64, 64 }, |
.max_vs_threads = 36, |
.max_gs_threads = 36, |
.max_wm_threads = (48-1) << IVB_PS_MAX_THREADS_SHIFT, |
.urb = { 128, 512, 192, 8 }, |
.gt = 1, |
}; |
static const struct gt_info hsw_gt_info = { |
.name = "Haswell (gen7.5)", |
.max_vs_threads = 8, |
.max_gs_threads = 8, |
.max_wm_threads = |
(8 - 1) << HSW_PS_MAX_THREADS_SHIFT | |
1 << HSW_PS_SAMPLE_MASK_SHIFT, |
.urb = { 128, 64, 64, 8 }, |
.gt = 0, |
}; |
static const struct gt_info hsw_gt1_info = { |
.name = "Haswell (gen7.5, gt1)", |
.max_vs_threads = 70, |
.max_gs_threads = 70, |
.max_wm_threads = |
(102 - 1) << HSW_PS_MAX_THREADS_SHIFT | |
1 << HSW_PS_SAMPLE_MASK_SHIFT, |
.urb = { 128, 640, 256, 8 }, |
.gt = 1, |
}; |
static const struct gt_info hsw_gt2_info = { |
.name = "Haswell (gen7.5, gt2)", |
.max_vs_threads = 140, |
.max_gs_threads = 140, |
.max_wm_threads = |
(140 - 1) << HSW_PS_MAX_THREADS_SHIFT | |
1 << HSW_PS_SAMPLE_MASK_SHIFT, |
.urb = { 256, 1664, 640, 8 }, |
.gt = 2, |
}; |
static const struct gt_info hsw_gt3_info = { |
.name = "Haswell (gen7.5, gt3)", |
.max_vs_threads = 280, |
.max_gs_threads = 280, |
.max_wm_threads = |
(280 - 1) << HSW_PS_MAX_THREADS_SHIFT | |
1 << HSW_PS_SAMPLE_MASK_SHIFT, |
.urb = { 512, 3328, 1280, 16 }, |
.gt = 3, |
}; |
inline static bool is_ivb(struct sna *sna) |
{ |
return sna->kgem.gen == 070; |
} |
inline static bool is_byt(struct sna *sna) |
{ |
return sna->kgem.gen == 071; |
} |
inline static bool is_hsw(struct sna *sna) |
{ |
return sna->kgem.gen == 075; |
} |
static const uint32_t ps_kernel_packed[][4] = { |
#include "exa_wm_src_affine.g7b" |
#include "exa_wm_src_sample_argb.g7b" |
#include "exa_wm_yuv_rgb.g7b" |
#include "exa_wm_write.g7b" |
}; |
static const uint32_t ps_kernel_planar[][4] = { |
#include "exa_wm_src_affine.g7b" |
#include "exa_wm_src_sample_planar.g7b" |
#include "exa_wm_yuv_rgb.g7b" |
#include "exa_wm_write.g7b" |
}; |
#define KERNEL(kernel_enum, kernel, num_surfaces) \ |
[GEN7_WM_KERNEL_##kernel_enum] = {#kernel_enum, kernel, sizeof(kernel), num_surfaces} |
#define NOKERNEL(kernel_enum, func, num_surfaces) \ |
[GEN7_WM_KERNEL_##kernel_enum] = {#kernel_enum, (void *)func, 0, num_surfaces} |
static const struct wm_kernel_info { |
const char *name; |
const void *data; |
unsigned int size; |
int num_surfaces; |
} wm_kernels[] = { |
NOKERNEL(NOMASK, brw_wm_kernel__affine, 2), |
NOKERNEL(NOMASK_P, brw_wm_kernel__projective, 2), |
NOKERNEL(MASK, brw_wm_kernel__affine_mask, 3), |
NOKERNEL(MASK_P, brw_wm_kernel__projective_mask, 3), |
NOKERNEL(MASKCA, brw_wm_kernel__affine_mask_ca, 3), |
NOKERNEL(MASKCA_P, brw_wm_kernel__projective_mask_ca, 3), |
NOKERNEL(MASKSA, brw_wm_kernel__affine_mask_sa, 3), |
NOKERNEL(MASKSA_P, brw_wm_kernel__projective_mask_sa, 3), |
NOKERNEL(OPACITY, brw_wm_kernel__affine_opacity, 2), |
NOKERNEL(OPACITY_P, brw_wm_kernel__projective_opacity, 2), |
KERNEL(VIDEO_PLANAR, ps_kernel_planar, 7), |
KERNEL(VIDEO_PACKED, ps_kernel_packed, 2), |
}; |
#undef KERNEL |
static const struct blendinfo { |
bool src_alpha; |
uint32_t src_blend; |
uint32_t dst_blend; |
} gen7_blend_op[] = { |
/* Clear */ {0, GEN7_BLENDFACTOR_ZERO, GEN7_BLENDFACTOR_ZERO}, |
/* Src */ {0, GEN7_BLENDFACTOR_ONE, GEN7_BLENDFACTOR_ZERO}, |
/* Dst */ {0, GEN7_BLENDFACTOR_ZERO, GEN7_BLENDFACTOR_ONE}, |
/* Over */ {1, GEN7_BLENDFACTOR_ONE, GEN7_BLENDFACTOR_INV_SRC_ALPHA}, |
/* OverReverse */ {0, GEN7_BLENDFACTOR_INV_DST_ALPHA, GEN7_BLENDFACTOR_ONE}, |
/* In */ {0, GEN7_BLENDFACTOR_DST_ALPHA, GEN7_BLENDFACTOR_ZERO}, |
/* InReverse */ {1, GEN7_BLENDFACTOR_ZERO, GEN7_BLENDFACTOR_SRC_ALPHA}, |
/* Out */ {0, GEN7_BLENDFACTOR_INV_DST_ALPHA, GEN7_BLENDFACTOR_ZERO}, |
/* OutReverse */ {1, GEN7_BLENDFACTOR_ZERO, GEN7_BLENDFACTOR_INV_SRC_ALPHA}, |
/* Atop */ {1, GEN7_BLENDFACTOR_DST_ALPHA, GEN7_BLENDFACTOR_INV_SRC_ALPHA}, |
/* AtopReverse */ {1, GEN7_BLENDFACTOR_INV_DST_ALPHA, GEN7_BLENDFACTOR_SRC_ALPHA}, |
/* Xor */ {1, GEN7_BLENDFACTOR_INV_DST_ALPHA, GEN7_BLENDFACTOR_INV_SRC_ALPHA}, |
/* Add */ {0, GEN7_BLENDFACTOR_ONE, GEN7_BLENDFACTOR_ONE}, |
}; |
/** |
* Highest-valued BLENDFACTOR used in gen7_blend_op. |
* |
* This leaves out GEN7_BLENDFACTOR_INV_DST_COLOR, |
* GEN7_BLENDFACTOR_INV_CONST_{COLOR,ALPHA}, |
* GEN7_BLENDFACTOR_INV_SRC1_{COLOR,ALPHA} |
*/ |
#define GEN7_BLENDFACTOR_COUNT (GEN7_BLENDFACTOR_INV_DST_ALPHA + 1) |
#define GEN7_BLEND_STATE_PADDED_SIZE ALIGN(sizeof(struct gen7_blend_state), 64) |
#define BLEND_OFFSET(s, d) \ |
((d != GEN7_BLENDFACTOR_ZERO) << 15 | \ |
(((s) * GEN7_BLENDFACTOR_COUNT + (d)) * GEN7_BLEND_STATE_PADDED_SIZE)) |
#define NO_BLEND BLEND_OFFSET(GEN7_BLENDFACTOR_ONE, GEN7_BLENDFACTOR_ZERO) |
#define CLEAR BLEND_OFFSET(GEN7_BLENDFACTOR_ZERO, GEN7_BLENDFACTOR_ZERO) |
#define SAMPLER_OFFSET(sf, se, mf, me) \ |
((((((sf) * EXTEND_COUNT + (se)) * FILTER_COUNT + (mf)) * EXTEND_COUNT + (me)) + 2) * 2 * sizeof(struct gen7_sampler_state)) |
#define VERTEX_2s2s 0 |
#define COPY_SAMPLER 0 |
#define COPY_VERTEX VERTEX_2s2s |
#define COPY_FLAGS(a) GEN7_SET_FLAGS(COPY_SAMPLER, (a) == GXcopy ? NO_BLEND : CLEAR, GEN7_WM_KERNEL_NOMASK, COPY_VERTEX) |
#define FILL_SAMPLER (2 * sizeof(struct gen7_sampler_state)) |
#define FILL_VERTEX VERTEX_2s2s |
#define FILL_FLAGS(op, format) GEN7_SET_FLAGS(FILL_SAMPLER, gen7_get_blend((op), false, (format)), GEN7_WM_KERNEL_NOMASK, FILL_VERTEX) |
#define FILL_FLAGS_NOBLEND GEN7_SET_FLAGS(FILL_SAMPLER, NO_BLEND, GEN7_WM_KERNEL_NOMASK, FILL_VERTEX) |
#define GEN7_SAMPLER(f) (((f) >> 16) & 0xfff0) |
#define GEN7_BLEND(f) (((f) >> 0) & 0x7ff0) |
#define GEN7_READS_DST(f) (((f) >> 15) & 1) |
#define GEN7_KERNEL(f) (((f) >> 16) & 0xf) |
#define GEN7_VERTEX(f) (((f) >> 0) & 0xf) |
#define GEN7_SET_FLAGS(S, B, K, V) (((S) | (K)) << 16 | ((B) | (V))) |
#define OUT_BATCH(v) batch_emit(sna, v) |
#define OUT_VERTEX(x,y) vertex_emit_2s(sna, x,y) |
#define OUT_VERTEX_F(v) vertex_emit(sna, v) |
static inline bool too_large(int width, int height) |
{ |
return width > GEN7_MAX_SIZE || height > GEN7_MAX_SIZE; |
} |
static uint32_t gen7_get_blend(int op, |
bool has_component_alpha, |
uint32_t dst_format) |
{ |
uint32_t src, dst; |
src = GEN7_BLENDFACTOR_ONE; //gen6_blend_op[op].src_blend; |
dst = GEN7_BLENDFACTOR_INV_SRC_ALPHA; //gen6_blend_op[op].dst_blend; |
#if 0 |
/* If there's no dst alpha channel, adjust the blend op so that |
* we'll treat it always as 1. |
*/ |
if (PICT_FORMAT_A(dst_format) == 0) { |
if (src == GEN7_BLENDFACTOR_DST_ALPHA) |
src = GEN7_BLENDFACTOR_ONE; |
else if (src == GEN7_BLENDFACTOR_INV_DST_ALPHA) |
src = GEN7_BLENDFACTOR_ZERO; |
} |
/* If the source alpha is being used, then we should only be in a |
* case where the source blend factor is 0, and the source blend |
* value is the mask channels multiplied by the source picture's alpha. |
*/ |
if (has_component_alpha && gen7_blend_op[op].src_alpha) { |
if (dst == GEN7_BLENDFACTOR_SRC_ALPHA) |
dst = GEN7_BLENDFACTOR_SRC_COLOR; |
else if (dst == GEN7_BLENDFACTOR_INV_SRC_ALPHA) |
dst = GEN7_BLENDFACTOR_INV_SRC_COLOR; |
} |
#endif |
DBG(("blend op=%d, dst=%x [A=%d] => src=%d, dst=%d => offset=%x\n", |
op, dst_format, PICT_FORMAT_A(dst_format), |
src, dst, (int)BLEND_OFFSET(src, dst))); |
return BLEND_OFFSET(src, dst); |
} |
static uint32_t gen7_get_card_format(PictFormat format) |
{ |
switch (format) { |
default: |
return -1; |
case PICT_a8r8g8b8: |
return GEN7_SURFACEFORMAT_B8G8R8A8_UNORM; |
case PICT_x8r8g8b8: |
return GEN7_SURFACEFORMAT_B8G8R8X8_UNORM; |
case PICT_a8b8g8r8: |
return GEN7_SURFACEFORMAT_R8G8B8A8_UNORM; |
case PICT_x8b8g8r8: |
return GEN7_SURFACEFORMAT_R8G8B8X8_UNORM; |
case PICT_a2r10g10b10: |
return GEN7_SURFACEFORMAT_B10G10R10A2_UNORM; |
case PICT_x2r10g10b10: |
return GEN7_SURFACEFORMAT_B10G10R10X2_UNORM; |
case PICT_r8g8b8: |
return GEN7_SURFACEFORMAT_R8G8B8_UNORM; |
case PICT_r5g6b5: |
return GEN7_SURFACEFORMAT_B5G6R5_UNORM; |
case PICT_a1r5g5b5: |
return GEN7_SURFACEFORMAT_B5G5R5A1_UNORM; |
case PICT_a8: |
return GEN7_SURFACEFORMAT_A8_UNORM; |
case PICT_a4r4g4b4: |
return GEN7_SURFACEFORMAT_B4G4R4A4_UNORM; |
} |
} |
static uint32_t gen7_get_dest_format(PictFormat format) |
{ |
switch (format) { |
default: |
return -1; |
case PICT_a8r8g8b8: |
case PICT_x8r8g8b8: |
return GEN7_SURFACEFORMAT_B8G8R8A8_UNORM; |
case PICT_a8b8g8r8: |
case PICT_x8b8g8r8: |
return GEN7_SURFACEFORMAT_R8G8B8A8_UNORM; |
case PICT_a2r10g10b10: |
case PICT_x2r10g10b10: |
return GEN7_SURFACEFORMAT_B10G10R10A2_UNORM; |
case PICT_r5g6b5: |
return GEN7_SURFACEFORMAT_B5G6R5_UNORM; |
case PICT_x1r5g5b5: |
case PICT_a1r5g5b5: |
return GEN7_SURFACEFORMAT_B5G5R5A1_UNORM; |
case PICT_a8: |
return GEN7_SURFACEFORMAT_A8_UNORM; |
case PICT_a4r4g4b4: |
case PICT_x4r4g4b4: |
return GEN7_SURFACEFORMAT_B4G4R4A4_UNORM; |
} |
} |
static int |
gen7_choose_composite_kernel(int op, bool has_mask, bool is_ca, bool is_affine) |
{ |
int base; |
if (has_mask) { |
if (is_ca) { |
if (gen7_blend_op[op].src_alpha) |
base = GEN7_WM_KERNEL_MASKSA; |
else |
base = GEN7_WM_KERNEL_MASKCA; |
} else |
base = GEN7_WM_KERNEL_MASK; |
} else |
base = GEN7_WM_KERNEL_NOMASK; |
return base + !is_affine; |
} |
static void |
gen7_emit_urb(struct sna *sna) |
{ |
OUT_BATCH(GEN7_3DSTATE_PUSH_CONSTANT_ALLOC_PS | (2 - 2)); |
OUT_BATCH(sna->render_state.gen7.info->urb.push_ps_size); |
/* num of VS entries must be divisible by 8 if size < 9 */ |
OUT_BATCH(GEN7_3DSTATE_URB_VS | (2 - 2)); |
OUT_BATCH((sna->render_state.gen7.info->urb.max_vs_entries << GEN7_URB_ENTRY_NUMBER_SHIFT) | |
(2 - 1) << GEN7_URB_ENTRY_SIZE_SHIFT | |
(1 << GEN7_URB_STARTING_ADDRESS_SHIFT)); |
OUT_BATCH(GEN7_3DSTATE_URB_HS | (2 - 2)); |
OUT_BATCH((0 << GEN7_URB_ENTRY_SIZE_SHIFT) | |
(2 << GEN7_URB_STARTING_ADDRESS_SHIFT)); |
OUT_BATCH(GEN7_3DSTATE_URB_DS | (2 - 2)); |
OUT_BATCH((0 << GEN7_URB_ENTRY_SIZE_SHIFT) | |
(2 << GEN7_URB_STARTING_ADDRESS_SHIFT)); |
OUT_BATCH(GEN7_3DSTATE_URB_GS | (2 - 2)); |
OUT_BATCH((0 << GEN7_URB_ENTRY_SIZE_SHIFT) | |
(1 << GEN7_URB_STARTING_ADDRESS_SHIFT)); |
} |
static void |
gen7_emit_state_base_address(struct sna *sna) |
{ |
uint32_t mocs; |
mocs = is_hsw(sna) ? 5 << 8 : 3 << 8; |
OUT_BATCH(GEN7_STATE_BASE_ADDRESS | (10 - 2)); |
OUT_BATCH(0); /* general */ |
OUT_BATCH(kgem_add_reloc(&sna->kgem, /* surface */ |
sna->kgem.nbatch, |
NULL, |
I915_GEM_DOMAIN_INSTRUCTION << 16, |
BASE_ADDRESS_MODIFY)); |
OUT_BATCH(kgem_add_reloc(&sna->kgem, /* dynamic */ |
sna->kgem.nbatch, |
sna->render_state.gen7.general_bo, |
I915_GEM_DOMAIN_INSTRUCTION << 16, |
mocs | BASE_ADDRESS_MODIFY)); |
OUT_BATCH(0); /* indirect */ |
OUT_BATCH(kgem_add_reloc(&sna->kgem, /* instruction */ |
sna->kgem.nbatch, |
sna->render_state.gen7.general_bo, |
I915_GEM_DOMAIN_INSTRUCTION << 16, |
mocs | BASE_ADDRESS_MODIFY)); |
/* upper bounds, disable */ |
OUT_BATCH(0); |
OUT_BATCH(BASE_ADDRESS_MODIFY); |
OUT_BATCH(0); |
OUT_BATCH(BASE_ADDRESS_MODIFY); |
} |
static void |
gen7_disable_vs(struct sna *sna) |
{ |
/* For future reference: |
* A PIPE_CONTROL with post-sync op set to 1 and a depth stall needs |
* to be emitted just prior to change VS state, i.e. 3DSTATE_VS, |
* 3DSTATE_URB_VS, 3DSTATE_CONSTANT_VS, |
* 3DSTATE_BINDING_TABLE_POINTER_VS, 3DSTATE_SAMPLER_STATE_POINTER_VS. |
* |
* Here we saved by the full-flush incurred when emitting |
* the batchbuffer. |
*/ |
OUT_BATCH(GEN7_3DSTATE_VS | (6 - 2)); |
OUT_BATCH(0); /* no VS kernel */ |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); /* pass-through */ |
#if 0 |
OUT_BATCH(GEN7_3DSTATE_CONSTANT_VS | (7 - 2)); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(GEN7_3DSTATE_BINDING_TABLE_POINTERS_VS | (2 - 2)); |
OUT_BATCH(0); |
OUT_BATCH(GEN7_3DSTATE_SAMPLER_STATE_POINTERS_VS | (2 - 2)); |
OUT_BATCH(0); |
#endif |
} |
static void |
gen7_disable_hs(struct sna *sna) |
{ |
OUT_BATCH(GEN7_3DSTATE_HS | (7 - 2)); |
OUT_BATCH(0); /* no HS kernel */ |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); /* pass-through */ |
#if 0 |
OUT_BATCH(GEN7_3DSTATE_CONSTANT_HS | (7 - 2)); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(GEN7_3DSTATE_BINDING_TABLE_POINTERS_HS | (2 - 2)); |
OUT_BATCH(0); |
OUT_BATCH(GEN7_3DSTATE_SAMPLER_STATE_POINTERS_HS | (2 - 2)); |
OUT_BATCH(0); |
#endif |
} |
static void |
gen7_disable_te(struct sna *sna) |
{ |
OUT_BATCH(GEN7_3DSTATE_TE | (4 - 2)); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
} |
static void |
gen7_disable_ds(struct sna *sna) |
{ |
OUT_BATCH(GEN7_3DSTATE_DS | (6 - 2)); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
#if 0 |
OUT_BATCH(GEN7_3DSTATE_CONSTANT_DS | (7 - 2)); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(GEN7_3DSTATE_BINDING_TABLE_POINTERS_DS | (2 - 2)); |
OUT_BATCH(0); |
OUT_BATCH(GEN7_3DSTATE_SAMPLER_STATE_POINTERS_DS | (2 - 2)); |
OUT_BATCH(0); |
#endif |
} |
static void |
gen7_disable_gs(struct sna *sna) |
{ |
OUT_BATCH(GEN7_3DSTATE_GS | (7 - 2)); |
OUT_BATCH(0); /* no GS kernel */ |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); /* pass-through */ |
#if 0 |
OUT_BATCH(GEN7_3DSTATE_CONSTANT_GS | (7 - 2)); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(GEN7_3DSTATE_BINDING_TABLE_POINTERS_GS | (2 - 2)); |
OUT_BATCH(0); |
OUT_BATCH(GEN7_3DSTATE_SAMPLER_STATE_POINTERS_GS | (2 - 2)); |
OUT_BATCH(0); |
#endif |
} |
static void |
gen7_disable_streamout(struct sna *sna) |
{ |
OUT_BATCH(GEN7_3DSTATE_STREAMOUT | (3 - 2)); |
OUT_BATCH(0); |
OUT_BATCH(0); |
} |
static void |
gen7_emit_sf_invariant(struct sna *sna) |
{ |
OUT_BATCH(GEN7_3DSTATE_SF | (7 - 2)); |
OUT_BATCH(0); |
OUT_BATCH(GEN7_3DSTATE_SF_CULL_NONE); |
OUT_BATCH(2 << GEN7_3DSTATE_SF_TRIFAN_PROVOKE_SHIFT); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
} |
static void |
gen7_emit_cc_invariant(struct sna *sna) |
{ |
#if 0 /* unused, no change */ |
OUT_BATCH(GEN7_3DSTATE_CC_STATE_POINTERS | (2 - 2)); |
OUT_BATCH(0); |
OUT_BATCH(GEN7_3DSTATE_DEPTH_STENCIL_STATE_POINTERS | (2 - 2)); |
OUT_BATCH(0); |
#endif |
/* XXX clear to be safe */ |
OUT_BATCH(GEN7_3DSTATE_VIEWPORT_STATE_POINTERS_CC | (2 - 2)); |
OUT_BATCH(0); |
} |
static void |
gen7_disable_clip(struct sna *sna) |
{ |
OUT_BATCH(GEN7_3DSTATE_CLIP | (4 - 2)); |
OUT_BATCH(0); |
OUT_BATCH(0); /* pass-through */ |
OUT_BATCH(0); |
OUT_BATCH(GEN7_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CL | (2 - 2)); |
OUT_BATCH(0); |
} |
static void |
gen7_emit_wm_invariant(struct sna *sna) |
{ |
OUT_BATCH(GEN7_3DSTATE_WM | (3 - 2)); |
OUT_BATCH(GEN7_WM_DISPATCH_ENABLE | |
GEN7_WM_PERSPECTIVE_PIXEL_BARYCENTRIC); |
OUT_BATCH(0); |
#if 0 |
/* XXX length bias of 7 in old spec? */ |
OUT_BATCH(GEN7_3DSTATE_CONSTANT_PS | (7 - 2)); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
#endif |
} |
static void |
gen7_emit_null_depth_buffer(struct sna *sna) |
{ |
OUT_BATCH(GEN7_3DSTATE_DEPTH_BUFFER | (7 - 2)); |
OUT_BATCH(GEN7_SURFACE_NULL << GEN7_3DSTATE_DEPTH_BUFFER_TYPE_SHIFT | |
GEN7_DEPTHFORMAT_D32_FLOAT << GEN7_3DSTATE_DEPTH_BUFFER_FORMAT_SHIFT); |
OUT_BATCH(0); /* disable depth, stencil and hiz */ |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
#if 0 |
OUT_BATCH(GEN7_3DSTATE_CLEAR_PARAMS | (3 - 2)); |
OUT_BATCH(0); |
OUT_BATCH(0); |
#endif |
} |
static void |
gen7_emit_invariant(struct sna *sna) |
{ |
OUT_BATCH(GEN7_PIPELINE_SELECT | PIPELINE_SELECT_3D); |
OUT_BATCH(GEN7_3DSTATE_MULTISAMPLE | (4 - 2)); |
OUT_BATCH(GEN7_3DSTATE_MULTISAMPLE_PIXEL_LOCATION_CENTER | |
GEN7_3DSTATE_MULTISAMPLE_NUMSAMPLES_1); /* 1 sample/pixel */ |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(GEN7_3DSTATE_SAMPLE_MASK | (2 - 2)); |
OUT_BATCH(1); |
gen7_emit_urb(sna); |
gen7_emit_state_base_address(sna); |
gen7_disable_vs(sna); |
gen7_disable_hs(sna); |
gen7_disable_te(sna); |
gen7_disable_ds(sna); |
gen7_disable_gs(sna); |
gen7_disable_clip(sna); |
gen7_emit_sf_invariant(sna); |
gen7_emit_wm_invariant(sna); |
gen7_emit_cc_invariant(sna); |
gen7_disable_streamout(sna); |
gen7_emit_null_depth_buffer(sna); |
sna->render_state.gen7.needs_invariant = false; |
} |
static void |
gen7_emit_cc(struct sna *sna, uint32_t blend_offset) |
{ |
struct gen7_render_state *render = &sna->render_state.gen7; |
if (render->blend == blend_offset) |
return; |
DBG(("%s: blend = %x\n", __FUNCTION__, blend_offset)); |
/* XXX can have upto 8 blend states preload, selectable via |
* Render Target Index. What other side-effects of Render Target Index? |
*/ |
assert (is_aligned(render->cc_blend + blend_offset, 64)); |
OUT_BATCH(GEN7_3DSTATE_BLEND_STATE_POINTERS | (2 - 2)); |
OUT_BATCH((render->cc_blend + blend_offset) | 1); |
render->blend = blend_offset; |
} |
static void |
gen7_emit_sampler(struct sna *sna, uint32_t state) |
{ |
if (sna->render_state.gen7.samplers == state) |
return; |
sna->render_state.gen7.samplers = state; |
DBG(("%s: sampler = %x\n", __FUNCTION__, state)); |
assert (is_aligned(sna->render_state.gen7.wm_state + state, 32)); |
OUT_BATCH(GEN7_3DSTATE_SAMPLER_STATE_POINTERS_PS | (2 - 2)); |
OUT_BATCH(sna->render_state.gen7.wm_state + state); |
} |
static void |
gen7_emit_sf(struct sna *sna, bool has_mask) |
{ |
int num_sf_outputs = has_mask ? 2 : 1; |
if (sna->render_state.gen7.num_sf_outputs == num_sf_outputs) |
return; |
DBG(("%s: num_sf_outputs=%d, read_length=%d, read_offset=%d\n", |
__FUNCTION__, num_sf_outputs, 1, 0)); |
sna->render_state.gen7.num_sf_outputs = num_sf_outputs; |
OUT_BATCH(GEN7_3DSTATE_SBE | (14 - 2)); |
OUT_BATCH(num_sf_outputs << GEN7_SBE_NUM_OUTPUTS_SHIFT | |
1 << GEN7_SBE_URB_ENTRY_READ_LENGTH_SHIFT | |
1 << GEN7_SBE_URB_ENTRY_READ_OFFSET_SHIFT); |
OUT_BATCH(0); |
OUT_BATCH(0); /* dw4 */ |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); /* dw8 */ |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); |
OUT_BATCH(0); /* dw12 */ |
OUT_BATCH(0); |
OUT_BATCH(0); |
} |
static void |
gen7_emit_wm(struct sna *sna, int kernel) |
{ |
const uint32_t *kernels; |
if (sna->render_state.gen7.kernel == kernel) |
return; |
sna->render_state.gen7.kernel = kernel; |
kernels = sna->render_state.gen7.wm_kernel[kernel]; |
DBG(("%s: switching to %s, num_surfaces=%d (8-wide? %d, 16-wide? %d, 32-wide? %d)\n", |
__FUNCTION__, |
wm_kernels[kernel].name, |
wm_kernels[kernel].num_surfaces, |
kernels[0], kernels[1], kernels[2])); |
OUT_BATCH(GEN7_3DSTATE_PS | (8 - 2)); |
OUT_BATCH(kernels[0] ?: kernels[1] ?: kernels[2]); |
OUT_BATCH(1 << GEN7_PS_SAMPLER_COUNT_SHIFT | |
wm_kernels[kernel].num_surfaces << GEN7_PS_BINDING_TABLE_ENTRY_COUNT_SHIFT); |
OUT_BATCH(0); /* scratch address */ |
OUT_BATCH(sna->render_state.gen7.info->max_wm_threads | |
(kernels[0] ? GEN7_PS_8_DISPATCH_ENABLE : 0) | |
(kernels[1] ? GEN7_PS_16_DISPATCH_ENABLE : 0) | |
(kernels[2] ? GEN7_PS_32_DISPATCH_ENABLE : 0) | |
GEN7_PS_ATTRIBUTE_ENABLE); |
OUT_BATCH((kernels[0] ? 4 : kernels[1] ? 6 : 8) << GEN7_PS_DISPATCH_START_GRF_SHIFT_0 | |
8 << GEN7_PS_DISPATCH_START_GRF_SHIFT_1 | |
6 << GEN7_PS_DISPATCH_START_GRF_SHIFT_2); |
OUT_BATCH(kernels[2]); |
OUT_BATCH(kernels[1]); |
} |
static bool |
gen7_emit_binding_table(struct sna *sna, uint16_t offset) |
{ |
if (sna->render_state.gen7.surface_table == offset) |
return false; |
/* Binding table pointers */ |
assert(is_aligned(4*offset, 32)); |
OUT_BATCH(GEN7_3DSTATE_BINDING_TABLE_POINTERS_PS | (2 - 2)); |
OUT_BATCH(offset*4); |
sna->render_state.gen7.surface_table = offset; |
return true; |
} |
static bool |
gen7_emit_drawing_rectangle(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
uint32_t limit = (op->dst.height - 1) << 16 | (op->dst.width - 1); |
uint32_t offset = (uint16_t)op->dst.y << 16 | (uint16_t)op->dst.x; |
assert(!too_large(op->dst.x, op->dst.y)); |
assert(!too_large(op->dst.width, op->dst.height)); |
if (sna->render_state.gen7.drawrect_limit == limit && |
sna->render_state.gen7.drawrect_offset == offset) |
return true; |
sna->render_state.gen7.drawrect_offset = offset; |
sna->render_state.gen7.drawrect_limit = limit; |
OUT_BATCH(GEN7_3DSTATE_DRAWING_RECTANGLE | (4 - 2)); |
OUT_BATCH(0); |
OUT_BATCH(limit); |
OUT_BATCH(offset); |
return false; |
} |
static void |
gen7_emit_vertex_elements(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
/* |
* vertex data in vertex buffer |
* position: (x, y) |
* texture coordinate 0: (u0, v0) if (is_affine is true) else (u0, v0, w0) |
* texture coordinate 1 if (has_mask is true): same as above |
*/ |
struct gen7_render_state *render = &sna->render_state.gen7; |
uint32_t src_format, dw; |
int id = GEN7_VERTEX(op->u.gen7.flags); |
bool has_mask; |
DBG(("%s: setup id=%d\n", __FUNCTION__, id)); |
if (render->ve_id == id) |
return; |
render->ve_id = id; |
/* The VUE layout |
* dword 0-3: pad (0.0, 0.0, 0.0. 0.0) |
* dword 4-7: position (x, y, 1.0, 1.0), |
* dword 8-11: texture coordinate 0 (u0, v0, w0, 1.0) |
* dword 12-15: texture coordinate 1 (u1, v1, w1, 1.0) |
* |
* dword 4-15 are fetched from vertex buffer |
*/ |
has_mask = (id >> 2) != 0; |
OUT_BATCH(GEN7_3DSTATE_VERTEX_ELEMENTS | |
((2 * (3 + has_mask)) + 1 - 2)); |
OUT_BATCH(id << GEN7_VE0_VERTEX_BUFFER_INDEX_SHIFT | GEN7_VE0_VALID | |
GEN7_SURFACEFORMAT_R32G32B32A32_FLOAT << GEN7_VE0_FORMAT_SHIFT | |
0 << GEN7_VE0_OFFSET_SHIFT); |
OUT_BATCH(GEN7_VFCOMPONENT_STORE_0 << GEN7_VE1_VFCOMPONENT_0_SHIFT | |
GEN7_VFCOMPONENT_STORE_0 << GEN7_VE1_VFCOMPONENT_1_SHIFT | |
GEN7_VFCOMPONENT_STORE_0 << GEN7_VE1_VFCOMPONENT_2_SHIFT | |
GEN7_VFCOMPONENT_STORE_0 << GEN7_VE1_VFCOMPONENT_3_SHIFT); |
/* x,y */ |
OUT_BATCH(id << GEN7_VE0_VERTEX_BUFFER_INDEX_SHIFT | GEN7_VE0_VALID | |
GEN7_SURFACEFORMAT_R16G16_SSCALED << GEN7_VE0_FORMAT_SHIFT | |
0 << GEN7_VE0_OFFSET_SHIFT); |
OUT_BATCH(GEN7_VFCOMPONENT_STORE_SRC << GEN7_VE1_VFCOMPONENT_0_SHIFT | |
GEN7_VFCOMPONENT_STORE_SRC << GEN7_VE1_VFCOMPONENT_1_SHIFT | |
GEN7_VFCOMPONENT_STORE_0 << GEN7_VE1_VFCOMPONENT_2_SHIFT | |
GEN7_VFCOMPONENT_STORE_1_FLT << GEN7_VE1_VFCOMPONENT_3_SHIFT); |
/* u0, v0, w0 */ |
DBG(("%s: first channel %d floats, offset=4b\n", __FUNCTION__, id & 3)); |
dw = GEN7_VFCOMPONENT_STORE_1_FLT << GEN7_VE1_VFCOMPONENT_3_SHIFT; |
switch (id & 3) { |
default: |
assert(0); |
case 0: |
src_format = GEN7_SURFACEFORMAT_R16G16_SSCALED; |
dw |= GEN7_VFCOMPONENT_STORE_SRC << GEN7_VE1_VFCOMPONENT_0_SHIFT; |
dw |= GEN7_VFCOMPONENT_STORE_SRC << GEN7_VE1_VFCOMPONENT_1_SHIFT; |
dw |= GEN7_VFCOMPONENT_STORE_0 << GEN7_VE1_VFCOMPONENT_2_SHIFT; |
break; |
case 1: |
src_format = GEN7_SURFACEFORMAT_R32_FLOAT; |
dw |= GEN7_VFCOMPONENT_STORE_SRC << GEN7_VE1_VFCOMPONENT_0_SHIFT; |
dw |= GEN7_VFCOMPONENT_STORE_0 << GEN7_VE1_VFCOMPONENT_1_SHIFT; |
dw |= GEN7_VFCOMPONENT_STORE_0 << GEN7_VE1_VFCOMPONENT_2_SHIFT; |
break; |
case 2: |
src_format = GEN7_SURFACEFORMAT_R32G32_FLOAT; |
dw |= GEN7_VFCOMPONENT_STORE_SRC << GEN7_VE1_VFCOMPONENT_0_SHIFT; |
dw |= GEN7_VFCOMPONENT_STORE_SRC << GEN7_VE1_VFCOMPONENT_1_SHIFT; |
dw |= GEN7_VFCOMPONENT_STORE_0 << GEN7_VE1_VFCOMPONENT_2_SHIFT; |
break; |
case 3: |
src_format = GEN7_SURFACEFORMAT_R32G32B32_FLOAT; |
dw |= GEN7_VFCOMPONENT_STORE_SRC << GEN7_VE1_VFCOMPONENT_0_SHIFT; |
dw |= GEN7_VFCOMPONENT_STORE_SRC << GEN7_VE1_VFCOMPONENT_1_SHIFT; |
dw |= GEN7_VFCOMPONENT_STORE_SRC << GEN7_VE1_VFCOMPONENT_2_SHIFT; |
break; |
} |
OUT_BATCH(id << GEN7_VE0_VERTEX_BUFFER_INDEX_SHIFT | GEN7_VE0_VALID | |
src_format << GEN7_VE0_FORMAT_SHIFT | |
4 << GEN7_VE0_OFFSET_SHIFT); |
OUT_BATCH(dw); |
/* u1, v1, w1 */ |
if (has_mask) { |
unsigned offset = 4 + ((id & 3) ?: 1) * sizeof(float); |
DBG(("%s: second channel %d floats, offset=%db\n", __FUNCTION__, id >> 2, offset)); |
dw = GEN7_VFCOMPONENT_STORE_1_FLT << GEN7_VE1_VFCOMPONENT_3_SHIFT; |
switch (id >> 2) { |
case 1: |
src_format = GEN7_SURFACEFORMAT_R32_FLOAT; |
dw |= GEN7_VFCOMPONENT_STORE_SRC << GEN7_VE1_VFCOMPONENT_0_SHIFT; |
dw |= GEN7_VFCOMPONENT_STORE_0 << GEN7_VE1_VFCOMPONENT_1_SHIFT; |
dw |= GEN7_VFCOMPONENT_STORE_0 << GEN7_VE1_VFCOMPONENT_2_SHIFT; |
break; |
default: |
assert(0); |
case 2: |
src_format = GEN7_SURFACEFORMAT_R32G32_FLOAT; |
dw |= GEN7_VFCOMPONENT_STORE_SRC << GEN7_VE1_VFCOMPONENT_0_SHIFT; |
dw |= GEN7_VFCOMPONENT_STORE_SRC << GEN7_VE1_VFCOMPONENT_1_SHIFT; |
dw |= GEN7_VFCOMPONENT_STORE_0 << GEN7_VE1_VFCOMPONENT_2_SHIFT; |
break; |
case 3: |
src_format = GEN7_SURFACEFORMAT_R32G32B32_FLOAT; |
dw |= GEN7_VFCOMPONENT_STORE_SRC << GEN7_VE1_VFCOMPONENT_0_SHIFT; |
dw |= GEN7_VFCOMPONENT_STORE_SRC << GEN7_VE1_VFCOMPONENT_1_SHIFT; |
dw |= GEN7_VFCOMPONENT_STORE_SRC << GEN7_VE1_VFCOMPONENT_2_SHIFT; |
break; |
} |
OUT_BATCH(id << GEN7_VE0_VERTEX_BUFFER_INDEX_SHIFT | GEN7_VE0_VALID | |
src_format << GEN7_VE0_FORMAT_SHIFT | |
offset << GEN7_VE0_OFFSET_SHIFT); |
OUT_BATCH(dw); |
} |
} |
inline static void |
gen7_emit_pipe_invalidate(struct sna *sna) |
{ |
OUT_BATCH(GEN7_PIPE_CONTROL | (4 - 2)); |
OUT_BATCH(GEN7_PIPE_CONTROL_WC_FLUSH | |
GEN7_PIPE_CONTROL_TC_FLUSH | |
GEN7_PIPE_CONTROL_CS_STALL); |
OUT_BATCH(0); |
OUT_BATCH(0); |
} |
inline static void |
gen7_emit_pipe_flush(struct sna *sna, bool need_stall) |
{ |
unsigned stall; |
stall = 0; |
if (need_stall) |
stall = (GEN7_PIPE_CONTROL_CS_STALL | |
GEN7_PIPE_CONTROL_STALL_AT_SCOREBOARD); |
OUT_BATCH(GEN7_PIPE_CONTROL | (4 - 2)); |
OUT_BATCH(GEN7_PIPE_CONTROL_WC_FLUSH | stall); |
OUT_BATCH(0); |
OUT_BATCH(0); |
} |
inline static void |
gen7_emit_pipe_stall(struct sna *sna) |
{ |
OUT_BATCH(GEN7_PIPE_CONTROL | (4 - 2)); |
OUT_BATCH(GEN7_PIPE_CONTROL_CS_STALL | |
GEN7_PIPE_CONTROL_STALL_AT_SCOREBOARD); |
OUT_BATCH(0); |
OUT_BATCH(0); |
} |
static void |
gen7_emit_state(struct sna *sna, |
const struct sna_composite_op *op, |
uint16_t wm_binding_table) |
{ |
bool need_stall; |
assert(op->dst.bo->exec); |
gen7_emit_cc(sna, GEN7_BLEND(op->u.gen7.flags)); |
gen7_emit_sampler(sna, GEN7_SAMPLER(op->u.gen7.flags)); |
gen7_emit_sf(sna, GEN7_VERTEX(op->u.gen7.flags) >> 2); |
gen7_emit_wm(sna, GEN7_KERNEL(op->u.gen7.flags)); |
gen7_emit_vertex_elements(sna, op); |
need_stall = gen7_emit_binding_table(sna, wm_binding_table); |
need_stall &= gen7_emit_drawing_rectangle(sna, op); |
if (ALWAYS_FLUSH || kgem_bo_is_dirty(op->src.bo) || kgem_bo_is_dirty(op->mask.bo)) { |
gen7_emit_pipe_invalidate(sna); |
kgem_clear_dirty(&sna->kgem); |
assert(op->dst.bo->exec); |
kgem_bo_mark_dirty(op->dst.bo); |
sna->render_state.gen7.emit_flush = false; |
need_stall = false; |
} |
if (sna->render_state.gen7.emit_flush) { |
gen7_emit_pipe_flush(sna, need_stall); |
need_stall = false; |
} |
if (need_stall) |
gen7_emit_pipe_stall(sna); |
sna->render_state.gen7.emit_flush = GEN7_READS_DST(op->u.gen7.flags); |
} |
static bool gen7_magic_ca_pass(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
struct gen7_render_state *state = &sna->render_state.gen7; |
if (!op->need_magic_ca_pass) |
return false; |
DBG(("%s: CA fixup (%d -> %d)\n", __FUNCTION__, |
sna->render.vertex_start, sna->render.vertex_index)); |
gen7_emit_pipe_stall(sna); |
gen7_emit_cc(sna, |
GEN7_BLEND(gen7_get_blend(PictOpAdd, true, |
op->dst.format))); |
gen7_emit_wm(sna, |
gen7_choose_composite_kernel(PictOpAdd, |
true, true, |
op->is_affine)); |
OUT_BATCH(GEN7_3DPRIMITIVE | (7- 2)); |
OUT_BATCH(GEN7_3DPRIMITIVE_VERTEX_SEQUENTIAL | _3DPRIM_RECTLIST); |
OUT_BATCH(sna->render.vertex_index - sna->render.vertex_start); |
OUT_BATCH(sna->render.vertex_start); |
OUT_BATCH(1); /* single instance */ |
OUT_BATCH(0); /* start instance location */ |
OUT_BATCH(0); /* index buffer offset, ignored */ |
state->last_primitive = sna->kgem.nbatch; |
return true; |
} |
static void null_create(struct sna_static_stream *stream) |
{ |
/* A bunch of zeros useful for legacy border color and depth-stencil */ |
sna_static_stream_map(stream, 64, 64); |
} |
static void |
sampler_state_init(struct gen7_sampler_state *sampler_state, |
sampler_filter_t filter, |
sampler_extend_t extend) |
{ |
sampler_state->ss0.lod_preclamp = 1; /* GL mode */ |
/* We use the legacy mode to get the semantics specified by |
* the Render extension. */ |
sampler_state->ss0.default_color_mode = GEN7_BORDER_COLOR_MODE_LEGACY; |
switch (filter) { |
default: |
case SAMPLER_FILTER_NEAREST: |
sampler_state->ss0.min_filter = GEN7_MAPFILTER_NEAREST; |
sampler_state->ss0.mag_filter = GEN7_MAPFILTER_NEAREST; |
break; |
case SAMPLER_FILTER_BILINEAR: |
sampler_state->ss0.min_filter = GEN7_MAPFILTER_LINEAR; |
sampler_state->ss0.mag_filter = GEN7_MAPFILTER_LINEAR; |
break; |
} |
switch (extend) { |
default: |
case SAMPLER_EXTEND_NONE: |
sampler_state->ss3.r_wrap_mode = GEN7_TEXCOORDMODE_CLAMP_BORDER; |
sampler_state->ss3.s_wrap_mode = GEN7_TEXCOORDMODE_CLAMP_BORDER; |
sampler_state->ss3.t_wrap_mode = GEN7_TEXCOORDMODE_CLAMP_BORDER; |
break; |
case SAMPLER_EXTEND_REPEAT: |
sampler_state->ss3.r_wrap_mode = GEN7_TEXCOORDMODE_WRAP; |
sampler_state->ss3.s_wrap_mode = GEN7_TEXCOORDMODE_WRAP; |
sampler_state->ss3.t_wrap_mode = GEN7_TEXCOORDMODE_WRAP; |
break; |
case SAMPLER_EXTEND_PAD: |
sampler_state->ss3.r_wrap_mode = GEN7_TEXCOORDMODE_CLAMP; |
sampler_state->ss3.s_wrap_mode = GEN7_TEXCOORDMODE_CLAMP; |
sampler_state->ss3.t_wrap_mode = GEN7_TEXCOORDMODE_CLAMP; |
break; |
case SAMPLER_EXTEND_REFLECT: |
sampler_state->ss3.r_wrap_mode = GEN7_TEXCOORDMODE_MIRROR; |
sampler_state->ss3.s_wrap_mode = GEN7_TEXCOORDMODE_MIRROR; |
sampler_state->ss3.t_wrap_mode = GEN7_TEXCOORDMODE_MIRROR; |
break; |
} |
} |
static void |
sampler_copy_init(struct gen7_sampler_state *ss) |
{ |
sampler_state_init(ss, SAMPLER_FILTER_NEAREST, SAMPLER_EXTEND_NONE); |
ss->ss3.non_normalized_coord = 1; |
sampler_state_init(ss+1, SAMPLER_FILTER_NEAREST, SAMPLER_EXTEND_NONE); |
} |
static void |
sampler_fill_init(struct gen7_sampler_state *ss) |
{ |
sampler_state_init(ss, SAMPLER_FILTER_NEAREST, SAMPLER_EXTEND_REPEAT); |
ss->ss3.non_normalized_coord = 1; |
sampler_state_init(ss+1, SAMPLER_FILTER_NEAREST, SAMPLER_EXTEND_NONE); |
} |
static uint32_t |
gen7_tiling_bits(uint32_t tiling) |
{ |
switch (tiling) { |
default: assert(0); |
case I915_TILING_NONE: return 0; |
case I915_TILING_X: return GEN7_SURFACE_TILED; |
case I915_TILING_Y: return GEN7_SURFACE_TILED | GEN7_SURFACE_TILED_Y; |
} |
} |
/** |
* Sets up the common fields for a surface state buffer for the given |
* picture in the given surface state buffer. |
*/ |
static uint32_t |
gen7_bind_bo(struct sna *sna, |
struct kgem_bo *bo, |
uint32_t width, |
uint32_t height, |
uint32_t format, |
bool is_dst) |
{ |
uint32_t *ss; |
uint32_t domains; |
int offset; |
uint32_t is_scanout = is_dst && bo->scanout; |
COMPILE_TIME_ASSERT(sizeof(struct gen7_surface_state) == 32); |
/* After the first bind, we manage the cache domains within the batch */ |
offset = kgem_bo_get_binding(bo, format | is_dst << 30 | is_scanout << 31); |
if (offset) { |
if (is_dst) |
kgem_bo_mark_dirty(bo); |
return offset * sizeof(uint32_t); |
} |
offset = sna->kgem.surface -= |
sizeof(struct gen7_surface_state) / sizeof(uint32_t); |
ss = sna->kgem.batch + offset; |
ss[0] = (GEN7_SURFACE_2D << GEN7_SURFACE_TYPE_SHIFT | |
gen7_tiling_bits(bo->tiling) | |
format << GEN7_SURFACE_FORMAT_SHIFT); |
if (bo->tiling == I915_TILING_Y) |
ss[0] |= GEN7_SURFACE_VALIGN_4; |
if (is_dst) { |
ss[0] |= GEN7_SURFACE_RC_READ_WRITE; |
domains = I915_GEM_DOMAIN_RENDER << 16 |I915_GEM_DOMAIN_RENDER; |
} else |
domains = I915_GEM_DOMAIN_SAMPLER << 16; |
ss[1] = kgem_add_reloc(&sna->kgem, offset + 1, bo, domains, 0); |
ss[2] = ((width - 1) << GEN7_SURFACE_WIDTH_SHIFT | |
(height - 1) << GEN7_SURFACE_HEIGHT_SHIFT); |
ss[3] = (bo->pitch - 1) << GEN7_SURFACE_PITCH_SHIFT; |
ss[4] = 0; |
ss[5] = (is_scanout || bo->io) ? 0 : is_hsw(sna) ? 5 << 16 : 3 << 16; |
ss[6] = 0; |
ss[7] = 0; |
if (is_hsw(sna)) |
ss[7] |= HSW_SURFACE_SWIZZLE(RED, GREEN, BLUE, ALPHA); |
kgem_bo_set_binding(bo, format | is_dst << 30 | is_scanout << 31, offset); |
DBG(("[%x] bind bo(handle=%d, addr=%d), format=%d, width=%d, height=%d, pitch=%d, tiling=%d -> %s\n", |
offset, bo->handle, ss[1], |
format, width, height, bo->pitch, bo->tiling, |
domains & 0xffff ? "render" : "sampler")); |
return offset * sizeof(uint32_t); |
} |
static void gen7_emit_vertex_buffer(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
int id = GEN7_VERTEX(op->u.gen7.flags); |
OUT_BATCH(GEN7_3DSTATE_VERTEX_BUFFERS | (5 - 2)); |
OUT_BATCH(id << GEN7_VB0_BUFFER_INDEX_SHIFT | |
GEN7_VB0_VERTEXDATA | |
GEN7_VB0_ADDRESS_MODIFY_ENABLE | |
4*op->floats_per_vertex << GEN7_VB0_BUFFER_PITCH_SHIFT); |
sna->render.vertex_reloc[sna->render.nvertex_reloc++] = sna->kgem.nbatch; |
OUT_BATCH(0); |
OUT_BATCH(~0); /* max address: disabled */ |
OUT_BATCH(0); |
sna->render.vb_id |= 1 << id; |
} |
static void gen7_emit_primitive(struct sna *sna) |
{ |
if (sna->kgem.nbatch == sna->render_state.gen7.last_primitive) { |
sna->render.vertex_offset = sna->kgem.nbatch - 5; |
return; |
} |
OUT_BATCH(GEN7_3DPRIMITIVE | (7- 2)); |
OUT_BATCH(GEN7_3DPRIMITIVE_VERTEX_SEQUENTIAL | _3DPRIM_RECTLIST); |
sna->render.vertex_offset = sna->kgem.nbatch; |
OUT_BATCH(0); /* vertex count, to be filled in later */ |
OUT_BATCH(sna->render.vertex_index); |
OUT_BATCH(1); /* single instance */ |
OUT_BATCH(0); /* start instance location */ |
OUT_BATCH(0); /* index buffer offset, ignored */ |
sna->render.vertex_start = sna->render.vertex_index; |
sna->render_state.gen7.last_primitive = sna->kgem.nbatch; |
} |
static bool gen7_rectangle_begin(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
int id = 1 << GEN7_VERTEX(op->u.gen7.flags); |
int ndwords; |
if (sna_vertex_wait__locked(&sna->render) && sna->render.vertex_offset) |
return true; |
ndwords = op->need_magic_ca_pass ? 60 : 6; |
if ((sna->render.vb_id & id) == 0) |
ndwords += 5; |
if (!kgem_check_batch(&sna->kgem, ndwords)) |
return false; |
if ((sna->render.vb_id & id) == 0) |
gen7_emit_vertex_buffer(sna, op); |
gen7_emit_primitive(sna); |
return true; |
} |
static int gen7_get_rectangles__flush(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
/* Preventing discarding new vbo after lock contention */ |
if (sna_vertex_wait__locked(&sna->render)) { |
int rem = vertex_space(sna); |
if (rem > op->floats_per_rect) |
return rem; |
} |
if (!kgem_check_batch(&sna->kgem, op->need_magic_ca_pass ? 65 : 6)) |
return 0; |
if (!kgem_check_reloc_and_exec(&sna->kgem, 2)) |
return 0; |
if (sna->render.vertex_offset) { |
gen4_vertex_flush(sna); |
if (gen7_magic_ca_pass(sna, op)) { |
gen7_emit_pipe_stall(sna); |
gen7_emit_cc(sna, GEN7_BLEND(op->u.gen7.flags)); |
gen7_emit_wm(sna, GEN7_KERNEL(op->u.gen7.flags)); |
} |
} |
return gen4_vertex_finish(sna); |
} |
inline static int gen7_get_rectangles(struct sna *sna, |
const struct sna_composite_op *op, |
int want, |
void (*emit_state)(struct sna *sna, const struct sna_composite_op *op)) |
{ |
int rem; |
assert(want); |
start: |
rem = vertex_space(sna); |
if (unlikely(rem < op->floats_per_rect)) { |
DBG(("flushing vbo for %s: %d < %d\n", |
__FUNCTION__, rem, op->floats_per_rect)); |
rem = gen7_get_rectangles__flush(sna, op); |
if (unlikely(rem == 0)) |
goto flush; |
} |
if (unlikely(sna->render.vertex_offset == 0)) { |
if (!gen7_rectangle_begin(sna, op)) |
goto flush; |
else |
goto start; |
} |
assert(rem <= vertex_space(sna)); |
assert(op->floats_per_rect <= rem); |
if (want > 1 && want * op->floats_per_rect > rem) |
want = rem / op->floats_per_rect; |
assert(want > 0); |
sna->render.vertex_index += 3*want; |
return want; |
flush: |
if (sna->render.vertex_offset) { |
gen4_vertex_flush(sna); |
gen7_magic_ca_pass(sna, op); |
} |
sna_vertex_wait__locked(&sna->render); |
_kgem_submit(&sna->kgem); |
emit_state(sna, op); |
goto start; |
} |
inline static uint32_t *gen7_composite_get_binding_table(struct sna *sna, |
uint16_t *offset) |
{ |
uint32_t *table; |
sna->kgem.surface -= |
sizeof(struct gen7_surface_state) / sizeof(uint32_t); |
/* Clear all surplus entries to zero in case of prefetch */ |
table = memset(sna->kgem.batch + sna->kgem.surface, |
0, sizeof(struct gen7_surface_state)); |
DBG(("%s(%x)\n", __FUNCTION__, 4*sna->kgem.surface)); |
*offset = sna->kgem.surface; |
return table; |
} |
static void |
gen7_get_batch(struct sna *sna, const struct sna_composite_op *op) |
{ |
kgem_set_mode(&sna->kgem, KGEM_RENDER, op->dst.bo); |
if (!kgem_check_batch_with_surfaces(&sna->kgem, 150, 4)) { |
DBG(("%s: flushing batch: %d < %d+%d\n", |
__FUNCTION__, sna->kgem.surface - sna->kgem.nbatch, |
150, 4*8)); |
_kgem_submit(&sna->kgem); |
_kgem_set_mode(&sna->kgem, KGEM_RENDER); |
} |
assert(sna->kgem.mode == KGEM_RENDER); |
assert(sna->kgem.ring == KGEM_RENDER); |
if (sna->render_state.gen7.needs_invariant) |
gen7_emit_invariant(sna); |
} |
static void gen7_emit_composite_state(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
uint32_t *binding_table; |
uint16_t offset; |
gen7_get_batch(sna, op); |
binding_table = gen7_composite_get_binding_table(sna, &offset); |
binding_table[0] = |
gen7_bind_bo(sna, |
op->dst.bo, op->dst.width, op->dst.height, |
gen7_get_dest_format(op->dst.format), |
true); |
binding_table[1] = |
gen7_bind_bo(sna, |
op->src.bo, op->src.width, op->src.height, |
op->src.card_format, |
false); |
if (op->mask.bo) { |
binding_table[2] = |
gen7_bind_bo(sna, |
op->mask.bo, |
op->mask.width, |
op->mask.height, |
op->mask.card_format, |
false); |
} |
if (sna->kgem.surface == offset && |
*(uint64_t *)(sna->kgem.batch + sna->render_state.gen7.surface_table) == *(uint64_t*)binding_table && |
(op->mask.bo == NULL || |
sna->kgem.batch[sna->render_state.gen7.surface_table+2] == binding_table[2])) { |
sna->kgem.surface += sizeof(struct gen7_surface_state) / sizeof(uint32_t); |
offset = sna->render_state.gen7.surface_table; |
} |
gen7_emit_state(sna, op, offset); |
} |
static void |
gen7_align_vertex(struct sna *sna, const struct sna_composite_op *op) |
{ |
if (op->floats_per_vertex != sna->render_state.gen7.floats_per_vertex) { |
if (sna->render.vertex_size - sna->render.vertex_used < 2*op->floats_per_rect) |
gen4_vertex_finish(sna); |
DBG(("aligning vertex: was %d, now %d floats per vertex, %d->%d\n", |
sna->render_state.gen7.floats_per_vertex, |
op->floats_per_vertex, |
sna->render.vertex_index, |
(sna->render.vertex_used + op->floats_per_vertex - 1) / op->floats_per_vertex)); |
sna->render.vertex_index = (sna->render.vertex_used + op->floats_per_vertex - 1) / op->floats_per_vertex; |
sna->render.vertex_used = sna->render.vertex_index * op->floats_per_vertex; |
sna->render_state.gen7.floats_per_vertex = op->floats_per_vertex; |
} |
} |
fastcall static void |
gen7_render_composite_blt(struct sna *sna, |
const struct sna_composite_op *op, |
const struct sna_composite_rectangles *r) |
{ |
gen7_get_rectangles(sna, op, 1, gen7_emit_composite_state); |
op->prim_emit(sna, op, r); |
} |
static uint32_t |
gen7_composite_create_blend_state(struct sna_static_stream *stream) |
{ |
char *base, *ptr; |
int src, dst; |
base = sna_static_stream_map(stream, |
GEN7_BLENDFACTOR_COUNT * GEN7_BLENDFACTOR_COUNT * GEN7_BLEND_STATE_PADDED_SIZE, |
64); |
ptr = base; |
for (src = 0; src < GEN7_BLENDFACTOR_COUNT; src++) { |
for (dst= 0; dst < GEN7_BLENDFACTOR_COUNT; dst++) { |
struct gen7_blend_state *blend = |
(struct gen7_blend_state *)ptr; |
blend->blend0.dest_blend_factor = dst; |
blend->blend0.source_blend_factor = src; |
blend->blend0.blend_func = GEN7_BLENDFUNCTION_ADD; |
blend->blend0.blend_enable = |
!(dst == GEN7_BLENDFACTOR_ZERO && src == GEN7_BLENDFACTOR_ONE); |
blend->blend1.post_blend_clamp_enable = 1; |
blend->blend1.pre_blend_clamp_enable = 1; |
ptr += GEN7_BLEND_STATE_PADDED_SIZE; |
} |
} |
return sna_static_stream_offsetof(stream, base); |
} |
#if 0 |
static uint32_t gen7_bind_video_source(struct sna *sna, |
struct kgem_bo *bo, |
uint32_t offset, |
int width, |
int height, |
int pitch, |
uint32_t format) |
{ |
uint32_t *ss, bind; |
bind = sna->kgem.surface -= |
sizeof(struct gen7_surface_state) / sizeof(uint32_t); |
assert(bo->tiling == I915_TILING_NONE); |
ss = sna->kgem.batch + bind; |
ss[0] = (GEN7_SURFACE_2D << GEN7_SURFACE_TYPE_SHIFT | |
format << GEN7_SURFACE_FORMAT_SHIFT); |
ss[1] = kgem_add_reloc(&sna->kgem, bind + 1, bo, |
I915_GEM_DOMAIN_SAMPLER << 16, |
offset); |
ss[2] = ((width - 1) << GEN7_SURFACE_WIDTH_SHIFT | |
(height - 1) << GEN7_SURFACE_HEIGHT_SHIFT); |
ss[3] = (pitch - 1) << GEN7_SURFACE_PITCH_SHIFT; |
ss[4] = 0; |
ss[5] = 0; |
ss[6] = 0; |
ss[7] = 0; |
if (is_hsw(sna)) |
ss[7] |= HSW_SURFACE_SWIZZLE(RED, GREEN, BLUE, ALPHA); |
DBG(("[%x] bind bo(handle=%d, addr=%d), format=%d, width=%d, height=%d, pitch=%d, offset=%d\n", |
bind, bo->handle, ss[1], |
format, width, height, pitch, offset)); |
return bind * sizeof(uint32_t); |
} |
static void gen7_emit_video_state(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
struct sna_video_frame *frame = op->priv; |
uint32_t src_surf_format; |
uint32_t src_surf_base[6]; |
int src_width[6]; |
int src_height[6]; |
int src_pitch[6]; |
uint32_t *binding_table; |
uint16_t offset; |
int n_src, n; |
gen7_get_batch(sna, op); |
src_surf_base[0] = 0; |
src_surf_base[1] = 0; |
src_surf_base[2] = frame->VBufOffset; |
src_surf_base[3] = frame->VBufOffset; |
src_surf_base[4] = frame->UBufOffset; |
src_surf_base[5] = frame->UBufOffset; |
if (is_planar_fourcc(frame->id)) { |
src_surf_format = GEN7_SURFACEFORMAT_R8_UNORM; |
src_width[1] = src_width[0] = frame->width; |
src_height[1] = src_height[0] = frame->height; |
src_pitch[1] = src_pitch[0] = frame->pitch[1]; |
src_width[4] = src_width[5] = src_width[2] = src_width[3] = |
frame->width / 2; |
src_height[4] = src_height[5] = src_height[2] = src_height[3] = |
frame->height / 2; |
src_pitch[4] = src_pitch[5] = src_pitch[2] = src_pitch[3] = |
frame->pitch[0]; |
n_src = 6; |
} else { |
if (frame->id == FOURCC_UYVY) |
src_surf_format = GEN7_SURFACEFORMAT_YCRCB_SWAPY; |
else |
src_surf_format = GEN7_SURFACEFORMAT_YCRCB_NORMAL; |
src_width[0] = frame->width; |
src_height[0] = frame->height; |
src_pitch[0] = frame->pitch[0]; |
n_src = 1; |
} |
binding_table = gen7_composite_get_binding_table(sna, &offset); |
binding_table[0] = |
gen7_bind_bo(sna, |
op->dst.bo, op->dst.width, op->dst.height, |
gen7_get_dest_format(op->dst.format), |
true); |
for (n = 0; n < n_src; n++) { |
binding_table[1+n] = |
gen7_bind_video_source(sna, |
frame->bo, |
src_surf_base[n], |
src_width[n], |
src_height[n], |
src_pitch[n], |
src_surf_format); |
} |
gen7_emit_state(sna, op, offset); |
} |
static bool |
gen7_render_video(struct sna *sna, |
struct sna_video *video, |
struct sna_video_frame *frame, |
RegionPtr dstRegion, |
PixmapPtr pixmap) |
{ |
struct sna_composite_op tmp; |
int dst_width = dstRegion->extents.x2 - dstRegion->extents.x1; |
int dst_height = dstRegion->extents.y2 - dstRegion->extents.y1; |
int src_width = frame->src.x2 - frame->src.x1; |
int src_height = frame->src.y2 - frame->src.y1; |
float src_offset_x, src_offset_y; |
float src_scale_x, src_scale_y; |
int nbox, pix_xoff, pix_yoff; |
struct sna_pixmap *priv; |
unsigned filter; |
BoxPtr box; |
DBG(("%s: src=(%d, %d), dst=(%d, %d), %ldx[(%d, %d), (%d, %d)...]\n", |
__FUNCTION__, |
src_width, src_height, dst_width, dst_height, |
(long)REGION_NUM_RECTS(dstRegion), |
REGION_EXTENTS(NULL, dstRegion)->x1, |
REGION_EXTENTS(NULL, dstRegion)->y1, |
REGION_EXTENTS(NULL, dstRegion)->x2, |
REGION_EXTENTS(NULL, dstRegion)->y2)); |
priv = sna_pixmap_force_to_gpu(pixmap, MOVE_READ | MOVE_WRITE); |
if (priv == NULL) |
return false; |
memset(&tmp, 0, sizeof(tmp)); |
tmp.dst.pixmap = pixmap; |
tmp.dst.width = pixmap->drawable.width; |
tmp.dst.height = pixmap->drawable.height; |
tmp.dst.format = sna_render_format_for_depth(pixmap->drawable.depth); |
tmp.dst.bo = priv->gpu_bo; |
tmp.src.bo = frame->bo; |
tmp.mask.bo = NULL; |
tmp.floats_per_vertex = 3; |
tmp.floats_per_rect = 9; |
if (src_width == dst_width && src_height == dst_height) |
filter = SAMPLER_FILTER_NEAREST; |
else |
filter = SAMPLER_FILTER_BILINEAR; |
tmp.u.gen7.flags = |
GEN7_SET_FLAGS(SAMPLER_OFFSET(filter, SAMPLER_EXTEND_PAD, |
SAMPLER_FILTER_NEAREST, SAMPLER_EXTEND_NONE), |
NO_BLEND, |
is_planar_fourcc(frame->id) ? |
GEN7_WM_KERNEL_VIDEO_PLANAR : |
GEN7_WM_KERNEL_VIDEO_PACKED, |
2); |
tmp.priv = frame; |
kgem_set_mode(&sna->kgem, KGEM_RENDER, tmp.dst.bo); |
if (!kgem_check_bo(&sna->kgem, tmp.dst.bo, frame->bo, NULL)) { |
kgem_submit(&sna->kgem); |
assert(kgem_check_bo(&sna->kgem, tmp.dst.bo, frame->bo, NULL)); |
_kgem_set_mode(&sna->kgem, KGEM_RENDER); |
} |
gen7_emit_video_state(sna, &tmp); |
gen7_align_vertex(sna, &tmp); |
/* Set up the offset for translating from the given region (in screen |
* coordinates) to the backing pixmap. |
*/ |
#ifdef COMPOSITE |
pix_xoff = -pixmap->screen_x + pixmap->drawable.x; |
pix_yoff = -pixmap->screen_y + pixmap->drawable.y; |
#else |
pix_xoff = 0; |
pix_yoff = 0; |
#endif |
DBG(("%s: src=(%d, %d)x(%d, %d); frame=(%dx%d), dst=(%dx%d)\n", |
__FUNCTION__, |
frame->src.x1, frame->src.y1, |
src_width, src_height, |
dst_width, dst_height, |
frame->width, frame->height)); |
src_scale_x = (float)src_width / dst_width / frame->width; |
src_offset_x = (float)frame->src.x1 / frame->width - dstRegion->extents.x1 * src_scale_x; |
src_scale_y = (float)src_height / dst_height / frame->height; |
src_offset_y = (float)frame->src.y1 / frame->height - dstRegion->extents.y1 * src_scale_y; |
DBG(("%s: scale=(%f, %f), offset=(%f, %f)\n", |
__FUNCTION__, |
src_scale_x, src_scale_y, |
src_offset_x, src_offset_y)); |
box = REGION_RECTS(dstRegion); |
nbox = REGION_NUM_RECTS(dstRegion); |
while (nbox--) { |
BoxRec r; |
DBG(("%s: dst=(%d, %d), (%d, %d) + (%d, %d); src=(%f, %f), (%f, %f)\n", |
__FUNCTION__, |
box->x1, box->y1, |
box->x2, box->y2, |
pix_xoff, pix_yoff, |
box->x1 * src_scale_x + src_offset_x, |
box->y1 * src_scale_y + src_offset_y, |
box->x2 * src_scale_x + src_offset_x, |
box->y2 * src_scale_y + src_offset_y)); |
r.x1 = box->x1 + pix_xoff; |
r.x2 = box->x2 + pix_xoff; |
r.y1 = box->y1 + pix_yoff; |
r.y2 = box->y2 + pix_yoff; |
gen7_get_rectangles(sna, &tmp, 1, gen7_emit_video_state); |
OUT_VERTEX(r.x2, r.y2); |
OUT_VERTEX_F(box->x2 * src_scale_x + src_offset_x); |
OUT_VERTEX_F(box->y2 * src_scale_y + src_offset_y); |
OUT_VERTEX(r.x1, r.y2); |
OUT_VERTEX_F(box->x1 * src_scale_x + src_offset_x); |
OUT_VERTEX_F(box->y2 * src_scale_y + src_offset_y); |
OUT_VERTEX(r.x1, r.y1); |
OUT_VERTEX_F(box->x1 * src_scale_x + src_offset_x); |
OUT_VERTEX_F(box->y1 * src_scale_y + src_offset_y); |
if (!DAMAGE_IS_ALL(priv->gpu_damage)) { |
sna_damage_add_box(&priv->gpu_damage, &r); |
sna_damage_subtract_box(&priv->cpu_damage, &r); |
} |
box++; |
} |
gen4_vertex_flush(sna); |
return true; |
} |
#endif |
static void gen7_render_composite_done(struct sna *sna, |
const struct sna_composite_op *op) |
{ |
if (sna->render.vertex_offset) { |
gen4_vertex_flush(sna); |
gen7_magic_ca_pass(sna, op); |
} |
} |
#if 0 |
static bool |
gen7_render_fill_boxes(struct sna *sna, |
CARD8 op, |
PictFormat format, |
const xRenderColor *color, |
PixmapPtr dst, struct kgem_bo *dst_bo, |
const BoxRec *box, int n) |
{ |
struct sna_composite_op tmp; |
uint32_t pixel; |
DBG(("%s (op=%d, color=(%04x, %04x, %04x, %04x) [%08x])\n", |
__FUNCTION__, op, |
color->red, color->green, color->blue, color->alpha, (int)format)); |
if (op >= ARRAY_SIZE(gen7_blend_op)) { |
DBG(("%s: fallback due to unhandled blend op: %d\n", |
__FUNCTION__, op)); |
return false; |
} |
if (prefer_blt_fill(sna, dst_bo) || !gen7_check_dst_format(format)) { |
uint8_t alu = GXinvalid; |
if (op <= PictOpSrc) { |
pixel = 0; |
if (op == PictOpClear) |
alu = GXclear; |
else if (sna_get_pixel_from_rgba(&pixel, |
color->red, |
color->green, |
color->blue, |
color->alpha, |
format)) |
alu = GXcopy; |
} |
if (alu != GXinvalid && |
sna_blt_fill_boxes(sna, alu, |
dst_bo, dst->drawable.bitsPerPixel, |
pixel, box, n)) |
return true; |
if (!gen7_check_dst_format(format)) |
return false; |
} |
if (op == PictOpClear) { |
pixel = 0; |
op = PictOpSrc; |
} else if (!sna_get_pixel_from_rgba(&pixel, |
color->red, |
color->green, |
color->blue, |
color->alpha, |
PICT_a8r8g8b8)) |
return false; |
DBG(("%s(%08x x %d [(%d, %d), (%d, %d) ...])\n", |
__FUNCTION__, pixel, n, |
box[0].x1, box[0].y1, box[0].x2, box[0].y2)); |
tmp.dst.pixmap = dst; |
tmp.dst.width = dst->drawable.width; |
tmp.dst.height = dst->drawable.height; |
tmp.dst.format = format; |
tmp.dst.bo = dst_bo; |
tmp.dst.x = tmp.dst.y = 0; |
tmp.damage = NULL; |
sna_render_composite_redirect_init(&tmp); |
if (too_large(dst->drawable.width, dst->drawable.height)) { |
BoxRec extents; |
boxes_extents(box, n, &extents); |
if (!sna_render_composite_redirect(sna, &tmp, |
extents.x1, extents.y1, |
extents.x2 - extents.x1, |
extents.y2 - extents.y1, |
n > 1)) |
return sna_tiling_fill_boxes(sna, op, format, color, |
dst, dst_bo, box, n); |
} |
tmp.src.bo = sna_render_get_solid(sna, pixel); |
tmp.mask.bo = NULL; |
tmp.floats_per_vertex = 2; |
tmp.floats_per_rect = 6; |
tmp.need_magic_ca_pass = false; |
tmp.u.gen7.flags = FILL_FLAGS(op, format); |
kgem_set_mode(&sna->kgem, KGEM_RENDER, dst_bo); |
if (!kgem_check_bo(&sna->kgem, dst_bo, NULL)) { |
kgem_submit(&sna->kgem); |
assert(kgem_check_bo(&sna->kgem, dst_bo, NULL)); |
} |
gen7_emit_fill_state(sna, &tmp); |
gen7_align_vertex(sna, &tmp); |
do { |
int n_this_time; |
int16_t *v; |
n_this_time = gen7_get_rectangles(sna, &tmp, n, |
gen7_emit_fill_state); |
n -= n_this_time; |
v = (int16_t *)(sna->render.vertices + sna->render.vertex_used); |
sna->render.vertex_used += 6 * n_this_time; |
assert(sna->render.vertex_used <= sna->render.vertex_size); |
do { |
DBG((" (%d, %d), (%d, %d)\n", |
box->x1, box->y1, box->x2, box->y2)); |
v[0] = box->x2; |
v[5] = v[1] = box->y2; |
v[8] = v[4] = box->x1; |
v[9] = box->y1; |
v[2] = v[3] = v[7] = 1; |
v[6] = v[10] = v[11] = 0; |
v += 12; box++; |
} while (--n_this_time); |
} while (n); |
gen4_vertex_flush(sna); |
kgem_bo_destroy(&sna->kgem, tmp.src.bo); |
sna_render_composite_redirect_done(sna, &tmp); |
return true; |
} |
#endif |
static void gen7_render_flush(struct sna *sna) |
{ |
gen4_vertex_close(sna); |
assert(sna->render.vb_id == 0); |
assert(sna->render.vertex_offset == 0); |
} |
static void |
gen7_render_context_switch(struct kgem *kgem, |
int new_mode) |
{ |
if (kgem->nbatch) { |
DBG(("%s: switch rings %d -> %d\n", |
__FUNCTION__, kgem->mode, new_mode)); |
_kgem_submit(kgem); |
} |
kgem->ring = new_mode; |
} |
static void |
gen7_render_retire(struct kgem *kgem) |
{ |
struct sna *sna; |
if (kgem->ring && (kgem->has_semaphores || !kgem->need_retire)) |
kgem->ring = kgem->mode; |
sna = container_of(kgem, struct sna, kgem); |
if (kgem->nbatch == 0 && sna->render.vbo && !kgem_bo_is_busy(sna->render.vbo)) { |
DBG(("%s: resetting idle vbo\n", __FUNCTION__)); |
sna->render.vertex_used = 0; |
sna->render.vertex_index = 0; |
} |
} |
static void |
gen7_render_expire(struct kgem *kgem) |
{ |
struct sna *sna; |
sna = container_of(kgem, struct sna, kgem); |
if (sna->render.vbo && !sna->render.vertex_used) { |
DBG(("%s: discarding vbo\n", __FUNCTION__)); |
kgem_bo_destroy(kgem, sna->render.vbo); |
sna->render.vbo = NULL; |
sna->render.vertices = sna->render.vertex_data; |
sna->render.vertex_size = ARRAY_SIZE(sna->render.vertex_data); |
sna->render.vertex_used = 0; |
sna->render.vertex_index = 0; |
} |
} |
static void gen7_render_reset(struct sna *sna) |
{ |
sna->render_state.gen7.emit_flush = false; |
sna->render_state.gen7.needs_invariant = true; |
sna->render_state.gen7.ve_id = 3 << 2; |
sna->render_state.gen7.last_primitive = -1; |
sna->render_state.gen7.num_sf_outputs = 0; |
sna->render_state.gen7.samplers = -1; |
sna->render_state.gen7.blend = -1; |
sna->render_state.gen7.kernel = -1; |
sna->render_state.gen7.drawrect_offset = -1; |
sna->render_state.gen7.drawrect_limit = -1; |
sna->render_state.gen7.surface_table = -1; |
sna->render.vertex_offset = 0; |
sna->render.nvertex_reloc = 0; |
sna->render.vb_id = 0; |
} |
static void gen7_render_fini(struct sna *sna) |
{ |
kgem_bo_destroy(&sna->kgem, sna->render_state.gen7.general_bo); |
} |
static bool is_gt3(struct sna *sna) |
{ |
assert(sna->kgem.gen == 075); |
return sna->PciInfo->device_id & 0x20; |
} |
static bool is_gt2(struct sna *sna) |
{ |
return sna->PciInfo->device_id & (is_hsw(sna)? 0x30 : 0x20); |
} |
static bool is_mobile(struct sna *sna) |
{ |
return (sna->PciInfo->device_id & 0xf) == 0x6; |
} |
static bool gen7_render_setup(struct sna *sna) |
{ |
struct gen7_render_state *state = &sna->render_state.gen7; |
struct sna_static_stream general; |
struct gen7_sampler_state *ss; |
int i, j, k, l, m; |
if (is_ivb(sna)) { |
state->info = &ivb_gt_info; |
if (sna->PciInfo->device_id & 0xf) { |
state->info = &ivb_gt1_info; |
if (is_gt2(sna)) |
state->info = &ivb_gt2_info; /* XXX requires GT_MODE WiZ disabled */ |
} |
} else if (is_byt(sna)) { |
state->info = &byt_gt_info; |
} else if (is_hsw(sna)) { |
state->info = &hsw_gt_info; |
if (sna->PciInfo->device_id & 0xf) { |
if (is_gt3(sna)) |
state->info = &hsw_gt3_info; |
else if (is_gt2(sna)) |
state->info = &hsw_gt2_info; |
else |
state->info = &hsw_gt1_info; |
} |
} else |
return false; |
sna_static_stream_init(&general); |
/* Zero pad the start. If you see an offset of 0x0 in the batchbuffer |
* dumps, you know it points to zero. |
*/ |
null_create(&general); |
for (m = 0; m < GEN7_WM_KERNEL_COUNT; m++) { |
if (wm_kernels[m].size) { |
state->wm_kernel[m][1] = |
sna_static_stream_add(&general, |
wm_kernels[m].data, |
wm_kernels[m].size, |
64); |
} else { |
if (USE_8_PIXEL_DISPATCH) { |
state->wm_kernel[m][0] = |
sna_static_stream_compile_wm(sna, &general, |
wm_kernels[m].data, 8); |
} |
if (USE_16_PIXEL_DISPATCH) { |
state->wm_kernel[m][1] = |
sna_static_stream_compile_wm(sna, &general, |
wm_kernels[m].data, 16); |
} |
if (USE_32_PIXEL_DISPATCH) { |
state->wm_kernel[m][2] = |
sna_static_stream_compile_wm(sna, &general, |
wm_kernels[m].data, 32); |
} |
} |
assert(state->wm_kernel[m][0]|state->wm_kernel[m][1]|state->wm_kernel[m][2]); |
} |
ss = sna_static_stream_map(&general, |
2 * sizeof(*ss) * |
(2 + |
FILTER_COUNT * EXTEND_COUNT * |
FILTER_COUNT * EXTEND_COUNT), |
32); |
state->wm_state = sna_static_stream_offsetof(&general, ss); |
sampler_copy_init(ss); ss += 2; |
sampler_fill_init(ss); ss += 2; |
for (i = 0; i < FILTER_COUNT; i++) { |
for (j = 0; j < EXTEND_COUNT; j++) { |
for (k = 0; k < FILTER_COUNT; k++) { |
for (l = 0; l < EXTEND_COUNT; l++) { |
sampler_state_init(ss++, i, j); |
sampler_state_init(ss++, k, l); |
} |
} |
} |
} |
state->cc_blend = gen7_composite_create_blend_state(&general); |
state->general_bo = sna_static_stream_fini(sna, &general); |
return state->general_bo != NULL; |
} |
const char *gen7_render_init(struct sna *sna, const char *backend) |
{ |
if (!gen7_render_setup(sna)) |
return backend; |
sna->kgem.context_switch = gen7_render_context_switch; |
sna->kgem.retire = gen7_render_retire; |
sna->kgem.expire = gen7_render_expire; |
#if 0 |
#if !NO_COMPOSITE |
sna->render.composite = gen7_render_composite; |
sna->render.prefer_gpu |= PREFER_GPU_RENDER; |
#endif |
#if !NO_COMPOSITE_SPANS |
sna->render.check_composite_spans = gen7_check_composite_spans; |
sna->render.composite_spans = gen7_render_composite_spans; |
if (is_mobile(sna) || is_gt2(sna) || is_byt(sna)) |
sna->render.prefer_gpu |= PREFER_GPU_SPANS; |
#endif |
sna->render.video = gen7_render_video; |
#if !NO_COPY_BOXES |
sna->render.copy_boxes = gen7_render_copy_boxes; |
#endif |
#if !NO_COPY |
sna->render.copy = gen7_render_copy; |
#endif |
#if !NO_FILL_BOXES |
sna->render.fill_boxes = gen7_render_fill_boxes; |
#endif |
#if !NO_FILL |
sna->render.fill = gen7_render_fill; |
#endif |
#if !NO_FILL_ONE |
sna->render.fill_one = gen7_render_fill_one; |
#endif |
#if !NO_FILL_CLEAR |
sna->render.clear = gen7_render_clear; |
#endif |
#endif |
sna->render.blit_tex = gen7_blit_tex; |
sna->render.caps = HW_BIT_BLIT | HW_TEX_BLIT; |
sna->render.flush = gen7_render_flush; |
sna->render.reset = gen7_render_reset; |
sna->render.fini = gen7_render_fini; |
sna->render.max_3d_size = GEN7_MAX_SIZE; |
sna->render.max_3d_pitch = 1 << 18; |
return sna->render_state.gen7.info->name; |
} |
static bool |
gen7_blit_tex(struct sna *sna, |
uint8_t op, bool scale, |
PixmapPtr src, struct kgem_bo *src_bo, |
PixmapPtr mask,struct kgem_bo *mask_bo, |
PixmapPtr dst, struct kgem_bo *dst_bo, |
int32_t src_x, int32_t src_y, |
int32_t msk_x, int32_t msk_y, |
int32_t dst_x, int32_t dst_y, |
int32_t width, int32_t height, |
struct sna_composite_op *tmp) |
{ |
tmp->op = PictOpSrc; |
tmp->dst.pixmap = dst; |
tmp->dst.bo = dst_bo; |
tmp->dst.width = dst->drawable.width; |
tmp->dst.height = dst->drawable.height; |
tmp->dst.format = PICT_a8r8g8b8; |
tmp->src.repeat = RepeatNone; |
tmp->src.filter = PictFilterNearest; |
tmp->src.is_affine = true; |
tmp->src.bo = src_bo; |
tmp->src.pict_format = PICT_x8r8g8b8; |
tmp->src.card_format = gen7_get_card_format(tmp->src.pict_format); |
tmp->src.width = src->drawable.width; |
tmp->src.height = src->drawable.height; |
tmp->is_affine = tmp->src.is_affine; |
tmp->has_component_alpha = false; |
tmp->need_magic_ca_pass = false; |
tmp->mask.repeat = SAMPLER_EXTEND_NONE; |
tmp->mask.filter = SAMPLER_FILTER_NEAREST; |
tmp->mask.is_affine = true; |
tmp->mask.bo = mask_bo; |
tmp->mask.pict_format = PIXMAN_a8; |
tmp->mask.card_format = gen7_get_card_format(tmp->mask.pict_format); |
tmp->mask.width = mask->drawable.width; |
tmp->mask.height = mask->drawable.height; |
if( scale ) |
{ |
tmp->src.scale[0] = 1.f/width; |
tmp->src.scale[1] = 1.f/height; |
} |
else |
{ |
tmp->src.scale[0] = 1.f/src->drawable.width; |
tmp->src.scale[1] = 1.f/src->drawable.height; |
} |
tmp->mask.scale[0] = 1.f/mask->drawable.width; |
tmp->mask.scale[1] = 1.f/mask->drawable.height; |
tmp->u.gen7.flags = |
GEN7_SET_FLAGS(SAMPLER_OFFSET(tmp->src.filter, |
tmp->src.repeat, |
tmp->mask.filter, |
tmp->mask.repeat), |
gen7_get_blend(tmp->op, |
tmp->has_component_alpha, |
tmp->dst.format), |
/* gen7_choose_composite_kernel(tmp->op, |
tmp->mask.bo != NULL, |
tmp->has_component_alpha, |
tmp->is_affine), */ |
GEN7_WM_KERNEL_MASK, |
gen4_choose_composite_emitter(sna, tmp)); |
tmp->blt = gen7_render_composite_blt; |
// tmp->box = gen7_render_composite_box; |
tmp->done = gen7_render_composite_done; |
kgem_set_mode(&sna->kgem, KGEM_RENDER, tmp->dst.bo); |
if (!kgem_check_bo(&sna->kgem, |
tmp->dst.bo, tmp->src.bo, tmp->mask.bo, |
NULL)) { |
kgem_submit(&sna->kgem); |
_kgem_set_mode(&sna->kgem, KGEM_RENDER); |
} |
gen7_emit_composite_state(sna, tmp); |
gen7_align_vertex(sna, tmp); |
return true; |
} |
/drivers/video/Intel-2D/sna/gen7_render.h |
---|
0,0 → 1,1366 |
#ifndef GEN7_RENDER_H |
#define GEN7_RENDER_H |
#define INTEL_MASK(high, low) (((1 << ((high) - (low) + 1)) - 1) << (low)) |
#define GEN7_3D(Pipeline,Opcode,Subopcode) ((3 << 29) | \ |
((Pipeline) << 27) | \ |
((Opcode) << 24) | \ |
((Subopcode) << 16)) |
#define GEN7_STATE_BASE_ADDRESS GEN7_3D(0, 1, 1) |
#define GEN7_STATE_SIP GEN7_3D(0, 1, 2) |
#define GEN7_PIPELINE_SELECT GEN7_3D(1, 1, 4) |
#define GEN7_MEDIA_STATE_POINTERS GEN7_3D(2, 0, 0) |
#define GEN7_MEDIA_OBJECT GEN7_3D(2, 1, 0) |
#define GEN7_3DSTATE_VERTEX_BUFFERS GEN7_3D(3, 0, 8) |
#define GEN7_3DSTATE_VERTEX_ELEMENTS GEN7_3D(3, 0, 9) |
#define GEN7_3DSTATE_INDEX_BUFFER GEN7_3D(3, 0, 0xa) |
#define GEN7_3DSTATE_VF_STATISTICS GEN7_3D(3, 0, 0xb) |
#define GEN7_3DSTATE_DRAWING_RECTANGLE GEN7_3D(3, 1, 0) |
#define GEN7_3DSTATE_CONSTANT_COLOR GEN7_3D(3, 1, 1) |
#define GEN7_3DSTATE_SAMPLER_PALETTE_LOAD GEN7_3D(3, 1, 2) |
#define GEN7_3DSTATE_CHROMA_KEY GEN7_3D(3, 1, 4) |
#define GEN7_3DSTATE_POLY_STIPPLE_OFFSET GEN7_3D(3, 1, 6) |
#define GEN7_3DSTATE_POLY_STIPPLE_PATTERN GEN7_3D(3, 1, 7) |
#define GEN7_3DSTATE_LINE_STIPPLE GEN7_3D(3, 1, 8) |
#define GEN7_3DSTATE_GLOBAL_DEPTH_OFFSET_CLAMP GEN7_3D(3, 1, 9) |
/* These two are BLC and CTG only, not BW or CL */ |
#define GEN7_3DSTATE_AA_LINE_PARAMS GEN7_3D(3, 1, 0xa) |
#define GEN7_3DSTATE_GS_SVB_INDEX GEN7_3D(3, 1, 0xb) |
#define GEN7_3DPRIMITIVE GEN7_3D(3, 3, 0) |
#define GEN7_3DSTATE_SAMPLER_STATE_POINTERS GEN7_3D(3, 0, 0x02) |
# define GEN7_3DSTATE_SAMPLER_STATE_MODIFY_PS (1 << 12) |
# define GEN7_3DSTATE_SAMPLER_STATE_MODIFY_GS (1 << 9) |
# define GEN7_3DSTATE_SAMPLER_STATE_MODIFY_VS (1 << 8) |
#define GEN7_3DSTATE_URB GEN7_3D(3, 0, 0x05) |
/* DW1 */ |
# define GEN7_3DSTATE_URB_VS_SIZE_SHIFT 16 |
# define GEN7_3DSTATE_URB_VS_ENTRIES_SHIFT 0 |
/* DW2 */ |
# define GEN7_3DSTATE_URB_GS_ENTRIES_SHIFT 8 |
# define GEN7_3DSTATE_URB_GS_SIZE_SHIFT 0 |
#define GEN7_3DSTATE_VIEWPORT_STATE_POINTERS GEN7_3D(3, 0, 0x0d) |
# define GEN7_3DSTATE_VIEWPORT_STATE_MODIFY_CC (1 << 12) |
# define GEN7_3DSTATE_VIEWPORT_STATE_MODIFY_SF (1 << 11) |
# define GEN7_3DSTATE_VIEWPORT_STATE_MODIFY_CLIP (1 << 10) |
#define GEN7_3DSTATE_CC_STATE_POINTERS GEN7_3D(3, 0, 0x0e) |
#define GEN7_3DSTATE_VS GEN7_3D(3, 0, 0x10) |
#define GEN7_3DSTATE_GS GEN7_3D(3, 0, 0x11) |
/* DW4 */ |
# define GEN7_3DSTATE_GS_DISPATCH_START_GRF_SHIFT 0 |
#define GEN7_3DSTATE_CLIP GEN7_3D(3, 0, 0x12) |
#define GEN7_3DSTATE_SF GEN7_3D(3, 0, 0x13) |
/* DW1 */ |
# define GEN7_3DSTATE_SF_NUM_OUTPUTS_SHIFT 22 |
# define GEN7_3DSTATE_SF_URB_ENTRY_READ_LENGTH_SHIFT 11 |
# define GEN7_3DSTATE_SF_URB_ENTRY_READ_OFFSET_SHIFT 4 |
/* DW2 */ |
/* DW3 */ |
# define GEN7_3DSTATE_SF_CULL_BOTH (0 << 29) |
# define GEN7_3DSTATE_SF_CULL_NONE (1 << 29) |
# define GEN7_3DSTATE_SF_CULL_FRONT (2 << 29) |
# define GEN7_3DSTATE_SF_CULL_BACK (3 << 29) |
/* DW4 */ |
# define GEN7_3DSTATE_SF_TRI_PROVOKE_SHIFT 29 |
# define GEN7_3DSTATE_SF_LINE_PROVOKE_SHIFT 27 |
# define GEN7_3DSTATE_SF_TRIFAN_PROVOKE_SHIFT 25 |
#define GEN7_3DSTATE_WM GEN7_3D(3, 0, 0x14) |
/* DW1 */ |
# define GEN7_WM_STATISTICS_ENABLE (1 << 31) |
# define GEN7_WM_DEPTH_CLEAR (1 << 30) |
# define GEN7_WM_DISPATCH_ENABLE (1 << 29) |
# define GEN7_WM_DEPTH_RESOLVE (1 << 28) |
# define GEN7_WM_HIERARCHICAL_DEPTH_RESOLVE (1 << 27) |
# define GEN7_WM_KILL_ENABLE (1 << 25) |
# define GEN7_WM_PSCDEPTH_OFF (0 << 23) |
# define GEN7_WM_PSCDEPTH_ON (1 << 23) |
# define GEN7_WM_PSCDEPTH_ON_GE (2 << 23) |
# define GEN7_WM_PSCDEPTH_ON_LE (3 << 23) |
# define GEN7_WM_USES_SOURCE_DEPTH (1 << 20) |
# define GEN7_WM_USES_SOURCE_W (1 << 19) |
# define GEN7_WM_POSITION_ZW_PIXEL (0 << 17) |
# define GEN7_WM_POSITION_ZW_CENTROID (2 << 17) |
# define GEN7_WM_POSITION_ZW_SAMPLE (3 << 17) |
# define GEN7_WM_NONPERSPECTIVE_SAMPLE_BARYCENTRIC (1 << 16) |
# define GEN7_WM_NONPERSPECTIVE_CENTROID_BARYCENTRIC (1 << 15) |
# define GEN7_WM_NONPERSPECTIVE_PIXEL_BARYCENTRIC (1 << 14) |
# define GEN7_WM_PERSPECTIVE_SAMPLE_BARYCENTRIC (1 << 13) |
# define GEN7_WM_PERSPECTIVE_CENTROID_BARYCENTRIC (1 << 12) |
# define GEN7_WM_PERSPECTIVE_PIXEL_BARYCENTRIC (1 << 11) |
# define GEN7_WM_USES_INPUT_COVERAGE_MASK (1 << 10) |
# define GEN7_WM_LINE_END_CAP_AA_WIDTH_0_5 (0 << 8) |
# define GEN7_WM_LINE_END_CAP_AA_WIDTH_1_0 (1 << 8) |
# define GEN7_WM_LINE_END_CAP_AA_WIDTH_2_0 (2 << 8) |
# define GEN7_WM_LINE_END_CAP_AA_WIDTH_4_0 (3 << 8) |
# define GEN7_WM_LINE_AA_WIDTH_0_5 (0 << 6) |
# define GEN7_WM_LINE_AA_WIDTH_1_0 (1 << 6) |
# define GEN7_WM_LINE_AA_WIDTH_2_0 (2 << 6) |
# define GEN7_WM_LINE_AA_WIDTH_4_0 (3 << 6) |
# define GEN7_WM_POLYGON_STIPPLE_ENABLE (1 << 4) |
# define GEN7_WM_LINE_STIPPLE_ENABLE (1 << 3) |
# define GEN7_WM_POINT_RASTRULE_UPPER_RIGHT (1 << 2) |
# define GEN7_WM_MSRAST_OFF_PIXEL (0 << 0) |
# define GEN7_WM_MSRAST_OFF_PATTERN (1 << 0) |
# define GEN7_WM_MSRAST_ON_PIXEL (2 << 0) |
# define GEN7_WM_MSRAST_ON_PATTERN (3 << 0) |
/* DW2 */ |
# define GEN7_WM_MSDISPMODE_PERPIXEL (1 << 31) |
#define GEN7_3DSTATE_CONSTANT_VS GEN7_3D(3, 0, 0x15) |
#define GEN7_3DSTATE_CONSTANT_GS GEN7_3D(3, 0, 0x16) |
#define GEN7_3DSTATE_CONSTANT_PS GEN7_3D(3, 0, 0x17) |
#define GEN7_3DSTATE_SAMPLE_MASK GEN7_3D(3, 0, 0x18) |
#define GEN7_3DSTATE_MULTISAMPLE GEN7_3D(3, 1, 0x0d) |
/* DW1 */ |
# define GEN7_3DSTATE_MULTISAMPLE_PIXEL_LOCATION_CENTER (0 << 4) |
# define GEN7_3DSTATE_MULTISAMPLE_PIXEL_LOCATION_UPPER_LEFT (1 << 4) |
# define GEN7_3DSTATE_MULTISAMPLE_NUMSAMPLES_1 (0 << 1) |
# define GEN7_3DSTATE_MULTISAMPLE_NUMSAMPLES_4 (2 << 1) |
# define GEN7_3DSTATE_MULTISAMPLE_NUMSAMPLES_8 (3 << 1) |
#define PIPELINE_SELECT_3D 0 |
#define PIPELINE_SELECT_MEDIA 1 |
/* for GEN7_STATE_BASE_ADDRESS */ |
#define BASE_ADDRESS_MODIFY (1 << 0) |
/* for GEN7_PIPE_CONTROL */ |
#define GEN7_PIPE_CONTROL GEN7_3D(3, 2, 0) |
#define GEN7_PIPE_CONTROL_CS_STALL (1 << 20) |
#define GEN7_PIPE_CONTROL_NOWRITE (0 << 14) |
#define GEN7_PIPE_CONTROL_WRITE_QWORD (1 << 14) |
#define GEN7_PIPE_CONTROL_WRITE_DEPTH (2 << 14) |
#define GEN7_PIPE_CONTROL_WRITE_TIME (3 << 14) |
#define GEN7_PIPE_CONTROL_DEPTH_STALL (1 << 13) |
#define GEN7_PIPE_CONTROL_WC_FLUSH (1 << 12) |
#define GEN7_PIPE_CONTROL_IS_FLUSH (1 << 11) |
#define GEN7_PIPE_CONTROL_TC_FLUSH (1 << 10) |
#define GEN7_PIPE_CONTROL_NOTIFY_ENABLE (1 << 8) |
#define GEN7_PIPE_CONTROL_GLOBAL_GTT (1 << 2) |
#define GEN7_PIPE_CONTROL_LOCAL_PGTT (0 << 2) |
#define GEN7_PIPE_CONTROL_STALL_AT_SCOREBOARD (1 << 1) |
#define GEN7_PIPE_CONTROL_DEPTH_CACHE_FLUSH (1 << 0) |
/* VERTEX_BUFFER_STATE Structure */ |
#define GEN7_VB0_BUFFER_INDEX_SHIFT 26 |
#define GEN7_VB0_VERTEXDATA (0 << 20) |
#define GEN7_VB0_INSTANCEDATA (1 << 20) |
#define GEN7_VB0_BUFFER_PITCH_SHIFT 0 |
#define GEN7_VB0_ADDRESS_MODIFY_ENABLE (1 << 14) |
/* VERTEX_ELEMENT_STATE Structure */ |
#define GEN7_VE0_VERTEX_BUFFER_INDEX_SHIFT 26 |
#define GEN7_VE0_VALID (1 << 25) |
#define GEN7_VE0_FORMAT_SHIFT 16 |
#define GEN7_VE0_OFFSET_SHIFT 0 |
#define GEN7_VE1_VFCOMPONENT_0_SHIFT 28 |
#define GEN7_VE1_VFCOMPONENT_1_SHIFT 24 |
#define GEN7_VE1_VFCOMPONENT_2_SHIFT 20 |
#define GEN7_VE1_VFCOMPONENT_3_SHIFT 16 |
#define GEN7_VE1_DESTINATION_ELEMENT_OFFSET_SHIFT 0 |
/* 3DPRIMITIVE bits */ |
#define GEN7_3DPRIMITIVE_VERTEX_SEQUENTIAL (0 << 15) |
#define GEN7_3DPRIMITIVE_VERTEX_RANDOM (1 << 15) |
#define GEN7_SVG_CTL 0x7400 |
#define GEN7_SVG_CTL_GS_BA (0 << 8) |
#define GEN7_SVG_CTL_SS_BA (1 << 8) |
#define GEN7_SVG_CTL_IO_BA (2 << 8) |
#define GEN7_SVG_CTL_GS_AUB (3 << 8) |
#define GEN7_SVG_CTL_IO_AUB (4 << 8) |
#define GEN7_SVG_CTL_SIP (5 << 8) |
#define GEN7_VF_CTL_SNAPSHOT_COMPLETE (1 << 31) |
#define GEN7_VF_CTL_SNAPSHOT_MUX_SELECT_THREADID (0 << 8) |
#define GEN7_VF_CTL_SNAPSHOT_MUX_SELECT_VF_DEBUG (1 << 8) |
#define GEN7_VF_CTL_SNAPSHOT_TYPE_VERTEX_SEQUENCE (0 << 4) |
#define GEN7_VF_CTL_SNAPSHOT_TYPE_VERTEX_INDEX (1 << 4) |
#define GEN7_VF_CTL_SKIP_INITIAL_PRIMITIVES (1 << 3) |
#define GEN7_VF_CTL_MAX_PRIMITIVES_LIMIT_ENABLE (1 << 2) |
#define GEN7_VF_CTL_VERTEX_RANGE_LIMIT_ENABLE (1 << 1) |
#define GEN7_VF_CTL_SNAPSHOT_ENABLE (1 << 0) |
#define GEN7_VF_STRG_VAL 0x7504 |
#define GEN7_VF_STR_VL_OVR 0x7508 |
#define GEN7_VF_VC_OVR 0x750c |
#define GEN7_VF_STR_PSKIP 0x7510 |
#define GEN7_VF_MAX_PRIM 0x7514 |
#define GEN7_VF_RDATA 0x7518 |
#define GEN7_VS_CTL 0x7600 |
#define GEN7_VS_CTL_SNAPSHOT_COMPLETE (1 << 31) |
#define GEN7_VS_CTL_SNAPSHOT_MUX_VERTEX_0 (0 << 8) |
#define GEN7_VS_CTL_SNAPSHOT_MUX_VERTEX_1 (1 << 8) |
#define GEN7_VS_CTL_SNAPSHOT_MUX_VALID_COUNT (2 << 8) |
#define GEN7_VS_CTL_SNAPSHOT_MUX_VS_KERNEL_POINTER (3 << 8) |
#define GEN7_VS_CTL_SNAPSHOT_ALL_THREADS (1 << 2) |
#define GEN7_VS_CTL_THREAD_SNAPSHOT_ENABLE (1 << 1) |
#define GEN7_VS_CTL_SNAPSHOT_ENABLE (1 << 0) |
#define GEN7_VS_STRG_VAL 0x7604 |
#define GEN7_VS_RDATA 0x7608 |
#define GEN7_SF_CTL 0x7b00 |
#define GEN7_SF_CTL_SNAPSHOT_COMPLETE (1 << 31) |
#define GEN7_SF_CTL_SNAPSHOT_MUX_VERTEX_0_FF_ID (0 << 8) |
#define GEN7_SF_CTL_SNAPSHOT_MUX_VERTEX_0_REL_COUNT (1 << 8) |
#define GEN7_SF_CTL_SNAPSHOT_MUX_VERTEX_1_FF_ID (2 << 8) |
#define GEN7_SF_CTL_SNAPSHOT_MUX_VERTEX_1_REL_COUNT (3 << 8) |
#define GEN7_SF_CTL_SNAPSHOT_MUX_VERTEX_2_FF_ID (4 << 8) |
#define GEN7_SF_CTL_SNAPSHOT_MUX_VERTEX_2_REL_COUNT (5 << 8) |
#define GEN7_SF_CTL_SNAPSHOT_MUX_VERTEX_COUNT (6 << 8) |
#define GEN7_SF_CTL_SNAPSHOT_MUX_SF_KERNEL_POINTER (7 << 8) |
#define GEN7_SF_CTL_MIN_MAX_PRIMITIVE_RANGE_ENABLE (1 << 4) |
#define GEN7_SF_CTL_DEBUG_CLIP_RECTANGLE_ENABLE (1 << 3) |
#define GEN7_SF_CTL_SNAPSHOT_ALL_THREADS (1 << 2) |
#define GEN7_SF_CTL_THREAD_SNAPSHOT_ENABLE (1 << 1) |
#define GEN7_SF_CTL_SNAPSHOT_ENABLE (1 << 0) |
#define GEN7_SF_STRG_VAL 0x7b04 |
#define GEN7_SF_RDATA 0x7b18 |
#define GEN7_WIZ_CTL 0x7c00 |
#define GEN7_WIZ_CTL_SNAPSHOT_COMPLETE (1 << 31) |
#define GEN7_WIZ_CTL_SUBSPAN_INSTANCE_SHIFT 16 |
#define GEN7_WIZ_CTL_SNAPSHOT_MUX_WIZ_KERNEL_POINTER (0 << 8) |
#define GEN7_WIZ_CTL_SNAPSHOT_MUX_SUBSPAN_INSTANCE (1 << 8) |
#define GEN7_WIZ_CTL_SNAPSHOT_MUX_PRIMITIVE_SEQUENCE (2 << 8) |
#define GEN7_WIZ_CTL_SINGLE_SUBSPAN_DISPATCH (1 << 6) |
#define GEN7_WIZ_CTL_IGNORE_COLOR_SCOREBOARD_STALLS (1 << 5) |
#define GEN7_WIZ_CTL_ENABLE_SUBSPAN_INSTANCE_COMPARE (1 << 4) |
#define GEN7_WIZ_CTL_USE_UPSTREAM_SNAPSHOT_FLAG (1 << 3) |
#define GEN7_WIZ_CTL_SNAPSHOT_ALL_THREADS (1 << 2) |
#define GEN7_WIZ_CTL_THREAD_SNAPSHOT_ENABLE (1 << 1) |
#define GEN7_WIZ_CTL_SNAPSHOT_ENABLE (1 << 0) |
#define GEN7_WIZ_STRG_VAL 0x7c04 |
#define GEN7_WIZ_RDATA 0x7c18 |
#define GEN7_TS_CTL 0x7e00 |
#define GEN7_TS_CTL_SNAPSHOT_COMPLETE (1 << 31) |
#define GEN7_TS_CTL_SNAPSHOT_MESSAGE_ERROR (0 << 8) |
#define GEN7_TS_CTL_SNAPSHOT_INTERFACE_DESCRIPTOR (3 << 8) |
#define GEN7_TS_CTL_SNAPSHOT_ALL_CHILD_THREADS (1 << 2) |
#define GEN7_TS_CTL_SNAPSHOT_ALL_ROOT_THREADS (1 << 1) |
#define GEN7_TS_CTL_SNAPSHOT_ENABLE (1 << 0) |
#define GEN7_TS_STRG_VAL 0x7e04 |
#define GEN7_TS_RDATA 0x7e08 |
#define GEN7_TD_CTL 0x8000 |
#define GEN7_TD_CTL_MUX_SHIFT 8 |
#define GEN7_TD_CTL_EXTERNAL_HALT_R0_DEBUG_MATCH (1 << 7) |
#define GEN7_TD_CTL_FORCE_EXTERNAL_HALT (1 << 6) |
#define GEN7_TD_CTL_EXCEPTION_MASK_OVERRIDE (1 << 5) |
#define GEN7_TD_CTL_FORCE_THREAD_BREAKPOINT_ENABLE (1 << 4) |
#define GEN7_TD_CTL_BREAKPOINT_ENABLE (1 << 2) |
#define GEN7_TD_CTL2 0x8004 |
#define GEN7_TD_CTL2_ILLEGAL_OPCODE_EXCEPTION_OVERRIDE (1 << 28) |
#define GEN7_TD_CTL2_MASKSTACK_EXCEPTION_OVERRIDE (1 << 26) |
#define GEN7_TD_CTL2_SOFTWARE_EXCEPTION_OVERRIDE (1 << 25) |
#define GEN7_TD_CTL2_ACTIVE_THREAD_LIMIT_SHIFT 16 |
#define GEN7_TD_CTL2_ACTIVE_THREAD_LIMIT_ENABLE (1 << 8) |
#define GEN7_TD_CTL2_THREAD_SPAWNER_EXECUTION_MASK_ENABLE (1 << 7) |
#define GEN7_TD_CTL2_WIZ_EXECUTION_MASK_ENABLE (1 << 6) |
#define GEN7_TD_CTL2_SF_EXECUTION_MASK_ENABLE (1 << 5) |
#define GEN7_TD_CTL2_CLIPPER_EXECUTION_MASK_ENABLE (1 << 4) |
#define GEN7_TD_CTL2_GS_EXECUTION_MASK_ENABLE (1 << 3) |
#define GEN7_TD_CTL2_VS_EXECUTION_MASK_ENABLE (1 << 0) |
#define GEN7_TD_VF_VS_EMSK 0x8008 |
#define GEN7_TD_GS_EMSK 0x800c |
#define GEN7_TD_CLIP_EMSK 0x8010 |
#define GEN7_TD_SF_EMSK 0x8014 |
#define GEN7_TD_WIZ_EMSK 0x8018 |
#define GEN7_TD_0_6_EHTRG_VAL 0x801c |
#define GEN7_TD_0_7_EHTRG_VAL 0x8020 |
#define GEN7_TD_0_6_EHTRG_MSK 0x8024 |
#define GEN7_TD_0_7_EHTRG_MSK 0x8028 |
#define GEN7_TD_RDATA 0x802c |
#define GEN7_TD_TS_EMSK 0x8030 |
#define GEN7_EU_CTL 0x8800 |
#define GEN7_EU_CTL_SELECT_SHIFT 16 |
#define GEN7_EU_CTL_DATA_MUX_SHIFT 8 |
#define GEN7_EU_ATT_0 0x8810 |
#define GEN7_EU_ATT_1 0x8814 |
#define GEN7_EU_ATT_DATA_0 0x8820 |
#define GEN7_EU_ATT_DATA_1 0x8824 |
#define GEN7_EU_ATT_CLR_0 0x8830 |
#define GEN7_EU_ATT_CLR_1 0x8834 |
#define GEN7_EU_RDATA 0x8840 |
#define _3DPRIM_POINTLIST 0x01 |
#define _3DPRIM_LINELIST 0x02 |
#define _3DPRIM_LINESTRIP 0x03 |
#define _3DPRIM_TRILIST 0x04 |
#define _3DPRIM_TRISTRIP 0x05 |
#define _3DPRIM_TRIFAN 0x06 |
#define _3DPRIM_QUADLIST 0x07 |
#define _3DPRIM_QUADSTRIP 0x08 |
#define _3DPRIM_LINELIST_ADJ 0x09 |
#define _3DPRIM_LINESTRIP_ADJ 0x0A |
#define _3DPRIM_TRILIST_ADJ 0x0B |
#define _3DPRIM_TRISTRIP_ADJ 0x0C |
#define _3DPRIM_TRISTRIP_REVERSE 0x0D |
#define _3DPRIM_POLYGON 0x0E |
#define _3DPRIM_RECTLIST 0x0F |
#define _3DPRIM_LINELOOP 0x10 |
#define _3DPRIM_POINTLIST_BF 0x11 |
#define _3DPRIM_LINESTRIP_CONT 0x12 |
#define _3DPRIM_LINESTRIP_BF 0x13 |
#define _3DPRIM_LINESTRIP_CONT_BF 0x14 |
#define _3DPRIM_TRIFAN_NOSTIPPLE 0x15 |
#define _3DPRIM_VERTEXBUFFER_ACCESS_SEQUENTIAL 0 |
#define _3DPRIM_VERTEXBUFFER_ACCESS_RANDOM 1 |
#define GEN7_ANISORATIO_2 0 |
#define GEN7_ANISORATIO_4 1 |
#define GEN7_ANISORATIO_6 2 |
#define GEN7_ANISORATIO_8 3 |
#define GEN7_ANISORATIO_10 4 |
#define GEN7_ANISORATIO_12 5 |
#define GEN7_ANISORATIO_14 6 |
#define GEN7_ANISORATIO_16 7 |
#define GEN7_BLENDFACTOR_ONE 0x1 |
#define GEN7_BLENDFACTOR_SRC_COLOR 0x2 |
#define GEN7_BLENDFACTOR_SRC_ALPHA 0x3 |
#define GEN7_BLENDFACTOR_DST_ALPHA 0x4 |
#define GEN7_BLENDFACTOR_DST_COLOR 0x5 |
#define GEN7_BLENDFACTOR_SRC_ALPHA_SATURATE 0x6 |
#define GEN7_BLENDFACTOR_CONST_COLOR 0x7 |
#define GEN7_BLENDFACTOR_CONST_ALPHA 0x8 |
#define GEN7_BLENDFACTOR_SRC1_COLOR 0x9 |
#define GEN7_BLENDFACTOR_SRC1_ALPHA 0x0A |
#define GEN7_BLENDFACTOR_ZERO 0x11 |
#define GEN7_BLENDFACTOR_INV_SRC_COLOR 0x12 |
#define GEN7_BLENDFACTOR_INV_SRC_ALPHA 0x13 |
#define GEN7_BLENDFACTOR_INV_DST_ALPHA 0x14 |
#define GEN7_BLENDFACTOR_INV_DST_COLOR 0x15 |
#define GEN7_BLENDFACTOR_INV_CONST_COLOR 0x17 |
#define GEN7_BLENDFACTOR_INV_CONST_ALPHA 0x18 |
#define GEN7_BLENDFACTOR_INV_SRC1_COLOR 0x19 |
#define GEN7_BLENDFACTOR_INV_SRC1_ALPHA 0x1A |
#define GEN7_BLENDFUNCTION_ADD 0 |
#define GEN7_BLENDFUNCTION_SUBTRACT 1 |
#define GEN7_BLENDFUNCTION_REVERSE_SUBTRACT 2 |
#define GEN7_BLENDFUNCTION_MIN 3 |
#define GEN7_BLENDFUNCTION_MAX 4 |
#define GEN7_ALPHATEST_FORMAT_UNORM8 0 |
#define GEN7_ALPHATEST_FORMAT_FLOAT32 1 |
#define GEN7_CHROMAKEY_KILL_ON_ANY_MATCH 0 |
#define GEN7_CHROMAKEY_REPLACE_BLACK 1 |
#define GEN7_CLIP_API_OGL 0 |
#define GEN7_CLIP_API_DX 1 |
#define GEN7_CLIPMODE_NORMAL 0 |
#define GEN7_CLIPMODE_CLIP_ALL 1 |
#define GEN7_CLIPMODE_CLIP_NON_REJECTED 2 |
#define GEN7_CLIPMODE_REJECT_ALL 3 |
#define GEN7_CLIPMODE_ACCEPT_ALL 4 |
#define GEN7_CLIP_NDCSPACE 0 |
#define GEN7_CLIP_SCREENSPACE 1 |
#define GEN7_COMPAREFUNCTION_ALWAYS 0 |
#define GEN7_COMPAREFUNCTION_NEVER 1 |
#define GEN7_COMPAREFUNCTION_LESS 2 |
#define GEN7_COMPAREFUNCTION_EQUAL 3 |
#define GEN7_COMPAREFUNCTION_LEQUAL 4 |
#define GEN7_COMPAREFUNCTION_GREATER 5 |
#define GEN7_COMPAREFUNCTION_NOTEQUAL 6 |
#define GEN7_COMPAREFUNCTION_GEQUAL 7 |
#define GEN7_COVERAGE_PIXELS_HALF 0 |
#define GEN7_COVERAGE_PIXELS_1 1 |
#define GEN7_COVERAGE_PIXELS_2 2 |
#define GEN7_COVERAGE_PIXELS_4 3 |
#define GEN7_CULLMODE_BOTH 0 |
#define GEN7_CULLMODE_NONE 1 |
#define GEN7_CULLMODE_FRONT 2 |
#define GEN7_CULLMODE_BACK 3 |
#define GEN7_DEFAULTCOLOR_R8G8B8A8_UNORM 0 |
#define GEN7_DEFAULTCOLOR_R32G32B32A32_FLOAT 1 |
#define GEN7_DEPTHFORMAT_D32_FLOAT_S8X24_UINT 0 |
#define GEN7_DEPTHFORMAT_D32_FLOAT 1 |
#define GEN7_DEPTHFORMAT_D24_UNORM_S8_UINT 2 |
#define GEN7_DEPTHFORMAT_D16_UNORM 5 |
#define GEN7_FLOATING_POINT_IEEE_754 0 |
#define GEN7_FLOATING_POINT_NON_IEEE_754 1 |
#define GEN7_FRONTWINDING_CW 0 |
#define GEN7_FRONTWINDING_CCW 1 |
#define GEN7_INDEX_BYTE 0 |
#define GEN7_INDEX_WORD 1 |
#define GEN7_INDEX_DWORD 2 |
#define GEN7_LOGICOPFUNCTION_CLEAR 0 |
#define GEN7_LOGICOPFUNCTION_NOR 1 |
#define GEN7_LOGICOPFUNCTION_AND_INVERTED 2 |
#define GEN7_LOGICOPFUNCTION_COPY_INVERTED 3 |
#define GEN7_LOGICOPFUNCTION_AND_REVERSE 4 |
#define GEN7_LOGICOPFUNCTION_INVERT 5 |
#define GEN7_LOGICOPFUNCTION_XOR 6 |
#define GEN7_LOGICOPFUNCTION_NAND 7 |
#define GEN7_LOGICOPFUNCTION_AND 8 |
#define GEN7_LOGICOPFUNCTION_EQUIV 9 |
#define GEN7_LOGICOPFUNCTION_NOOP 10 |
#define GEN7_LOGICOPFUNCTION_OR_INVERTED 11 |
#define GEN7_LOGICOPFUNCTION_COPY 12 |
#define GEN7_LOGICOPFUNCTION_OR_REVERSE 13 |
#define GEN7_LOGICOPFUNCTION_OR 14 |
#define GEN7_LOGICOPFUNCTION_SET 15 |
#define GEN7_MAPFILTER_NEAREST 0x0 |
#define GEN7_MAPFILTER_LINEAR 0x1 |
#define GEN7_MAPFILTER_ANISOTROPIC 0x2 |
#define GEN7_MIPFILTER_NONE 0 |
#define GEN7_MIPFILTER_NEAREST 1 |
#define GEN7_MIPFILTER_LINEAR 3 |
#define GEN7_POLYGON_FRONT_FACING 0 |
#define GEN7_POLYGON_BACK_FACING 1 |
#define GEN7_PREFILTER_ALWAYS 0x0 |
#define GEN7_PREFILTER_NEVER 0x1 |
#define GEN7_PREFILTER_LESS 0x2 |
#define GEN7_PREFILTER_EQUAL 0x3 |
#define GEN7_PREFILTER_LEQUAL 0x4 |
#define GEN7_PREFILTER_GREATER 0x5 |
#define GEN7_PREFILTER_NOTEQUAL 0x6 |
#define GEN7_PREFILTER_GEQUAL 0x7 |
#define GEN7_PROVOKING_VERTEX_0 0 |
#define GEN7_PROVOKING_VERTEX_1 1 |
#define GEN7_PROVOKING_VERTEX_2 2 |
#define GEN7_RASTRULE_UPPER_LEFT 0 |
#define GEN7_RASTRULE_UPPER_RIGHT 1 |
#define GEN7_RENDERTARGET_CLAMPRANGE_UNORM 0 |
#define GEN7_RENDERTARGET_CLAMPRANGE_SNORM 1 |
#define GEN7_RENDERTARGET_CLAMPRANGE_FORMAT 2 |
#define GEN7_STENCILOP_KEEP 0 |
#define GEN7_STENCILOP_ZERO 1 |
#define GEN7_STENCILOP_REPLACE 2 |
#define GEN7_STENCILOP_INCRSAT 3 |
#define GEN7_STENCILOP_DECRSAT 4 |
#define GEN7_STENCILOP_INCR 5 |
#define GEN7_STENCILOP_DECR 6 |
#define GEN7_STENCILOP_INVERT 7 |
#define GEN7_SURFACE_MIPMAPLAYOUT_BELOW 0 |
#define GEN7_SURFACE_MIPMAPLAYOUT_RIGHT 1 |
#define GEN7_SURFACEFORMAT_R32G32B32A32_FLOAT 0x000 |
#define GEN7_SURFACEFORMAT_R32G32B32A32_SINT 0x001 |
#define GEN7_SURFACEFORMAT_R32G32B32A32_UINT 0x002 |
#define GEN7_SURFACEFORMAT_R32G32B32A32_UNORM 0x003 |
#define GEN7_SURFACEFORMAT_R32G32B32A32_SNORM 0x004 |
#define GEN7_SURFACEFORMAT_R64G64_FLOAT 0x005 |
#define GEN7_SURFACEFORMAT_R32G32B32X32_FLOAT 0x006 |
#define GEN7_SURFACEFORMAT_R32G32B32A32_SSCALED 0x007 |
#define GEN7_SURFACEFORMAT_R32G32B32A32_USCALED 0x008 |
#define GEN7_SURFACEFORMAT_R32G32B32_FLOAT 0x040 |
#define GEN7_SURFACEFORMAT_R32G32B32_SINT 0x041 |
#define GEN7_SURFACEFORMAT_R32G32B32_UINT 0x042 |
#define GEN7_SURFACEFORMAT_R32G32B32_UNORM 0x043 |
#define GEN7_SURFACEFORMAT_R32G32B32_SNORM 0x044 |
#define GEN7_SURFACEFORMAT_R32G32B32_SSCALED 0x045 |
#define GEN7_SURFACEFORMAT_R32G32B32_USCALED 0x046 |
#define GEN7_SURFACEFORMAT_R16G16B16A16_UNORM 0x080 |
#define GEN7_SURFACEFORMAT_R16G16B16A16_SNORM 0x081 |
#define GEN7_SURFACEFORMAT_R16G16B16A16_SINT 0x082 |
#define GEN7_SURFACEFORMAT_R16G16B16A16_UINT 0x083 |
#define GEN7_SURFACEFORMAT_R16G16B16A16_FLOAT 0x084 |
#define GEN7_SURFACEFORMAT_R32G32_FLOAT 0x085 |
#define GEN7_SURFACEFORMAT_R32G32_SINT 0x086 |
#define GEN7_SURFACEFORMAT_R32G32_UINT 0x087 |
#define GEN7_SURFACEFORMAT_R32_FLOAT_X8X24_TYPELESS 0x088 |
#define GEN7_SURFACEFORMAT_X32_TYPELESS_G8X24_UINT 0x089 |
#define GEN7_SURFACEFORMAT_L32A32_FLOAT 0x08A |
#define GEN7_SURFACEFORMAT_R32G32_UNORM 0x08B |
#define GEN7_SURFACEFORMAT_R32G32_SNORM 0x08C |
#define GEN7_SURFACEFORMAT_R64_FLOAT 0x08D |
#define GEN7_SURFACEFORMAT_R16G16B16X16_UNORM 0x08E |
#define GEN7_SURFACEFORMAT_R16G16B16X16_FLOAT 0x08F |
#define GEN7_SURFACEFORMAT_A32X32_FLOAT 0x090 |
#define GEN7_SURFACEFORMAT_L32X32_FLOAT 0x091 |
#define GEN7_SURFACEFORMAT_I32X32_FLOAT 0x092 |
#define GEN7_SURFACEFORMAT_R16G16B16A16_SSCALED 0x093 |
#define GEN7_SURFACEFORMAT_R16G16B16A16_USCALED 0x094 |
#define GEN7_SURFACEFORMAT_R32G32_SSCALED 0x095 |
#define GEN7_SURFACEFORMAT_R32G32_USCALED 0x096 |
#define GEN7_SURFACEFORMAT_B8G8R8A8_UNORM 0x0C0 |
#define GEN7_SURFACEFORMAT_B8G8R8A8_UNORM_SRGB 0x0C1 |
#define GEN7_SURFACEFORMAT_R10G10B10A2_UNORM 0x0C2 |
#define GEN7_SURFACEFORMAT_R10G10B10A2_UNORM_SRGB 0x0C3 |
#define GEN7_SURFACEFORMAT_R10G10B10A2_UINT 0x0C4 |
#define GEN7_SURFACEFORMAT_R10G10B10_SNORM_A2_UNORM 0x0C5 |
#define GEN7_SURFACEFORMAT_R8G8B8A8_UNORM 0x0C7 |
#define GEN7_SURFACEFORMAT_R8G8B8A8_UNORM_SRGB 0x0C8 |
#define GEN7_SURFACEFORMAT_R8G8B8A8_SNORM 0x0C9 |
#define GEN7_SURFACEFORMAT_R8G8B8A8_SINT 0x0CA |
#define GEN7_SURFACEFORMAT_R8G8B8A8_UINT 0x0CB |
#define GEN7_SURFACEFORMAT_R16G16_UNORM 0x0CC |
#define GEN7_SURFACEFORMAT_R16G16_SNORM 0x0CD |
#define GEN7_SURFACEFORMAT_R16G16_SINT 0x0CE |
#define GEN7_SURFACEFORMAT_R16G16_UINT 0x0CF |
#define GEN7_SURFACEFORMAT_R16G16_FLOAT 0x0D0 |
#define GEN7_SURFACEFORMAT_B10G10R10A2_UNORM 0x0D1 |
#define GEN7_SURFACEFORMAT_B10G10R10A2_UNORM_SRGB 0x0D2 |
#define GEN7_SURFACEFORMAT_R11G11B10_FLOAT 0x0D3 |
#define GEN7_SURFACEFORMAT_R32_SINT 0x0D6 |
#define GEN7_SURFACEFORMAT_R32_UINT 0x0D7 |
#define GEN7_SURFACEFORMAT_R32_FLOAT 0x0D8 |
#define GEN7_SURFACEFORMAT_R24_UNORM_X8_TYPELESS 0x0D9 |
#define GEN7_SURFACEFORMAT_X24_TYPELESS_G8_UINT 0x0DA |
#define GEN7_SURFACEFORMAT_L16A16_UNORM 0x0DF |
#define GEN7_SURFACEFORMAT_I24X8_UNORM 0x0E0 |
#define GEN7_SURFACEFORMAT_L24X8_UNORM 0x0E1 |
#define GEN7_SURFACEFORMAT_A24X8_UNORM 0x0E2 |
#define GEN7_SURFACEFORMAT_I32_FLOAT 0x0E3 |
#define GEN7_SURFACEFORMAT_L32_FLOAT 0x0E4 |
#define GEN7_SURFACEFORMAT_A32_FLOAT 0x0E5 |
#define GEN7_SURFACEFORMAT_B8G8R8X8_UNORM 0x0E9 |
#define GEN7_SURFACEFORMAT_B8G8R8X8_UNORM_SRGB 0x0EA |
#define GEN7_SURFACEFORMAT_R8G8B8X8_UNORM 0x0EB |
#define GEN7_SURFACEFORMAT_R8G8B8X8_UNORM_SRGB 0x0EC |
#define GEN7_SURFACEFORMAT_R9G9B9E5_SHAREDEXP 0x0ED |
#define GEN7_SURFACEFORMAT_B10G10R10X2_UNORM 0x0EE |
#define GEN7_SURFACEFORMAT_L16A16_FLOAT 0x0F0 |
#define GEN7_SURFACEFORMAT_R32_UNORM 0x0F1 |
#define GEN7_SURFACEFORMAT_R32_SNORM 0x0F2 |
#define GEN7_SURFACEFORMAT_R10G10B10X2_USCALED 0x0F3 |
#define GEN7_SURFACEFORMAT_R8G8B8A8_SSCALED 0x0F4 |
#define GEN7_SURFACEFORMAT_R8G8B8A8_USCALED 0x0F5 |
#define GEN7_SURFACEFORMAT_R16G16_SSCALED 0x0F6 |
#define GEN7_SURFACEFORMAT_R16G16_USCALED 0x0F7 |
#define GEN7_SURFACEFORMAT_R32_SSCALED 0x0F8 |
#define GEN7_SURFACEFORMAT_R32_USCALED 0x0F9 |
#define GEN7_SURFACEFORMAT_B5G6R5_UNORM 0x100 |
#define GEN7_SURFACEFORMAT_B5G6R5_UNORM_SRGB 0x101 |
#define GEN7_SURFACEFORMAT_B5G5R5A1_UNORM 0x102 |
#define GEN7_SURFACEFORMAT_B5G5R5A1_UNORM_SRGB 0x103 |
#define GEN7_SURFACEFORMAT_B4G4R4A4_UNORM 0x104 |
#define GEN7_SURFACEFORMAT_B4G4R4A4_UNORM_SRGB 0x105 |
#define GEN7_SURFACEFORMAT_R8G8_UNORM 0x106 |
#define GEN7_SURFACEFORMAT_R8G8_SNORM 0x107 |
#define GEN7_SURFACEFORMAT_R8G8_SINT 0x108 |
#define GEN7_SURFACEFORMAT_R8G8_UINT 0x109 |
#define GEN7_SURFACEFORMAT_R16_UNORM 0x10A |
#define GEN7_SURFACEFORMAT_R16_SNORM 0x10B |
#define GEN7_SURFACEFORMAT_R16_SINT 0x10C |
#define GEN7_SURFACEFORMAT_R16_UINT 0x10D |
#define GEN7_SURFACEFORMAT_R16_FLOAT 0x10E |
#define GEN7_SURFACEFORMAT_I16_UNORM 0x111 |
#define GEN7_SURFACEFORMAT_L16_UNORM 0x112 |
#define GEN7_SURFACEFORMAT_A16_UNORM 0x113 |
#define GEN7_SURFACEFORMAT_L8A8_UNORM 0x114 |
#define GEN7_SURFACEFORMAT_I16_FLOAT 0x115 |
#define GEN7_SURFACEFORMAT_L16_FLOAT 0x116 |
#define GEN7_SURFACEFORMAT_A16_FLOAT 0x117 |
#define GEN7_SURFACEFORMAT_R5G5_SNORM_B6_UNORM 0x119 |
#define GEN7_SURFACEFORMAT_B5G5R5X1_UNORM 0x11A |
#define GEN7_SURFACEFORMAT_B5G5R5X1_UNORM_SRGB 0x11B |
#define GEN7_SURFACEFORMAT_R8G8_SSCALED 0x11C |
#define GEN7_SURFACEFORMAT_R8G8_USCALED 0x11D |
#define GEN7_SURFACEFORMAT_R16_SSCALED 0x11E |
#define GEN7_SURFACEFORMAT_R16_USCALED 0x11F |
#define GEN7_SURFACEFORMAT_R8_UNORM 0x140 |
#define GEN7_SURFACEFORMAT_R8_SNORM 0x141 |
#define GEN7_SURFACEFORMAT_R8_SINT 0x142 |
#define GEN7_SURFACEFORMAT_R8_UINT 0x143 |
#define GEN7_SURFACEFORMAT_A8_UNORM 0x144 |
#define GEN7_SURFACEFORMAT_I8_UNORM 0x145 |
#define GEN7_SURFACEFORMAT_L8_UNORM 0x146 |
#define GEN7_SURFACEFORMAT_P4A4_UNORM 0x147 |
#define GEN7_SURFACEFORMAT_A4P4_UNORM 0x148 |
#define GEN7_SURFACEFORMAT_R8_SSCALED 0x149 |
#define GEN7_SURFACEFORMAT_R8_USCALED 0x14A |
#define GEN7_SURFACEFORMAT_R1_UINT 0x181 |
#define GEN7_SURFACEFORMAT_YCRCB_NORMAL 0x182 |
#define GEN7_SURFACEFORMAT_YCRCB_SWAPUVY 0x183 |
#define GEN7_SURFACEFORMAT_BC1_UNORM 0x186 |
#define GEN7_SURFACEFORMAT_BC2_UNORM 0x187 |
#define GEN7_SURFACEFORMAT_BC3_UNORM 0x188 |
#define GEN7_SURFACEFORMAT_BC4_UNORM 0x189 |
#define GEN7_SURFACEFORMAT_BC5_UNORM 0x18A |
#define GEN7_SURFACEFORMAT_BC1_UNORM_SRGB 0x18B |
#define GEN7_SURFACEFORMAT_BC2_UNORM_SRGB 0x18C |
#define GEN7_SURFACEFORMAT_BC3_UNORM_SRGB 0x18D |
#define GEN7_SURFACEFORMAT_MONO8 0x18E |
#define GEN7_SURFACEFORMAT_YCRCB_SWAPUV 0x18F |
#define GEN7_SURFACEFORMAT_YCRCB_SWAPY 0x190 |
#define GEN7_SURFACEFORMAT_DXT1_RGB 0x191 |
#define GEN7_SURFACEFORMAT_FXT1 0x192 |
#define GEN7_SURFACEFORMAT_R8G8B8_UNORM 0x193 |
#define GEN7_SURFACEFORMAT_R8G8B8_SNORM 0x194 |
#define GEN7_SURFACEFORMAT_R8G8B8_SSCALED 0x195 |
#define GEN7_SURFACEFORMAT_R8G8B8_USCALED 0x196 |
#define GEN7_SURFACEFORMAT_R64G64B64A64_FLOAT 0x197 |
#define GEN7_SURFACEFORMAT_R64G64B64_FLOAT 0x198 |
#define GEN7_SURFACEFORMAT_BC4_SNORM 0x199 |
#define GEN7_SURFACEFORMAT_BC5_SNORM 0x19A |
#define GEN7_SURFACEFORMAT_R16G16B16_UNORM 0x19C |
#define GEN7_SURFACEFORMAT_R16G16B16_SNORM 0x19D |
#define GEN7_SURFACEFORMAT_R16G16B16_SSCALED 0x19E |
#define GEN7_SURFACEFORMAT_R16G16B16_USCALED 0x19F |
#define GEN7_SURFACERETURNFORMAT_FLOAT32 0 |
#define GEN7_SURFACERETURNFORMAT_S1 1 |
#define GEN7_SURFACE_1D 0 |
#define GEN7_SURFACE_2D 1 |
#define GEN7_SURFACE_3D 2 |
#define GEN7_SURFACE_CUBE 3 |
#define GEN7_SURFACE_BUFFER 4 |
#define GEN7_SURFACE_NULL 7 |
#define GEN7_BORDER_COLOR_MODE_DEFAULT 0 |
#define GEN7_BORDER_COLOR_MODE_LEGACY 1 |
#define GEN7_TEXCOORDMODE_WRAP 0 |
#define GEN7_TEXCOORDMODE_MIRROR 1 |
#define GEN7_TEXCOORDMODE_CLAMP 2 |
#define GEN7_TEXCOORDMODE_CUBE 3 |
#define GEN7_TEXCOORDMODE_CLAMP_BORDER 4 |
#define GEN7_TEXCOORDMODE_MIRROR_ONCE 5 |
#define GEN7_THREAD_PRIORITY_NORMAL 0 |
#define GEN7_THREAD_PRIORITY_HIGH 1 |
#define GEN7_TILEWALK_XMAJOR 0 |
#define GEN7_TILEWALK_YMAJOR 1 |
#define GEN7_VERTEX_SUBPIXEL_PRECISION_8BITS 0 |
#define GEN7_VERTEX_SUBPIXEL_PRECISION_4BITS 1 |
#define GEN7_VERTEXBUFFER_ACCESS_VERTEXDATA 0 |
#define GEN7_VERTEXBUFFER_ACCESS_INSTANCEDATA 1 |
#define GEN7_VFCOMPONENT_NOSTORE 0 |
#define GEN7_VFCOMPONENT_STORE_SRC 1 |
#define GEN7_VFCOMPONENT_STORE_0 2 |
#define GEN7_VFCOMPONENT_STORE_1_FLT 3 |
#define GEN7_VFCOMPONENT_STORE_1_INT 4 |
#define GEN7_VFCOMPONENT_STORE_VID 5 |
#define GEN7_VFCOMPONENT_STORE_IID 6 |
#define GEN7_VFCOMPONENT_STORE_PID 7 |
/* Execution Unit (EU) defines |
*/ |
#define GEN7_ALIGN_1 0 |
#define GEN7_ALIGN_16 1 |
#define GEN7_ADDRESS_DIRECT 0 |
#define GEN7_ADDRESS_REGISTER_INDIRECT_REGISTER 1 |
#define GEN7_CHANNEL_X 0 |
#define GEN7_CHANNEL_Y 1 |
#define GEN7_CHANNEL_Z 2 |
#define GEN7_CHANNEL_W 3 |
#define GEN7_COMPRESSION_NONE 0 |
#define GEN7_COMPRESSION_2NDHALF 1 |
#define GEN7_COMPRESSION_COMPRESSED 2 |
#define GEN7_CONDITIONAL_NONE 0 |
#define GEN7_CONDITIONAL_Z 1 |
#define GEN7_CONDITIONAL_NZ 2 |
#define GEN7_CONDITIONAL_EQ 1 /* Z */ |
#define GEN7_CONDITIONAL_NEQ 2 /* NZ */ |
#define GEN7_CONDITIONAL_G 3 |
#define GEN7_CONDITIONAL_GE 4 |
#define GEN7_CONDITIONAL_L 5 |
#define GEN7_CONDITIONAL_LE 6 |
#define GEN7_CONDITIONAL_C 7 |
#define GEN7_CONDITIONAL_O 8 |
#define GEN7_DEBUG_NONE 0 |
#define GEN7_DEBUG_BREAKPOINT 1 |
#define GEN7_DEPENDENCY_NORMAL 0 |
#define GEN7_DEPENDENCY_NOTCLEARED 1 |
#define GEN7_DEPENDENCY_NOTCHECKED 2 |
#define GEN7_DEPENDENCY_DISABLE 3 |
#define GEN7_EXECUTE_1 0 |
#define GEN7_EXECUTE_2 1 |
#define GEN7_EXECUTE_4 2 |
#define GEN7_EXECUTE_8 3 |
#define GEN7_EXECUTE_16 4 |
#define GEN7_EXECUTE_32 5 |
#define GEN7_HORIZONTAL_STRIDE_0 0 |
#define GEN7_HORIZONTAL_STRIDE_1 1 |
#define GEN7_HORIZONTAL_STRIDE_2 2 |
#define GEN7_HORIZONTAL_STRIDE_4 3 |
#define GEN7_INSTRUCTION_NORMAL 0 |
#define GEN7_INSTRUCTION_SATURATE 1 |
#define INTEL_MASK_ENABLE 0 |
#define INTEL_MASK_DISABLE 1 |
#define GEN7_OPCODE_MOV 1 |
#define GEN7_OPCODE_SEL 2 |
#define GEN7_OPCODE_NOT 4 |
#define GEN7_OPCODE_AND 5 |
#define GEN7_OPCODE_OR 6 |
#define GEN7_OPCODE_XOR 7 |
#define GEN7_OPCODE_SHR 8 |
#define GEN7_OPCODE_SHL 9 |
#define GEN7_OPCODE_RSR 10 |
#define GEN7_OPCODE_RSL 11 |
#define GEN7_OPCODE_ASR 12 |
#define GEN7_OPCODE_CMP 16 |
#define GEN7_OPCODE_JMPI 32 |
#define GEN7_OPCODE_IF 34 |
#define GEN7_OPCODE_IFF 35 |
#define GEN7_OPCODE_ELSE 36 |
#define GEN7_OPCODE_ENDIF 37 |
#define GEN7_OPCODE_DO 38 |
#define GEN7_OPCODE_WHILE 39 |
#define GEN7_OPCODE_BREAK 40 |
#define GEN7_OPCODE_CONTINUE 41 |
#define GEN7_OPCODE_HALT 42 |
#define GEN7_OPCODE_MSAVE 44 |
#define GEN7_OPCODE_MRESTORE 45 |
#define GEN7_OPCODE_PUSH 46 |
#define GEN7_OPCODE_POP 47 |
#define GEN7_OPCODE_WAIT 48 |
#define GEN7_OPCODE_SEND 49 |
#define GEN7_OPCODE_ADD 64 |
#define GEN7_OPCODE_MUL 65 |
#define GEN7_OPCODE_AVG 66 |
#define GEN7_OPCODE_FRC 67 |
#define GEN7_OPCODE_RNDU 68 |
#define GEN7_OPCODE_RNDD 69 |
#define GEN7_OPCODE_RNDE 70 |
#define GEN7_OPCODE_RNDZ 71 |
#define GEN7_OPCODE_MAC 72 |
#define GEN7_OPCODE_MACH 73 |
#define GEN7_OPCODE_LZD 74 |
#define GEN7_OPCODE_SAD2 80 |
#define GEN7_OPCODE_SADA2 81 |
#define GEN7_OPCODE_DP4 84 |
#define GEN7_OPCODE_DPH 85 |
#define GEN7_OPCODE_DP3 86 |
#define GEN7_OPCODE_DP2 87 |
#define GEN7_OPCODE_DPA2 88 |
#define GEN7_OPCODE_LINE 89 |
#define GEN7_OPCODE_NOP 126 |
#define GEN7_PREDICATE_NONE 0 |
#define GEN7_PREDICATE_NORMAL 1 |
#define GEN7_PREDICATE_ALIGN1_ANYV 2 |
#define GEN7_PREDICATE_ALIGN1_ALLV 3 |
#define GEN7_PREDICATE_ALIGN1_ANY2H 4 |
#define GEN7_PREDICATE_ALIGN1_ALL2H 5 |
#define GEN7_PREDICATE_ALIGN1_ANY4H 6 |
#define GEN7_PREDICATE_ALIGN1_ALL4H 7 |
#define GEN7_PREDICATE_ALIGN1_ANY8H 8 |
#define GEN7_PREDICATE_ALIGN1_ALL8H 9 |
#define GEN7_PREDICATE_ALIGN1_ANY16H 10 |
#define GEN7_PREDICATE_ALIGN1_ALL16H 11 |
#define GEN7_PREDICATE_ALIGN16_REPLICATE_X 2 |
#define GEN7_PREDICATE_ALIGN16_REPLICATE_Y 3 |
#define GEN7_PREDICATE_ALIGN16_REPLICATE_Z 4 |
#define GEN7_PREDICATE_ALIGN16_REPLICATE_W 5 |
#define GEN7_PREDICATE_ALIGN16_ANY4H 6 |
#define GEN7_PREDICATE_ALIGN16_ALL4H 7 |
#define GEN7_ARCHITECTURE_REGISTER_FILE 0 |
#define GEN7_GENERAL_REGISTER_FILE 1 |
#define GEN7_MESSAGE_REGISTER_FILE 2 |
#define GEN7_IMMEDIATE_VALUE 3 |
#define GEN7_REGISTER_TYPE_UD 0 |
#define GEN7_REGISTER_TYPE_D 1 |
#define GEN7_REGISTER_TYPE_UW 2 |
#define GEN7_REGISTER_TYPE_W 3 |
#define GEN7_REGISTER_TYPE_UB 4 |
#define GEN7_REGISTER_TYPE_B 5 |
#define GEN7_REGISTER_TYPE_VF 5 /* packed float vector, immediates only? */ |
#define GEN7_REGISTER_TYPE_HF 6 |
#define GEN7_REGISTER_TYPE_V 6 /* packed int vector, immediates only, uword dest only */ |
#define GEN7_REGISTER_TYPE_F 7 |
#define GEN7_ARF_NULL 0x00 |
#define GEN7_ARF_ADDRESS 0x10 |
#define GEN7_ARF_ACCUMULATOR 0x20 |
#define GEN7_ARF_FLAG 0x30 |
#define GEN7_ARF_MASK 0x40 |
#define GEN7_ARF_MASK_STACK 0x50 |
#define GEN7_ARF_MASK_STACK_DEPTH 0x60 |
#define GEN7_ARF_STATE 0x70 |
#define GEN7_ARF_CONTROL 0x80 |
#define GEN7_ARF_NOTIFICATION_COUNT 0x90 |
#define GEN7_ARF_IP 0xA0 |
#define GEN7_AMASK 0 |
#define GEN7_IMASK 1 |
#define GEN7_LMASK 2 |
#define GEN7_CMASK 3 |
#define GEN7_THREAD_NORMAL 0 |
#define GEN7_THREAD_ATOMIC 1 |
#define GEN7_THREAD_SWITCH 2 |
#define GEN7_VERTICAL_STRIDE_0 0 |
#define GEN7_VERTICAL_STRIDE_1 1 |
#define GEN7_VERTICAL_STRIDE_2 2 |
#define GEN7_VERTICAL_STRIDE_4 3 |
#define GEN7_VERTICAL_STRIDE_8 4 |
#define GEN7_VERTICAL_STRIDE_16 5 |
#define GEN7_VERTICAL_STRIDE_32 6 |
#define GEN7_VERTICAL_STRIDE_64 7 |
#define GEN7_VERTICAL_STRIDE_128 8 |
#define GEN7_VERTICAL_STRIDE_256 9 |
#define GEN7_VERTICAL_STRIDE_ONE_DIMENSIONAL 0xF |
#define GEN7_WIDTH_1 0 |
#define GEN7_WIDTH_2 1 |
#define GEN7_WIDTH_4 2 |
#define GEN7_WIDTH_8 3 |
#define GEN7_WIDTH_16 4 |
#define GEN7_STATELESS_BUFFER_BOUNDARY_1K 0 |
#define GEN7_STATELESS_BUFFER_BOUNDARY_2K 1 |
#define GEN7_STATELESS_BUFFER_BOUNDARY_4K 2 |
#define GEN7_STATELESS_BUFFER_BOUNDARY_8K 3 |
#define GEN7_STATELESS_BUFFER_BOUNDARY_16K 4 |
#define GEN7_STATELESS_BUFFER_BOUNDARY_32K 5 |
#define GEN7_STATELESS_BUFFER_BOUNDARY_64K 6 |
#define GEN7_STATELESS_BUFFER_BOUNDARY_128K 7 |
#define GEN7_STATELESS_BUFFER_BOUNDARY_256K 8 |
#define GEN7_STATELESS_BUFFER_BOUNDARY_512K 9 |
#define GEN7_STATELESS_BUFFER_BOUNDARY_1M 10 |
#define GEN7_STATELESS_BUFFER_BOUNDARY_2M 11 |
#define GEN7_POLYGON_FACING_FRONT 0 |
#define GEN7_POLYGON_FACING_BACK 1 |
#define GEN7_MESSAGE_TARGET_NULL 0 |
#define GEN7_MESSAGE_TARGET_MATH 1 |
#define GEN7_MESSAGE_TARGET_SAMPLER 2 |
#define GEN7_MESSAGE_TARGET_GATEWAY 3 |
#define GEN7_MESSAGE_TARGET_DATAPORT_READ 4 |
#define GEN7_MESSAGE_TARGET_DATAPORT_WRITE 5 |
#define GEN7_MESSAGE_TARGET_URB 6 |
#define GEN7_MESSAGE_TARGET_THREAD_SPAWNER 7 |
#define GEN7_SAMPLER_RETURN_FORMAT_FLOAT32 0 |
#define GEN7_SAMPLER_RETURN_FORMAT_UINT32 2 |
#define GEN7_SAMPLER_RETURN_FORMAT_SINT32 3 |
#define GEN7_SAMPLER_MESSAGE_SIMD8_SAMPLE 0 |
#define GEN7_SAMPLER_MESSAGE_SIMD16_SAMPLE 0 |
#define GEN7_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS 0 |
#define GEN7_SAMPLER_MESSAGE_SIMD8_KILLPIX 1 |
#define GEN7_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_LOD 1 |
#define GEN7_SAMPLER_MESSAGE_SIMD16_SAMPLE_LOD 1 |
#define GEN7_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_GRADIENTS 2 |
#define GEN7_SAMPLER_MESSAGE_SIMD8_SAMPLE_GRADIENTS 2 |
#define GEN7_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_COMPARE 0 |
#define GEN7_SAMPLER_MESSAGE_SIMD16_SAMPLE_COMPARE 2 |
#define GEN7_SAMPLER_MESSAGE_SIMD4X2_RESINFO 2 |
#define GEN7_SAMPLER_MESSAGE_SIMD8_RESINFO 2 |
#define GEN7_SAMPLER_MESSAGE_SIMD16_RESINFO 2 |
#define GEN7_SAMPLER_MESSAGE_SIMD4X2_LD 3 |
#define GEN7_SAMPLER_MESSAGE_SIMD8_LD 3 |
#define GEN7_SAMPLER_MESSAGE_SIMD16_LD 3 |
#define GEN7_DATAPORT_OWORD_BLOCK_1_OWORDLOW 0 |
#define GEN7_DATAPORT_OWORD_BLOCK_1_OWORDHIGH 1 |
#define GEN7_DATAPORT_OWORD_BLOCK_2_OWORDS 2 |
#define GEN7_DATAPORT_OWORD_BLOCK_4_OWORDS 3 |
#define GEN7_DATAPORT_OWORD_BLOCK_8_OWORDS 4 |
#define GEN7_DATAPORT_OWORD_DUAL_BLOCK_1OWORD 0 |
#define GEN7_DATAPORT_OWORD_DUAL_BLOCK_4OWORDS 2 |
#define GEN7_DATAPORT_DWORD_SCATTERED_BLOCK_8DWORDS 2 |
#define GEN7_DATAPORT_DWORD_SCATTERED_BLOCK_16DWORDS 3 |
#define GEN7_DATAPORT_READ_MESSAGE_OWORD_BLOCK_READ 0 |
#define GEN7_DATAPORT_READ_MESSAGE_OWORD_DUAL_BLOCK_READ 1 |
#define GEN7_DATAPORT_READ_MESSAGE_DWORD_BLOCK_READ 2 |
#define GEN7_DATAPORT_READ_MESSAGE_DWORD_SCATTERED_READ 3 |
#define GEN7_DATAPORT_READ_TARGET_DATA_CACHE 0 |
#define GEN7_DATAPORT_READ_TARGET_RENDER_CACHE 1 |
#define GEN7_DATAPORT_READ_TARGET_SAMPLER_CACHE 2 |
#define GEN7_DATAPORT_RENDER_TARGET_WRITE_SIMD16_SINGLE_SOURCE 0 |
#define GEN7_DATAPORT_RENDER_TARGET_WRITE_SIMD16_SINGLE_SOURCE_REPLICATED 1 |
#define GEN7_DATAPORT_RENDER_TARGET_WRITE_SIMD8_DUAL_SOURCE_SUBSPAN01 2 |
#define GEN7_DATAPORT_RENDER_TARGET_WRITE_SIMD8_DUAL_SOURCE_SUBSPAN23 3 |
#define GEN7_DATAPORT_RENDER_TARGET_WRITE_SIMD8_SINGLE_SOURCE_SUBSPAN01 4 |
#define GEN7_DATAPORT_WRITE_MESSAGE_OWORD_BLOCK_WRITE 0 |
#define GEN7_DATAPORT_WRITE_MESSAGE_OWORD_DUAL_BLOCK_WRITE 1 |
#define GEN7_DATAPORT_WRITE_MESSAGE_DWORD_BLOCK_WRITE 2 |
#define GEN7_DATAPORT_WRITE_MESSAGE_DWORD_SCATTERED_WRITE 3 |
#define GEN7_DATAPORT_WRITE_MESSAGE_RENDER_TARGET_WRITE 4 |
#define GEN7_DATAPORT_WRITE_MESSAGE_STREAMED_VERTEX_BUFFER_WRITE 5 |
#define GEN7_DATAPORT_WRITE_MESSAGE_FLUSH_RENDER_CACHE 7 |
#define GEN7_MATH_FUNCTION_INV 1 |
#define GEN7_MATH_FUNCTION_LOG 2 |
#define GEN7_MATH_FUNCTION_EXP 3 |
#define GEN7_MATH_FUNCTION_SQRT 4 |
#define GEN7_MATH_FUNCTION_RSQ 5 |
#define GEN7_MATH_FUNCTION_SIN 6 /* was 7 */ |
#define GEN7_MATH_FUNCTION_COS 7 /* was 8 */ |
#define GEN7_MATH_FUNCTION_SINCOS 8 /* was 6 */ |
#define GEN7_MATH_FUNCTION_TAN 9 |
#define GEN7_MATH_FUNCTION_POW 10 |
#define GEN7_MATH_FUNCTION_INT_DIV_QUOTIENT_AND_REMAINDER 11 |
#define GEN7_MATH_FUNCTION_INT_DIV_QUOTIENT 12 |
#define GEN7_MATH_FUNCTION_INT_DIV_REMAINDER 13 |
#define GEN7_MATH_INTEGER_UNSIGNED 0 |
#define GEN7_MATH_INTEGER_SIGNED 1 |
#define GEN7_MATH_PRECISION_FULL 0 |
#define GEN7_MATH_PRECISION_PARTIAL 1 |
#define GEN7_MATH_SATURATE_NONE 0 |
#define GEN7_MATH_SATURATE_SATURATE 1 |
#define GEN7_MATH_DATA_VECTOR 0 |
#define GEN7_MATH_DATA_SCALAR 1 |
#define GEN7_URB_OPCODE_WRITE 0 |
#define GEN7_URB_SWIZZLE_NONE 0 |
#define GEN7_URB_SWIZZLE_INTERLEAVE 1 |
#define GEN7_URB_SWIZZLE_TRANSPOSE 2 |
#define GEN7_SCRATCH_SPACE_SIZE_1K 0 |
#define GEN7_SCRATCH_SPACE_SIZE_2K 1 |
#define GEN7_SCRATCH_SPACE_SIZE_4K 2 |
#define GEN7_SCRATCH_SPACE_SIZE_8K 3 |
#define GEN7_SCRATCH_SPACE_SIZE_16K 4 |
#define GEN7_SCRATCH_SPACE_SIZE_32K 5 |
#define GEN7_SCRATCH_SPACE_SIZE_64K 6 |
#define GEN7_SCRATCH_SPACE_SIZE_128K 7 |
#define GEN7_SCRATCH_SPACE_SIZE_256K 8 |
#define GEN7_SCRATCH_SPACE_SIZE_512K 9 |
#define GEN7_SCRATCH_SPACE_SIZE_1M 10 |
#define GEN7_SCRATCH_SPACE_SIZE_2M 11 |
/* The hardware supports two different modes for border color. The |
* default (OpenGL) mode uses floating-point color channels, while the |
* legacy mode uses 4 bytes. |
* |
* More significantly, the legacy mode respects the components of the |
* border color for channels not present in the source, (whereas the |
* default mode will ignore the border color's alpha channel and use |
* alpha==1 for an RGB source, for example). |
* |
* The legacy mode matches the semantics specified by the Render |
* extension. |
*/ |
struct gen7_sampler_default_border_color { |
float color[4]; |
}; |
struct gen7_sampler_legacy_border_color { |
uint8_t color[4]; |
}; |
struct gen7_blend_state { |
struct { |
uint32_t dest_blend_factor:5; |
uint32_t source_blend_factor:5; |
uint32_t pad3:1; |
uint32_t blend_func:3; |
uint32_t pad2:1; |
uint32_t ia_dest_blend_factor:5; |
uint32_t ia_source_blend_factor:5; |
uint32_t pad1:1; |
uint32_t ia_blend_func:3; |
uint32_t pad0:1; |
uint32_t ia_blend_enable:1; |
uint32_t blend_enable:1; |
} blend0; |
struct { |
uint32_t post_blend_clamp_enable:1; |
uint32_t pre_blend_clamp_enable:1; |
uint32_t clamp_range:2; |
uint32_t pad0:4; |
uint32_t x_dither_offset:2; |
uint32_t y_dither_offset:2; |
uint32_t dither_enable:1; |
uint32_t alpha_test_func:3; |
uint32_t alpha_test_enable:1; |
uint32_t pad1:1; |
uint32_t logic_op_func:4; |
uint32_t logic_op_enable:1; |
uint32_t pad2:1; |
uint32_t write_disable_b:1; |
uint32_t write_disable_g:1; |
uint32_t write_disable_r:1; |
uint32_t write_disable_a:1; |
uint32_t pad3:1; |
uint32_t alpha_to_coverage_dither:1; |
uint32_t alpha_to_one:1; |
uint32_t alpha_to_coverage:1; |
} blend1; |
}; |
struct gen7_color_calc_state { |
struct { |
uint32_t alpha_test_format:1; |
uint32_t pad0:14; |
uint32_t round_disable:1; |
uint32_t bf_stencil_ref:8; |
uint32_t stencil_ref:8; |
} cc0; |
union { |
float alpha_ref_f; |
struct { |
uint32_t ui:8; |
uint32_t pad0:24; |
} alpha_ref_fi; |
} cc1; |
float constant_r; |
float constant_g; |
float constant_b; |
float constant_a; |
}; |
struct gen7_depth_stencil_state { |
struct { |
uint32_t pad0:3; |
uint32_t bf_stencil_pass_depth_pass_op:3; |
uint32_t bf_stencil_pass_depth_fail_op:3; |
uint32_t bf_stencil_fail_op:3; |
uint32_t bf_stencil_func:3; |
uint32_t bf_stencil_enable:1; |
uint32_t pad1:2; |
uint32_t stencil_write_enable:1; |
uint32_t stencil_pass_depth_pass_op:3; |
uint32_t stencil_pass_depth_fail_op:3; |
uint32_t stencil_fail_op:3; |
uint32_t stencil_func:3; |
uint32_t stencil_enable:1; |
} ds0; |
struct { |
uint32_t bf_stencil_write_mask:8; |
uint32_t bf_stencil_test_mask:8; |
uint32_t stencil_write_mask:8; |
uint32_t stencil_test_mask:8; |
} ds1; |
struct { |
uint32_t pad0:26; |
uint32_t depth_write_enable:1; |
uint32_t depth_test_func:3; |
uint32_t pad1:1; |
uint32_t depth_test_enable:1; |
} ds2; |
}; |
struct gen7_surface_state { |
struct { |
unsigned int cube_pos_z:1; |
unsigned int cube_neg_z:1; |
unsigned int cube_pos_y:1; |
unsigned int cube_neg_y:1; |
unsigned int cube_pos_x:1; |
unsigned int cube_neg_x:1; |
unsigned int pad2:2; |
unsigned int render_cache_read_write:1; |
unsigned int pad1:1; |
unsigned int surface_array_spacing:1; |
unsigned int vert_line_stride_ofs:1; |
unsigned int vert_line_stride:1; |
unsigned int tile_walk:1; |
unsigned int tiled_surface:1; |
unsigned int horizontal_alignment:1; |
unsigned int vertical_alignment:2; |
unsigned int surface_format:9; /**< BRW_SURFACEFORMAT_x */ |
unsigned int pad0:1; |
unsigned int is_array:1; |
unsigned int surface_type:3; /**< BRW_SURFACE_1D/2D/3D/CUBE */ |
} ss0; |
struct { |
unsigned int base_addr; |
} ss1; |
struct { |
unsigned int width:14; |
unsigned int pad1:2; |
unsigned int height:14; |
unsigned int pad0:2; |
} ss2; |
struct { |
unsigned int pitch:18; |
unsigned int pad:3; |
unsigned int depth:11; |
} ss3; |
struct { |
unsigned int multisample_position_palette_index:3; |
unsigned int num_multisamples:3; |
unsigned int multisampled_surface_storage_format:1; |
unsigned int render_target_view_extent:11; |
unsigned int min_array_elt:11; |
unsigned int rotation:2; |
unsigned int pad0:1; |
} ss4; |
struct { |
unsigned int mip_count:4; |
unsigned int min_lod:4; |
unsigned int pad1:12; |
unsigned int y_offset:4; |
unsigned int pad0:1; |
unsigned int x_offset:7; |
} ss5; |
struct { |
unsigned int pad; /* Multisample Control Surface stuff */ |
} ss6; |
struct { |
unsigned int resource_min_lod:12; |
unsigned int pad0:16; |
unsigned int alpha_clear_color:1; |
unsigned int blue_clear_color:1; |
unsigned int green_clear_color:1; |
unsigned int red_clear_color:1; |
} ss7; |
}; |
struct gen7_sampler_state { |
struct { |
unsigned int aniso_algorithm:1; |
unsigned int lod_bias:13; |
unsigned int min_filter:3; |
unsigned int mag_filter:3; |
unsigned int mip_filter:2; |
unsigned int base_level:5; |
unsigned int pad1:1; |
unsigned int lod_preclamp:1; |
unsigned int default_color_mode:1; |
unsigned int pad0:1; |
unsigned int disable:1; |
} ss0; |
struct { |
unsigned int cube_control_mode:1; |
unsigned int shadow_function:3; |
unsigned int pad:4; |
unsigned int max_lod:12; |
unsigned int min_lod:12; |
} ss1; |
struct { |
unsigned int pad:5; |
unsigned int default_color_pointer:27; |
} ss2; |
struct { |
unsigned int r_wrap_mode:3; |
unsigned int t_wrap_mode:3; |
unsigned int s_wrap_mode:3; |
unsigned int pad:1; |
unsigned int non_normalized_coord:1; |
unsigned int trilinear_quality:2; |
unsigned int address_round:6; |
unsigned int max_aniso:3; |
unsigned int chroma_key_mode:1; |
unsigned int chroma_key_index:2; |
unsigned int chroma_key_enable:1; |
unsigned int pad0:6; |
} ss3; |
}; |
/* Surface state DW0 */ |
#define GEN7_SURFACE_RC_READ_WRITE (1 << 8) |
#define GEN7_SURFACE_VALIGN_4 (1 << 16) |
#define GEN7_SURFACE_HALIGN_8 (1 << 15) |
#define GEN7_SURFACE_TILED (1 << 14) |
#define GEN7_SURFACE_TILED_Y (1 << 13) |
#define GEN7_SURFACE_FORMAT_SHIFT 18 |
#define GEN7_SURFACE_TYPE_SHIFT 29 |
/* Surface state DW2 */ |
#define GEN7_SURFACE_HEIGHT_SHIFT 16 |
#define GEN7_SURFACE_WIDTH_SHIFT 0 |
/* Surface state DW3 */ |
#define GEN7_SURFACE_DEPTH_SHIFT 21 |
#define GEN7_SURFACE_PITCH_SHIFT 0 |
#define HSW_SWIZZLE_ZERO 0 |
#define HSW_SWIZZLE_ONE 1 |
#define HSW_SWIZZLE_RED 4 |
#define HSW_SWIZZLE_GREEN 5 |
#define HSW_SWIZZLE_BLUE 6 |
#define HSW_SWIZZLE_ALPHA 7 |
#define __HSW_SURFACE_SWIZZLE(r,g,b,a) \ |
((a) << 16 | (b) << 19 | (g) << 22 | (r) << 25) |
#define HSW_SURFACE_SWIZZLE(r,g,b,a) \ |
__HSW_SURFACE_SWIZZLE(HSW_SWIZZLE_##r, HSW_SWIZZLE_##g, HSW_SWIZZLE_##b, HSW_SWIZZLE_##a) |
/* _3DSTATE_VERTEX_BUFFERS on GEN7*/ |
/* DW1 */ |
#define GEN7_VB0_ADDRESS_MODIFYENABLE (1 << 14) |
/* _3DPRIMITIVE on GEN7 */ |
/* DW1 */ |
# define GEN7_3DPRIM_VERTEXBUFFER_ACCESS_SEQUENTIAL (0 << 8) |
# define GEN7_3DPRIM_VERTEXBUFFER_ACCESS_RANDOM (1 << 8) |
#define GEN7_3DSTATE_CLEAR_PARAMS GEN7_3D(3, 0, 0x04) |
#define GEN7_3DSTATE_DEPTH_BUFFER GEN7_3D(3, 0, 0x05) |
# define GEN7_3DSTATE_DEPTH_BUFFER_TYPE_SHIFT 29 |
# define GEN7_3DSTATE_DEPTH_BUFFER_FORMAT_SHIFT 18 |
/* DW1 */ |
# define GEN7_3DSTATE_DEPTH_CLEAR_VALID (1 << 15) |
#define GEN7_3DSTATE_CONSTANT_HS GEN7_3D(3, 0, 0x19) |
#define GEN7_3DSTATE_CONSTANT_DS GEN7_3D(3, 0, 0x1a) |
#define GEN7_3DSTATE_HS GEN7_3D(3, 0, 0x1b) |
#define GEN7_3DSTATE_TE GEN7_3D(3, 0, 0x1c) |
#define GEN7_3DSTATE_DS GEN7_3D(3, 0, 0x1d) |
#define GEN7_3DSTATE_STREAMOUT GEN7_3D(3, 0, 0x1e) |
#define GEN7_3DSTATE_SBE GEN7_3D(3, 0, 0x1f) |
/* DW1 */ |
# define GEN7_SBE_SWIZZLE_CONTROL_MODE (1 << 28) |
# define GEN7_SBE_NUM_OUTPUTS_SHIFT 22 |
# define GEN7_SBE_SWIZZLE_ENABLE (1 << 21) |
# define GEN7_SBE_POINT_SPRITE_LOWERLEFT (1 << 20) |
# define GEN7_SBE_URB_ENTRY_READ_LENGTH_SHIFT 11 |
# define GEN7_SBE_URB_ENTRY_READ_OFFSET_SHIFT 4 |
#define GEN7_3DSTATE_PS GEN7_3D(3, 0, 0x20) |
/* DW1: kernel pointer */ |
/* DW2 */ |
# define GEN7_PS_SPF_MODE (1 << 31) |
# define GEN7_PS_VECTOR_MASK_ENABLE (1 << 30) |
# define GEN7_PS_SAMPLER_COUNT_SHIFT 27 |
# define GEN7_PS_BINDING_TABLE_ENTRY_COUNT_SHIFT 18 |
# define GEN7_PS_FLOATING_POINT_MODE_IEEE_754 (0 << 16) |
# define GEN7_PS_FLOATING_POINT_MODE_ALT (1 << 16) |
/* DW3: scratch space */ |
/* DW4 */ |
# define IVB_PS_MAX_THREADS_SHIFT 24 |
# define HSW_PS_MAX_THREADS_SHIFT 23 |
# define HSW_PS_SAMPLE_MASK_SHIFT 12 |
# define GEN7_PS_PUSH_CONSTANT_ENABLE (1 << 11) |
# define GEN7_PS_ATTRIBUTE_ENABLE (1 << 10) |
# define GEN7_PS_OMASK_TO_RENDER_TARGET (1 << 9) |
# define GEN7_PS_DUAL_SOURCE_BLEND_ENABLE (1 << 7) |
# define GEN7_PS_POSOFFSET_NONE (0 << 3) |
# define GEN7_PS_POSOFFSET_CENTROID (2 << 3) |
# define GEN7_PS_POSOFFSET_SAMPLE (3 << 3) |
# define GEN7_PS_32_DISPATCH_ENABLE (1 << 2) |
# define GEN7_PS_16_DISPATCH_ENABLE (1 << 1) |
# define GEN7_PS_8_DISPATCH_ENABLE (1 << 0) |
/* DW5 */ |
# define GEN7_PS_DISPATCH_START_GRF_SHIFT_0 16 |
# define GEN7_PS_DISPATCH_START_GRF_SHIFT_1 8 |
# define GEN7_PS_DISPATCH_START_GRF_SHIFT_2 0 |
/* DW6: kernel 1 pointer */ |
/* DW7: kernel 2 pointer */ |
#define GEN7_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CL GEN7_3D(3, 0, 0x21) |
#define GEN7_3DSTATE_VIEWPORT_STATE_POINTERS_CC GEN7_3D(3, 0, 0x23) |
#define GEN7_3DSTATE_BLEND_STATE_POINTERS GEN7_3D(3, 0, 0x24) |
#define GEN7_3DSTATE_DEPTH_STENCIL_STATE_POINTERS GEN7_3D(3, 0, 0x25) |
#define GEN7_3DSTATE_BINDING_TABLE_POINTERS_VS GEN7_3D(3, 0, 0x26) |
#define GEN7_3DSTATE_BINDING_TABLE_POINTERS_HS GEN7_3D(3, 0, 0x27) |
#define GEN7_3DSTATE_BINDING_TABLE_POINTERS_DS GEN7_3D(3, 0, 0x28) |
#define GEN7_3DSTATE_BINDING_TABLE_POINTERS_GS GEN7_3D(3, 0, 0x29) |
#define GEN7_3DSTATE_BINDING_TABLE_POINTERS_PS GEN7_3D(3, 0, 0x2a) |
#define GEN7_3DSTATE_SAMPLER_STATE_POINTERS_VS GEN7_3D(3, 0, 0x2b) |
#define GEN7_3DSTATE_SAMPLER_STATE_POINTERS_GS GEN7_3D(3, 0, 0x2e) |
#define GEN7_3DSTATE_SAMPLER_STATE_POINTERS_PS GEN7_3D(3, 0, 0x2f) |
#define GEN7_3DSTATE_URB_VS GEN7_3D(3, 0, 0x30) |
#define GEN7_3DSTATE_URB_HS GEN7_3D(3, 0, 0x31) |
#define GEN7_3DSTATE_URB_DS GEN7_3D(3, 0, 0x32) |
#define GEN7_3DSTATE_URB_GS GEN7_3D(3, 0, 0x33) |
/* DW1 */ |
# define GEN7_URB_ENTRY_NUMBER_SHIFT 0 |
# define GEN7_URB_ENTRY_SIZE_SHIFT 16 |
# define GEN7_URB_STARTING_ADDRESS_SHIFT 25 |
#define GEN7_3DSTATE_PUSH_CONSTANT_ALLOC_VS GEN7_3D(3, 1, 0x12) |
#define GEN7_3DSTATE_PUSH_CONSTANT_ALLOC_PS GEN7_3D(3, 1, 0x16) |
/* DW1 */ |
# define GEN7_PUSH_CONSTANT_BUFFER_OFFSET_SHIFT 16 |
struct gen7_cc_viewport { |
float min_depth; |
float max_depth; |
}; |
typedef enum { |
SAMPLER_FILTER_NEAREST = 0, |
SAMPLER_FILTER_BILINEAR, |
FILTER_COUNT |
} sampler_filter_t; |
typedef enum { |
SAMPLER_EXTEND_NONE = 0, |
SAMPLER_EXTEND_REPEAT, |
SAMPLER_EXTEND_PAD, |
SAMPLER_EXTEND_REFLECT, |
EXTEND_COUNT |
} sampler_extend_t; |
#endif |
/drivers/video/Intel-2D/sna/intel.h |
---|
0,0 → 1,152 |
/** enumeration of 3d consumers so some can maintain invariant state. */ |
enum last_3d { |
LAST_3D_OTHER, |
LAST_3D_VIDEO, |
LAST_3D_RENDER, |
LAST_3D_ROTATION |
}; |
typedef struct intel_screen_private { |
int cpp; |
#define RENDER_BATCH I915_EXEC_RENDER |
#define BLT_BATCH I915_EXEC_BLT |
unsigned int current_batch; |
dri_bufmgr *bufmgr; |
uint32_t batch_ptr[4096]; |
/** Byte offset in batch_ptr for the next dword to be emitted. */ |
unsigned int batch_used; |
/** Position in batch_ptr at the start of the current BEGIN_BATCH */ |
unsigned int batch_emit_start; |
/** Number of bytes to be emitted in the current BEGIN_BATCH. */ |
uint32_t batch_emitting; |
dri_bo *batch_bo, *last_batch_bo[2]; |
/** Whether we're in a section of code that can't tolerate flushing */ |
Bool in_batch_atomic; |
/** Ending batch_used that was verified by intel_start_batch_atomic() */ |
int batch_atomic_limit; |
struct list batch_pixmaps; |
drm_intel_bo *wa_scratch_bo; |
unsigned int tiling; |
#define INTEL_TILING_FB 0x1 |
#define INTEL_TILING_2D 0x2 |
#define INTEL_TILING_3D 0x4 |
#define INTEL_TILING_ALL (~0) |
Bool has_relaxed_fencing; |
int Chipset; |
unsigned int BR[20]; |
void (*vertex_flush) (struct intel_screen_private *intel); |
void (*batch_flush) (struct intel_screen_private *intel); |
void (*batch_commit_notify) (struct intel_screen_private *intel); |
Bool need_sync; |
int accel_pixmap_offset_alignment; |
int accel_max_x; |
int accel_max_y; |
int max_bo_size; |
int max_gtt_map_size; |
int max_tiling_size; |
struct { |
drm_intel_bo *gen4_vs_bo; |
drm_intel_bo *gen4_sf_bo; |
drm_intel_bo *gen4_wm_packed_bo; |
drm_intel_bo *gen4_wm_planar_bo; |
drm_intel_bo *gen4_cc_bo; |
drm_intel_bo *gen4_cc_vp_bo; |
drm_intel_bo *gen4_sampler_bo; |
drm_intel_bo *gen4_sip_kernel_bo; |
drm_intel_bo *wm_prog_packed_bo; |
drm_intel_bo *wm_prog_planar_bo; |
drm_intel_bo *gen6_blend_bo; |
drm_intel_bo *gen6_depth_stencil_bo; |
} video; |
/* Render accel state */ |
float scale_units[2][2]; |
/** Transform pointers for src/mask, or NULL if identity */ |
PictTransform *transform[2]; |
PixmapPtr render_source, render_mask, render_dest; |
PicturePtr render_source_picture, render_mask_picture, render_dest_picture; |
Bool needs_3d_invariant; |
Bool needs_render_state_emit; |
Bool needs_render_vertex_emit; |
/* i830 render accel state */ |
uint32_t render_dest_format; |
uint32_t cblend, ablend, s8_blendctl; |
/* i915 render accel state */ |
PixmapPtr texture[2]; |
uint32_t mapstate[6]; |
uint32_t samplerstate[6]; |
struct { |
int op; |
uint32_t dst_format; |
} i915_render_state; |
struct { |
int num_sf_outputs; |
int drawrect; |
uint32_t blend; |
dri_bo *samplers; |
dri_bo *kernel; |
} gen6_render_state; |
uint32_t prim_offset; |
void (*prim_emit)(struct intel_screen_private *intel, |
int srcX, int srcY, |
int maskX, int maskY, |
int dstX, int dstY, |
int w, int h); |
int floats_per_vertex; |
int last_floats_per_vertex; |
uint16_t vertex_offset; |
uint16_t vertex_count; |
uint16_t vertex_index; |
uint16_t vertex_used; |
uint32_t vertex_id; |
float vertex_ptr[4*1024]; |
dri_bo *vertex_bo; |
uint8_t surface_data[16*1024]; |
uint16_t surface_used; |
uint16_t surface_table; |
uint32_t surface_reloc; |
dri_bo *surface_bo; |
/* 965 render acceleration state */ |
struct gen4_render_state *gen4_render_state; |
Bool use_pageflipping; |
Bool use_triple_buffer; |
Bool force_fallback; |
Bool has_kernel_flush; |
Bool needs_flush; |
enum last_3d last_3d; |
/** |
* User option to print acceleration fallback info to the server log. |
*/ |
Bool fallback_debug; |
unsigned debug_flush; |
Bool has_prime_vmap_flush; |
} intel_screen_private; |
/drivers/video/Intel-2D/sna/kgem.c |
---|
0,0 → 1,5420 |
/* |
* Copyright (c) 2011 Intel Corporation |
* |
* Permission is hereby granted, free of charge, to any person obtaining a |
* copy of this software and associated documentation files (the "Software"), |
* to deal in the Software without restriction, including without limitation |
* the rights to use, copy, modify, merge, publish, distribute, sublicense, |
* and/or sell copies of the Software, and to permit persons to whom the |
* Software is furnished to do so, subject to the following conditions: |
* |
* The above copyright notice and this permission notice (including the next |
* paragraph) shall be included in all copies or substantial portions of the |
* Software. |
* |
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL |
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, |
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE |
* SOFTWARE. |
* |
* Authors: |
* Chris Wilson <chris@chris-wilson.co.uk> |
* |
*/ |
#ifdef HAVE_CONFIG_H |
#include "config.h" |
#endif |
#include "sna.h" |
#include "sna_reg.h" |
#include <time.h> |
#include <errno.h> |
#include <fcntl.h> |
#ifdef HAVE_VALGRIND |
#include <valgrind.h> |
#include <memcheck.h> |
#endif |
#ifdef HAVE_STRUCT_SYSINFO_TOTALRAM |
#include <sys/sysinfo.h> |
#endif |
#include "sna_cpuid.h" |
static struct kgem_bo * |
search_linear_cache(struct kgem *kgem, unsigned int num_pages, unsigned flags); |
static struct kgem_bo * |
search_snoop_cache(struct kgem *kgem, unsigned int num_pages, unsigned flags); |
#define DBG_NO_HW 0 |
#define DBG_NO_TILING 0 |
#define DBG_NO_CACHE 0 |
#define DBG_NO_CACHE_LEVEL 0 |
#define DBG_NO_CPU 0 |
#define DBG_NO_CREATE2 1 |
#define DBG_NO_USERPTR 0 |
#define DBG_NO_UNSYNCHRONIZED_USERPTR 0 |
#define DBG_NO_LLC 0 |
#define DBG_NO_SEMAPHORES 0 |
#define DBG_NO_MADV 1 |
#define DBG_NO_UPLOAD_CACHE 0 |
#define DBG_NO_UPLOAD_ACTIVE 0 |
#define DBG_NO_MAP_UPLOAD 0 |
#define DBG_NO_RELAXED_FENCING 0 |
#define DBG_NO_SECURE_BATCHES 0 |
#define DBG_NO_PINNED_BATCHES 0 |
#define DBG_NO_FAST_RELOC 0 |
#define DBG_NO_HANDLE_LUT 1 |
#define DBG_NO_WT 0 |
#define DBG_DUMP 0 |
#define FORCE_MMAP_SYNC 0 /* ((1 << DOMAIN_CPU) | (1 << DOMAIN_GTT)) */ |
#ifndef DEBUG_SYNC |
#define DEBUG_SYNC 0 |
#endif |
#if 0 |
#define ASSERT_IDLE(kgem__, handle__) assert(!__kgem_busy(kgem__, handle__)) |
#define ASSERT_MAYBE_IDLE(kgem__, handle__, expect__) assert(!(expect__) || !__kgem_busy(kgem__, handle__)) |
#else |
#define ASSERT_IDLE(kgem__, handle__) |
#define ASSERT_MAYBE_IDLE(kgem__, handle__, expect__) |
#endif |
/* Worst case seems to be 965gm where we cannot write within a cacheline that |
* is being simultaneously being read by the GPU, or within the sampler |
* prefetch. In general, the chipsets seem to have a requirement that sampler |
* offsets be aligned to a cacheline (64 bytes). |
*/ |
#define UPLOAD_ALIGNMENT 128 |
#define PAGE_ALIGN(x) ALIGN(x, PAGE_SIZE) |
#define NUM_PAGES(x) (((x) + PAGE_SIZE-1) / PAGE_SIZE) |
#define MAX_GTT_VMA_CACHE 512 |
#define MAX_CPU_VMA_CACHE INT16_MAX |
#define MAP_PRESERVE_TIME 10 |
#define MAKE_CPU_MAP(ptr) ((void*)((uintptr_t)(ptr) | 1)) |
#define MAKE_USER_MAP(ptr) ((void*)((uintptr_t)(ptr) | 3)) |
#define IS_USER_MAP(ptr) ((uintptr_t)(ptr) & 2) |
#define __MAP_TYPE(ptr) ((uintptr_t)(ptr) & 3) |
#define MAKE_REQUEST(rq, ring) ((struct kgem_request *)((uintptr_t)(rq) | (ring))) |
#define LOCAL_I915_PARAM_HAS_BLT 11 |
#define LOCAL_I915_PARAM_HAS_RELAXED_FENCING 12 |
#define LOCAL_I915_PARAM_HAS_RELAXED_DELTA 15 |
#define LOCAL_I915_PARAM_HAS_SEMAPHORES 20 |
#define LOCAL_I915_PARAM_HAS_SECURE_BATCHES 23 |
#define LOCAL_I915_PARAM_HAS_PINNED_BATCHES 24 |
#define LOCAL_I915_PARAM_HAS_NO_RELOC 25 |
#define LOCAL_I915_PARAM_HAS_HANDLE_LUT 26 |
#define LOCAL_I915_PARAM_HAS_WT 27 |
#define LOCAL_I915_EXEC_IS_PINNED (1<<10) |
#define LOCAL_I915_EXEC_NO_RELOC (1<<11) |
#define LOCAL_I915_EXEC_HANDLE_LUT (1<<12) |
struct local_i915_gem_userptr { |
uint64_t user_ptr; |
uint64_t user_size; |
uint32_t flags; |
#define I915_USERPTR_READ_ONLY (1<<0) |
#define I915_USERPTR_UNSYNCHRONIZED (1<<31) |
uint32_t handle; |
}; |
#define UNCACHED 0 |
#define SNOOPED 1 |
#define DISPLAY 2 |
struct local_i915_gem_caching { |
uint32_t handle; |
uint32_t caching; |
}; |
#define LOCAL_IOCTL_I915_GEM_SET_CACHING SRV_I915_GEM_SET_CACHING |
struct local_fbinfo { |
int width; |
int height; |
int pitch; |
int tiling; |
}; |
struct kgem_buffer { |
struct kgem_bo base; |
void *mem; |
uint32_t used; |
uint32_t need_io : 1; |
uint32_t write : 2; |
uint32_t mmapped : 1; |
}; |
static struct kgem_bo *__kgem_freed_bo; |
static struct kgem_request *__kgem_freed_request; |
static struct drm_i915_gem_exec_object2 _kgem_dummy_exec; |
static inline int bytes(struct kgem_bo *bo) |
{ |
return __kgem_bo_size(bo); |
} |
#define bucket(B) (B)->size.pages.bucket |
#define num_pages(B) (B)->size.pages.count |
#ifdef DEBUG_MEMORY |
static void debug_alloc(struct kgem *kgem, size_t size) |
{ |
kgem->debug_memory.bo_allocs++; |
kgem->debug_memory.bo_bytes += size; |
} |
static void debug_alloc__bo(struct kgem *kgem, struct kgem_bo *bo) |
{ |
debug_alloc(kgem, bytes(bo)); |
} |
#else |
#define debug_alloc(k, b) |
#define debug_alloc__bo(k, b) |
#endif |
#ifndef NDEBUG |
static void assert_tiling(struct kgem *kgem, struct kgem_bo *bo) |
{ |
struct drm_i915_gem_get_tiling tiling; |
assert(bo); |
VG_CLEAR(tiling); |
tiling.handle = bo->handle; |
tiling.tiling_mode = -1; |
(void)drmIoctl(kgem->fd, DRM_IOCTL_I915_GEM_GET_TILING, &tiling); |
assert(tiling.tiling_mode == bo->tiling); |
} |
#else |
#define assert_tiling(kgem, bo) |
#endif |
static void kgem_sna_reset(struct kgem *kgem) |
{ |
struct sna *sna = container_of(kgem, struct sna, kgem); |
sna->render.reset(sna); |
sna->blt_state.fill_bo = 0; |
} |
static void kgem_sna_flush(struct kgem *kgem) |
{ |
struct sna *sna = container_of(kgem, struct sna, kgem); |
sna->render.flush(sna); |
// if (sna->render.solid_cache.dirty) |
// sna_render_flush_solid(sna); |
} |
static bool gem_set_tiling(int fd, uint32_t handle, int tiling, int stride) |
{ |
struct drm_i915_gem_set_tiling set_tiling; |
int ret; |
if (DBG_NO_TILING) |
return false; |
VG_CLEAR(set_tiling); |
do { |
set_tiling.handle = handle; |
set_tiling.tiling_mode = tiling; |
set_tiling.stride = stride; |
ret = drmIoctl(fd, DRM_IOCTL_I915_GEM_SET_TILING, &set_tiling); |
} while (ret != 0); |
return ret == 0; |
} |
static bool gem_set_caching(int fd, uint32_t handle, int caching) |
{ |
struct local_i915_gem_caching arg; |
VG_CLEAR(arg); |
arg.handle = handle; |
arg.caching = caching; |
return drmIoctl(fd, LOCAL_IOCTL_I915_GEM_SET_CACHING, &arg) == 0; |
} |
static bool __kgem_throttle_retire(struct kgem *kgem, unsigned flags) |
{ |
if (flags & CREATE_NO_RETIRE) { |
DBG(("%s: not retiring per-request\n", __FUNCTION__)); |
return false; |
} |
if (!kgem->need_retire) { |
DBG(("%s: nothing to retire\n", __FUNCTION__)); |
return false; |
} |
if (kgem_retire(kgem)) |
return true; |
if (flags & CREATE_NO_THROTTLE || !kgem->need_throttle) { |
DBG(("%s: not throttling\n", __FUNCTION__)); |
return false; |
} |
kgem_throttle(kgem); |
return kgem_retire(kgem); |
} |
static void *__kgem_bo_map__gtt(struct kgem *kgem, struct kgem_bo *bo) |
{ |
struct drm_i915_gem_mmap_gtt mmap_arg; |
void *ptr; |
DBG(("%s(handle=%d, size=%d)\n", __FUNCTION__, |
bo->handle, bytes(bo))); |
assert(bo->proxy == NULL); |
assert(!bo->snoop); |
assert(kgem_bo_can_map(kgem, bo)); |
retry_gtt: |
VG_CLEAR(mmap_arg); |
mmap_arg.handle = bo->handle; |
if (drmIoctl(kgem->fd, DRM_IOCTL_I915_GEM_MMAP_GTT, &mmap_arg)) { |
(void)__kgem_throttle_retire(kgem, 0); |
if (kgem_expire_cache(kgem)) |
goto retry_gtt; |
if (kgem->need_expire) { |
kgem_cleanup_cache(kgem); |
goto retry_gtt; |
} |
printf("%s: failed to retrieve GTT offset for handle=%d\n", |
__FUNCTION__, bo->handle); |
return NULL; |
} |
retry_mmap: |
ptr = (void*)(int)mmap_arg.offset; |
if (ptr == NULL) { |
ErrorF("%s: failed to mmap handle=%d, %d bytes, into GTT domain\n", |
__FUNCTION__, bo->handle, bytes(bo)); |
ptr = NULL; |
} |
return ptr; |
} |
static int __gem_write(int fd, uint32_t handle, |
int offset, int length, |
const void *src) |
{ |
struct drm_i915_gem_pwrite pwrite; |
DBG(("%s(handle=%d, offset=%d, len=%d)\n", __FUNCTION__, |
handle, offset, length)); |
VG_CLEAR(pwrite); |
pwrite.handle = handle; |
pwrite.offset = offset; |
pwrite.size = length; |
pwrite.data_ptr = (uintptr_t)src; |
return drmIoctl(fd, DRM_IOCTL_I915_GEM_PWRITE, &pwrite); |
} |
static int gem_write(int fd, uint32_t handle, |
int offset, int length, |
const void *src) |
{ |
struct drm_i915_gem_pwrite pwrite; |
DBG(("%s(handle=%d, offset=%d, len=%d)\n", __FUNCTION__, |
handle, offset, length)); |
VG_CLEAR(pwrite); |
pwrite.handle = handle; |
/* align the transfer to cachelines; fortuitously this is safe! */ |
if ((offset | length) & 63) { |
pwrite.offset = offset & ~63; |
pwrite.size = ALIGN(offset+length, 64) - pwrite.offset; |
pwrite.data_ptr = (uintptr_t)src + pwrite.offset - offset; |
} else { |
pwrite.offset = offset; |
pwrite.size = length; |
pwrite.data_ptr = (uintptr_t)src; |
} |
return drmIoctl(fd, DRM_IOCTL_I915_GEM_PWRITE, &pwrite); |
} |
bool __kgem_busy(struct kgem *kgem, int handle) |
{ |
struct drm_i915_gem_busy busy; |
VG_CLEAR(busy); |
busy.handle = handle; |
busy.busy = !kgem->wedged; |
(void)drmIoctl(kgem->fd, DRM_IOCTL_I915_GEM_BUSY, &busy); |
DBG(("%s: handle=%d, busy=%d, wedged=%d\n", |
__FUNCTION__, handle, busy.busy, kgem->wedged)); |
return busy.busy; |
} |
static void kgem_bo_retire(struct kgem *kgem, struct kgem_bo *bo) |
{ |
DBG(("%s: retiring bo handle=%d (needed flush? %d), rq? %d [busy?=%d]\n", |
__FUNCTION__, bo->handle, bo->needs_flush, bo->rq != NULL, |
__kgem_busy(kgem, bo->handle))); |
assert(bo->exec == NULL); |
assert(list_is_empty(&bo->vma)); |
if (bo->rq) { |
if (!__kgem_busy(kgem, bo->handle)) { |
__kgem_bo_clear_busy(bo); |
kgem_retire(kgem); |
} |
} else { |
assert(!bo->needs_flush); |
ASSERT_IDLE(kgem, bo->handle); |
} |
} |
bool kgem_bo_write(struct kgem *kgem, struct kgem_bo *bo, |
const void *data, int length) |
{ |
assert(bo->refcnt); |
assert(!bo->purged); |
assert(bo->proxy == NULL); |
ASSERT_IDLE(kgem, bo->handle); |
assert(length <= bytes(bo)); |
if (gem_write(kgem->fd, bo->handle, 0, length, data)) |
return false; |
DBG(("%s: flush=%d, domain=%d\n", __FUNCTION__, bo->flush, bo->domain)); |
if (bo->exec == NULL) { |
kgem_bo_retire(kgem, bo); |
bo->domain = DOMAIN_NONE; |
} |
bo->gtt_dirty = true; |
return true; |
} |
static uint32_t gem_create(int fd, int num_pages) |
{ |
struct drm_i915_gem_create create; |
VG_CLEAR(create); |
create.handle = 0; |
create.size = PAGE_SIZE * num_pages; |
(void)drmIoctl(fd, DRM_IOCTL_I915_GEM_CREATE, &create); |
return create.handle; |
} |
static bool |
kgem_bo_set_purgeable(struct kgem *kgem, struct kgem_bo *bo) |
{ |
#if DBG_NO_MADV |
return true; |
#else |
struct drm_i915_gem_madvise madv; |
assert(bo->exec == NULL); |
assert(!bo->purged); |
VG_CLEAR(madv); |
madv.handle = bo->handle; |
madv.madv = I915_MADV_DONTNEED; |
if (drmIoctl(kgem->fd, DRM_IOCTL_I915_GEM_MADVISE, &madv) == 0) { |
bo->purged = 1; |
kgem->need_purge |= !madv.retained && bo->domain == DOMAIN_GPU; |
return madv.retained; |
} |
return true; |
#endif |
} |
static bool |
kgem_bo_is_retained(struct kgem *kgem, struct kgem_bo *bo) |
{ |
#if DBG_NO_MADV |
return true; |
#else |
struct drm_i915_gem_madvise madv; |
if (!bo->purged) |
return true; |
VG_CLEAR(madv); |
madv.handle = bo->handle; |
madv.madv = I915_MADV_DONTNEED; |
if (drmIoctl(kgem->fd, DRM_IOCTL_I915_GEM_MADVISE, &madv) == 0) |
return madv.retained; |
return false; |
#endif |
} |
static bool |
kgem_bo_clear_purgeable(struct kgem *kgem, struct kgem_bo *bo) |
{ |
#if DBG_NO_MADV |
return true; |
#else |
struct drm_i915_gem_madvise madv; |
assert(bo->purged); |
VG_CLEAR(madv); |
madv.handle = bo->handle; |
madv.madv = I915_MADV_WILLNEED; |
if (drmIoctl(kgem->fd, DRM_IOCTL_I915_GEM_MADVISE, &madv) == 0) { |
bo->purged = !madv.retained; |
kgem->need_purge |= !madv.retained && bo->domain == DOMAIN_GPU; |
return madv.retained; |
} |
return false; |
#endif |
} |
static void gem_close(int fd, uint32_t handle) |
{ |
struct drm_gem_close close; |
VG_CLEAR(close); |
close.handle = handle; |
(void)drmIoctl(fd, DRM_IOCTL_GEM_CLOSE, &close); |
} |
constant inline static unsigned long __fls(unsigned long word) |
{ |
#if defined(__GNUC__) && (defined(__i386__) || defined(__x86__) || defined(__x86_64__)) |
asm("bsr %1,%0" |
: "=r" (word) |
: "rm" (word)); |
return word; |
#else |
unsigned int v = 0; |
while (word >>= 1) |
v++; |
return v; |
#endif |
} |
constant inline static int cache_bucket(int num_pages) |
{ |
return __fls(num_pages); |
} |
static struct kgem_bo *__kgem_bo_init(struct kgem_bo *bo, |
int handle, int num_pages) |
{ |
assert(num_pages); |
memset(bo, 0, sizeof(*bo)); |
bo->refcnt = 1; |
bo->handle = handle; |
bo->target_handle = -1; |
num_pages(bo) = num_pages; |
bucket(bo) = cache_bucket(num_pages); |
bo->reusable = true; |
bo->domain = DOMAIN_CPU; |
list_init(&bo->request); |
list_init(&bo->list); |
list_init(&bo->vma); |
return bo; |
} |
static struct kgem_bo *__kgem_bo_alloc(int handle, int num_pages) |
{ |
struct kgem_bo *bo; |
if (__kgem_freed_bo) { |
bo = __kgem_freed_bo; |
__kgem_freed_bo = *(struct kgem_bo **)bo; |
} else { |
bo = malloc(sizeof(*bo)); |
if (bo == NULL) |
return NULL; |
} |
return __kgem_bo_init(bo, handle, num_pages); |
} |
static struct kgem_request *__kgem_request_alloc(struct kgem *kgem) |
{ |
struct kgem_request *rq; |
rq = __kgem_freed_request; |
if (rq) { |
__kgem_freed_request = *(struct kgem_request **)rq; |
} else { |
rq = malloc(sizeof(*rq)); |
if (rq == NULL) |
rq = &kgem->static_request; |
} |
list_init(&rq->buffers); |
rq->bo = NULL; |
rq->ring = 0; |
return rq; |
} |
static void __kgem_request_free(struct kgem_request *rq) |
{ |
_list_del(&rq->list); |
*(struct kgem_request **)rq = __kgem_freed_request; |
__kgem_freed_request = rq; |
} |
static struct list *inactive(struct kgem *kgem, int num_pages) |
{ |
assert(num_pages < MAX_CACHE_SIZE / PAGE_SIZE); |
assert(cache_bucket(num_pages) < NUM_CACHE_BUCKETS); |
return &kgem->inactive[cache_bucket(num_pages)]; |
} |
static struct list *active(struct kgem *kgem, int num_pages, int tiling) |
{ |
assert(num_pages < MAX_CACHE_SIZE / PAGE_SIZE); |
assert(cache_bucket(num_pages) < NUM_CACHE_BUCKETS); |
return &kgem->active[cache_bucket(num_pages)][tiling]; |
} |
static size_t |
agp_aperture_size(struct pci_device *dev, unsigned gen) |
{ |
/* XXX assume that only future chipsets are unknown and follow |
* the post gen2 PCI layout. |
*/ |
return 0; |
} |
static size_t |
total_ram_size(void) |
{ |
uint32_t data[9]; |
size_t size = 0; |
asm volatile("int $0x40" |
: "=a" (size) |
: "a" (18),"b"(20), "c" (data) |
: "memory"); |
return size != -1 ? size : 0; |
} |
static unsigned |
cpu_cache_size__cpuid4(void) |
{ |
/* Deterministic Cache Parmaeters (Function 04h)": |
* When EAX is initialized to a value of 4, the CPUID instruction |
* returns deterministic cache information in the EAX, EBX, ECX |
* and EDX registers. This function requires ECX be initialized |
* with an index which indicates which cache to return information |
* about. The OS is expected to call this function (CPUID.4) with |
* ECX = 0, 1, 2, until EAX[4:0] == 0, indicating no more caches. |
* The order in which the caches are returned is not specified |
* and may change at Intel's discretion. |
* |
* Calculating the Cache Size in bytes: |
* = (Ways +1) * (Partitions +1) * (Line Size +1) * (Sets +1) |
*/ |
unsigned int eax, ebx, ecx, edx; |
unsigned int llc_size = 0; |
int cnt = 0; |
if (__get_cpuid_max(BASIC_CPUID, NULL) < 4) |
return 0; |
do { |
unsigned associativity, line_partitions, line_size, sets; |
__cpuid_count(4, cnt++, eax, ebx, ecx, edx); |
if ((eax & 0x1f) == 0) |
break; |
associativity = ((ebx >> 22) & 0x3ff) + 1; |
line_partitions = ((ebx >> 12) & 0x3ff) + 1; |
line_size = (ebx & 0xfff) + 1; |
sets = ecx + 1; |
llc_size = associativity * line_partitions * line_size * sets; |
} while (1); |
return llc_size; |
} |
static int gem_param(struct kgem *kgem, int name) |
{ |
drm_i915_getparam_t gp; |
int v = -1; /* No param uses the sign bit, reserve it for errors */ |
VG_CLEAR(gp); |
gp.param = name; |
gp.value = &v; |
if (drmIoctl(kgem->fd, DRM_IOCTL_I915_GETPARAM, &gp)) |
return -1; |
VG(VALGRIND_MAKE_MEM_DEFINED(&v, sizeof(v))); |
return v; |
} |
static bool test_has_execbuffer2(struct kgem *kgem) |
{ |
return 1; |
} |
static bool test_has_no_reloc(struct kgem *kgem) |
{ |
if (DBG_NO_FAST_RELOC) |
return false; |
return gem_param(kgem, LOCAL_I915_PARAM_HAS_NO_RELOC) > 0; |
} |
static bool test_has_handle_lut(struct kgem *kgem) |
{ |
if (DBG_NO_HANDLE_LUT) |
return false; |
return gem_param(kgem, LOCAL_I915_PARAM_HAS_HANDLE_LUT) > 0; |
} |
static bool test_has_wt(struct kgem *kgem) |
{ |
if (DBG_NO_WT) |
return false; |
return gem_param(kgem, LOCAL_I915_PARAM_HAS_WT) > 0; |
} |
static bool test_has_semaphores_enabled(struct kgem *kgem) |
{ |
bool detected = false; |
int ret; |
if (DBG_NO_SEMAPHORES) |
return false; |
ret = gem_param(kgem, LOCAL_I915_PARAM_HAS_SEMAPHORES); |
if (ret != -1) |
return ret > 0; |
return detected; |
} |
static bool __kgem_throttle(struct kgem *kgem) |
{ |
if (drmIoctl(kgem->fd, DRM_IOCTL_I915_GEM_THROTTLE, NULL) == 0) |
return false; |
return errno == EIO; |
} |
static bool is_hw_supported(struct kgem *kgem, |
struct pci_device *dev) |
{ |
if (DBG_NO_HW) |
return false; |
if (!test_has_execbuffer2(kgem)) |
return false; |
if (kgem->gen == (unsigned)-1) /* unknown chipset, assume future gen */ |
return kgem->has_blt; |
/* Although pre-855gm the GMCH is fubar, it works mostly. So |
* let the user decide through "NoAccel" whether or not to risk |
* hw acceleration. |
*/ |
if (kgem->gen == 060 && dev->revision < 8) { |
/* pre-production SNB with dysfunctional BLT */ |
return false; |
} |
if (kgem->gen >= 060) /* Only if the kernel supports the BLT ring */ |
return kgem->has_blt; |
return true; |
} |
static bool test_has_relaxed_fencing(struct kgem *kgem) |
{ |
if (kgem->gen < 040) { |
if (DBG_NO_RELAXED_FENCING) |
return false; |
return gem_param(kgem, LOCAL_I915_PARAM_HAS_RELAXED_FENCING) > 0; |
} else |
return true; |
} |
static bool test_has_llc(struct kgem *kgem) |
{ |
int has_llc = -1; |
if (DBG_NO_LLC) |
return false; |
#if defined(I915_PARAM_HAS_LLC) /* Expected in libdrm-2.4.31 */ |
has_llc = gem_param(kgem, I915_PARAM_HAS_LLC); |
#endif |
if (has_llc == -1) { |
DBG(("%s: no kernel/drm support for HAS_LLC, assuming support for LLC based on GPU generation\n", __FUNCTION__)); |
has_llc = kgem->gen >= 060; |
} |
return has_llc; |
} |
static bool test_has_caching(struct kgem *kgem) |
{ |
uint32_t handle; |
bool ret; |
if (DBG_NO_CACHE_LEVEL) |
return false; |
/* Incoherent blt and sampler hangs the GPU */ |
if (kgem->gen == 040) |
return false; |
handle = gem_create(kgem->fd, 1); |
if (handle == 0) |
return false; |
ret = gem_set_caching(kgem->fd, handle, UNCACHED); |
gem_close(kgem->fd, handle); |
return ret; |
} |
static bool test_has_userptr(struct kgem *kgem) |
{ |
#if defined(USE_USERPTR) |
uint32_t handle; |
void *ptr; |
if (DBG_NO_USERPTR) |
return false; |
/* Incoherent blt and sampler hangs the GPU */ |
if (kgem->gen == 040) |
return false; |
if (posix_memalign(&ptr, PAGE_SIZE, PAGE_SIZE)) |
return false; |
handle = gem_userptr(kgem->fd, ptr, PAGE_SIZE, false); |
gem_close(kgem->fd, handle); |
free(ptr); |
return handle != 0; |
#else |
return false; |
#endif |
} |
static bool test_has_create2(struct kgem *kgem) |
{ |
#if defined(USE_CREATE2) |
struct local_i915_gem_create2 args; |
if (DBG_NO_CREATE2) |
return false; |
memset(&args, 0, sizeof(args)); |
args.size = PAGE_SIZE; |
args.caching = DISPLAY; |
if (drmIoctl(kgem->fd, LOCAL_IOCTL_I915_GEM_CREATE2, &args) == 0) |
gem_close(kgem->fd, args.handle); |
return args.handle != 0; |
#else |
return false; |
#endif |
} |
static bool test_has_secure_batches(struct kgem *kgem) |
{ |
if (DBG_NO_SECURE_BATCHES) |
return false; |
return gem_param(kgem, LOCAL_I915_PARAM_HAS_SECURE_BATCHES) > 0; |
} |
static bool test_has_pinned_batches(struct kgem *kgem) |
{ |
if (DBG_NO_PINNED_BATCHES) |
return false; |
return gem_param(kgem, LOCAL_I915_PARAM_HAS_PINNED_BATCHES) > 0; |
} |
static bool kgem_init_pinned_batches(struct kgem *kgem) |
{ |
int count[2] = { 2, 1 }; |
int size[2] = { 1, 2 }; |
int n, i; |
if (kgem->wedged) |
return true; |
for (n = 0; n < ARRAY_SIZE(count); n++) { |
for (i = 0; i < count[n]; i++) { |
struct drm_i915_gem_pin pin; |
struct kgem_bo *bo; |
VG_CLEAR(pin); |
pin.handle = gem_create(kgem->fd, size[n]); |
if (pin.handle == 0) |
goto err; |
DBG(("%s: new handle=%d, num_pages=%d\n", |
__FUNCTION__, pin.handle, size[n])); |
bo = __kgem_bo_alloc(pin.handle, size[n]); |
if (bo == NULL) { |
gem_close(kgem->fd, pin.handle); |
goto err; |
} |
pin.alignment = 0; |
if (drmIoctl(kgem->fd, DRM_IOCTL_I915_GEM_PIN, &pin)) { |
gem_close(kgem->fd, pin.handle); |
goto err; |
} |
bo->presumed_offset = pin.offset; |
debug_alloc__bo(kgem, bo); |
list_add(&bo->list, &kgem->pinned_batches[n]); |
} |
} |
return true; |
err: |
for (n = 0; n < ARRAY_SIZE(kgem->pinned_batches); n++) { |
while (!list_is_empty(&kgem->pinned_batches[n])) { |
kgem_bo_destroy(kgem, |
list_first_entry(&kgem->pinned_batches[n], |
struct kgem_bo, list)); |
} |
} |
/* For simplicity populate the lists with a single unpinned bo */ |
for (n = 0; n < ARRAY_SIZE(count); n++) { |
struct kgem_bo *bo; |
uint32_t handle; |
handle = gem_create(kgem->fd, size[n]); |
if (handle == 0) |
break; |
bo = __kgem_bo_alloc(handle, size[n]); |
if (bo == NULL) { |
gem_close(kgem->fd, handle); |
break; |
} |
debug_alloc__bo(kgem, bo); |
list_add(&bo->list, &kgem->pinned_batches[n]); |
} |
return false; |
} |
void kgem_init(struct kgem *kgem, int fd, struct pci_device *dev, unsigned gen) |
{ |
struct drm_i915_gem_get_aperture aperture; |
size_t totalram; |
unsigned half_gpu_max; |
unsigned int i, j; |
DBG(("%s: fd=%d, gen=%d\n", __FUNCTION__, fd, gen)); |
memset(kgem, 0, sizeof(*kgem)); |
kgem->fd = fd; |
kgem->gen = gen; |
list_init(&kgem->requests[0]); |
list_init(&kgem->requests[1]); |
list_init(&kgem->batch_buffers); |
list_init(&kgem->active_buffers); |
list_init(&kgem->flushing); |
list_init(&kgem->large); |
list_init(&kgem->large_inactive); |
list_init(&kgem->snoop); |
list_init(&kgem->scanout); |
for (i = 0; i < ARRAY_SIZE(kgem->pinned_batches); i++) |
list_init(&kgem->pinned_batches[i]); |
for (i = 0; i < ARRAY_SIZE(kgem->inactive); i++) |
list_init(&kgem->inactive[i]); |
for (i = 0; i < ARRAY_SIZE(kgem->active); i++) { |
for (j = 0; j < ARRAY_SIZE(kgem->active[i]); j++) |
list_init(&kgem->active[i][j]); |
} |
for (i = 0; i < ARRAY_SIZE(kgem->vma); i++) { |
for (j = 0; j < ARRAY_SIZE(kgem->vma[i].inactive); j++) |
list_init(&kgem->vma[i].inactive[j]); |
} |
kgem->vma[MAP_GTT].count = -MAX_GTT_VMA_CACHE; |
kgem->vma[MAP_CPU].count = -MAX_CPU_VMA_CACHE; |
kgem->has_blt = gem_param(kgem, LOCAL_I915_PARAM_HAS_BLT) > 0; |
DBG(("%s: has BLT ring? %d\n", __FUNCTION__, |
kgem->has_blt)); |
kgem->has_relaxed_delta = |
gem_param(kgem, LOCAL_I915_PARAM_HAS_RELAXED_DELTA) > 0; |
DBG(("%s: has relaxed delta? %d\n", __FUNCTION__, |
kgem->has_relaxed_delta)); |
kgem->has_relaxed_fencing = test_has_relaxed_fencing(kgem); |
DBG(("%s: has relaxed fencing? %d\n", __FUNCTION__, |
kgem->has_relaxed_fencing)); |
kgem->has_llc = test_has_llc(kgem); |
DBG(("%s: has shared last-level-cache? %d\n", __FUNCTION__, |
kgem->has_llc)); |
kgem->has_wt = test_has_wt(kgem); |
DBG(("%s: has write-through caching for scanouts? %d\n", __FUNCTION__, |
kgem->has_wt)); |
kgem->has_caching = test_has_caching(kgem); |
DBG(("%s: has set-cache-level? %d\n", __FUNCTION__, |
kgem->has_caching)); |
kgem->has_userptr = test_has_userptr(kgem); |
DBG(("%s: has userptr? %d\n", __FUNCTION__, |
kgem->has_userptr)); |
kgem->has_create2 = test_has_create2(kgem); |
kgem->has_create2 = 0; |
DBG(("%s: has create2? %d\n", __FUNCTION__, |
kgem->has_create2)); |
kgem->has_no_reloc = test_has_no_reloc(kgem); |
DBG(("%s: has no-reloc? %d\n", __FUNCTION__, |
kgem->has_no_reloc)); |
kgem->has_handle_lut = test_has_handle_lut(kgem); |
kgem->has_handle_lut = 0; |
DBG(("%s: has handle-lut? %d\n", __FUNCTION__, |
kgem->has_handle_lut)); |
kgem->has_semaphores = false; |
if (kgem->has_blt && test_has_semaphores_enabled(kgem)) |
kgem->has_semaphores = true; |
DBG(("%s: semaphores enabled? %d\n", __FUNCTION__, |
kgem->has_semaphores)); |
kgem->can_blt_cpu = gen >= 030; |
DBG(("%s: can blt to cpu? %d\n", __FUNCTION__, |
kgem->can_blt_cpu)); |
kgem->has_secure_batches = test_has_secure_batches(kgem); |
DBG(("%s: can use privileged batchbuffers? %d\n", __FUNCTION__, |
kgem->has_secure_batches)); |
kgem->has_pinned_batches = test_has_pinned_batches(kgem); |
DBG(("%s: can use pinned batchbuffers (to avoid CS w/a)? %d\n", __FUNCTION__, |
kgem->has_pinned_batches)); |
if (!is_hw_supported(kgem, dev)) { |
printf("Detected unsupported/dysfunctional hardware, disabling acceleration.\n"); |
kgem->wedged = 1; |
} else if (__kgem_throttle(kgem)) { |
printf("Detected a hung GPU, disabling acceleration.\n"); |
kgem->wedged = 1; |
} |
kgem->batch_size = ARRAY_SIZE(kgem->batch); |
if (gen == 020 && !kgem->has_pinned_batches) |
/* Limited to what we can pin */ |
kgem->batch_size = 4*1024; |
if (gen == 022) |
/* 865g cannot handle a batch spanning multiple pages */ |
kgem->batch_size = PAGE_SIZE / sizeof(uint32_t); |
if ((gen >> 3) == 7) |
kgem->batch_size = 16*1024; |
if (!kgem->has_relaxed_delta && kgem->batch_size > 4*1024) |
kgem->batch_size = 4*1024; |
if (!kgem_init_pinned_batches(kgem) && gen == 020) { |
printf("Unable to reserve memory for GPU, disabling acceleration.\n"); |
kgem->wedged = 1; |
} |
DBG(("%s: maximum batch size? %d\n", __FUNCTION__, |
kgem->batch_size)); |
kgem->min_alignment = 4; |
if (gen < 040) |
kgem->min_alignment = 64; |
kgem->half_cpu_cache_pages = cpu_cache_size() >> 13; |
DBG(("%s: last-level cache size: %d bytes, threshold in pages: %d\n", |
__FUNCTION__, cpu_cache_size(), kgem->half_cpu_cache_pages)); |
kgem->next_request = __kgem_request_alloc(kgem); |
DBG(("%s: cpu bo enabled %d: llc? %d, set-cache-level? %d, userptr? %d\n", __FUNCTION__, |
!DBG_NO_CPU && (kgem->has_llc | kgem->has_userptr | kgem->has_caching), |
kgem->has_llc, kgem->has_caching, kgem->has_userptr)); |
VG_CLEAR(aperture); |
aperture.aper_size = 0; |
(void)drmIoctl(fd, DRM_IOCTL_I915_GEM_GET_APERTURE, &aperture); |
if (aperture.aper_size == 0) |
aperture.aper_size = 64*1024*1024; |
DBG(("%s: aperture size %lld, available now %lld\n", |
__FUNCTION__, |
(long long)aperture.aper_size, |
(long long)aperture.aper_available_size)); |
kgem->aperture_total = aperture.aper_size; |
kgem->aperture_high = aperture.aper_size * 3/4; |
kgem->aperture_low = aperture.aper_size * 1/3; |
if (gen < 033) { |
/* Severe alignment penalties */ |
kgem->aperture_high /= 2; |
kgem->aperture_low /= 2; |
} |
DBG(("%s: aperture low=%d [%d], high=%d [%d]\n", __FUNCTION__, |
kgem->aperture_low, kgem->aperture_low / (1024*1024), |
kgem->aperture_high, kgem->aperture_high / (1024*1024))); |
kgem->aperture_mappable = agp_aperture_size(dev, gen); |
if (kgem->aperture_mappable == 0 || |
kgem->aperture_mappable > aperture.aper_size) |
kgem->aperture_mappable = aperture.aper_size; |
DBG(("%s: aperture mappable=%d [%d MiB]\n", __FUNCTION__, |
kgem->aperture_mappable, kgem->aperture_mappable / (1024*1024))); |
kgem->buffer_size = 64 * 1024; |
while (kgem->buffer_size < kgem->aperture_mappable >> 10) |
kgem->buffer_size *= 2; |
if (kgem->buffer_size >> 12 > kgem->half_cpu_cache_pages) |
kgem->buffer_size = kgem->half_cpu_cache_pages << 12; |
kgem->buffer_size = 1 << __fls(kgem->buffer_size); |
DBG(("%s: buffer size=%d [%d KiB]\n", __FUNCTION__, |
kgem->buffer_size, kgem->buffer_size / 1024)); |
assert(kgem->buffer_size); |
kgem->max_object_size = 3 * (kgem->aperture_high >> 12) << 10; |
kgem->max_gpu_size = kgem->max_object_size; |
if (!kgem->has_llc && kgem->max_gpu_size > MAX_CACHE_SIZE) |
kgem->max_gpu_size = MAX_CACHE_SIZE; |
totalram = total_ram_size(); |
if (totalram == 0) { |
DBG(("%s: total ram size unknown, assuming maximum of total aperture\n", |
__FUNCTION__)); |
totalram = kgem->aperture_total; |
} |
DBG(("%s: total ram=%ld\n", __FUNCTION__, (long)totalram)); |
if (kgem->max_object_size > totalram / 2) |
kgem->max_object_size = totalram / 2; |
if (kgem->max_gpu_size > totalram / 4) |
kgem->max_gpu_size = totalram / 4; |
kgem->max_cpu_size = kgem->max_object_size; |
half_gpu_max = kgem->max_gpu_size / 2; |
kgem->max_copy_tile_size = (MAX_CACHE_SIZE + 1)/2; |
if (kgem->max_copy_tile_size > half_gpu_max) |
kgem->max_copy_tile_size = half_gpu_max; |
if (kgem->has_llc) |
kgem->max_upload_tile_size = kgem->max_copy_tile_size; |
else |
kgem->max_upload_tile_size = kgem->aperture_mappable / 4; |
if (kgem->max_upload_tile_size > half_gpu_max) |
kgem->max_upload_tile_size = half_gpu_max; |
if (kgem->max_upload_tile_size > kgem->aperture_high/2) |
kgem->max_upload_tile_size = kgem->aperture_high/2; |
if (kgem->max_upload_tile_size > kgem->aperture_low) |
kgem->max_upload_tile_size = kgem->aperture_low; |
if (kgem->max_upload_tile_size < 16*PAGE_SIZE) |
kgem->max_upload_tile_size = 16*PAGE_SIZE; |
kgem->large_object_size = MAX_CACHE_SIZE; |
if (kgem->large_object_size > half_gpu_max) |
kgem->large_object_size = half_gpu_max; |
if (kgem->max_copy_tile_size > kgem->aperture_high/2) |
kgem->max_copy_tile_size = kgem->aperture_high/2; |
if (kgem->max_copy_tile_size > kgem->aperture_low) |
kgem->max_copy_tile_size = kgem->aperture_low; |
if (kgem->max_copy_tile_size < 16*PAGE_SIZE) |
kgem->max_copy_tile_size = 16*PAGE_SIZE; |
if (kgem->has_llc | kgem->has_caching | kgem->has_userptr) { |
if (kgem->large_object_size > kgem->max_cpu_size) |
kgem->large_object_size = kgem->max_cpu_size; |
} else |
kgem->max_cpu_size = 0; |
if (DBG_NO_CPU) |
kgem->max_cpu_size = 0; |
DBG(("%s: maximum object size=%d\n", |
__FUNCTION__, kgem->max_object_size)); |
DBG(("%s: large object thresold=%d\n", |
__FUNCTION__, kgem->large_object_size)); |
DBG(("%s: max object sizes (gpu=%d, cpu=%d, tile upload=%d, copy=%d)\n", |
__FUNCTION__, |
kgem->max_gpu_size, kgem->max_cpu_size, |
kgem->max_upload_tile_size, kgem->max_copy_tile_size)); |
/* Convert the aperture thresholds to pages */ |
kgem->aperture_low /= PAGE_SIZE; |
kgem->aperture_high /= PAGE_SIZE; |
kgem->fence_max = gem_param(kgem, I915_PARAM_NUM_FENCES_AVAIL) - 2; |
if ((int)kgem->fence_max < 0) |
kgem->fence_max = 5; /* minimum safe value for all hw */ |
DBG(("%s: max fences=%d\n", __FUNCTION__, kgem->fence_max)); |
kgem->batch_flags_base = 0; |
if (kgem->has_no_reloc) |
kgem->batch_flags_base |= LOCAL_I915_EXEC_NO_RELOC; |
if (kgem->has_handle_lut) |
kgem->batch_flags_base |= LOCAL_I915_EXEC_HANDLE_LUT; |
if (kgem->has_pinned_batches) |
kgem->batch_flags_base |= LOCAL_I915_EXEC_IS_PINNED; |
} |
/* XXX hopefully a good approximation */ |
uint32_t kgem_get_unique_id(struct kgem *kgem) |
{ |
uint32_t id; |
id = ++kgem->unique_id; |
if (id == 0) |
id = ++kgem->unique_id; |
return id; |
} |
inline static uint32_t kgem_pitch_alignment(struct kgem *kgem, unsigned flags) |
{ |
if (flags & CREATE_PRIME) |
return 256; |
if (flags & CREATE_SCANOUT) |
return 64; |
return kgem->min_alignment; |
} |
void kgem_get_tile_size(struct kgem *kgem, int tiling, |
int *tile_width, int *tile_height, int *tile_size) |
{ |
if (kgem->gen <= 030) { |
if (tiling) { |
if (kgem->gen < 030) { |
*tile_width = 128; |
*tile_height = 16; |
*tile_size = 2048; |
} else { |
*tile_width = 512; |
*tile_height = 8; |
*tile_size = 4096; |
} |
} else { |
*tile_width = 1; |
*tile_height = 1; |
*tile_size = 1; |
} |
} else switch (tiling) { |
default: |
case I915_TILING_NONE: |
*tile_width = 1; |
*tile_height = 1; |
*tile_size = 1; |
break; |
case I915_TILING_X: |
*tile_width = 512; |
*tile_height = 8; |
*tile_size = 4096; |
break; |
case I915_TILING_Y: |
*tile_width = 128; |
*tile_height = 32; |
*tile_size = 4096; |
break; |
} |
} |
uint32_t kgem_surface_size(struct kgem *kgem, |
bool relaxed_fencing, |
unsigned flags, |
uint32_t width, |
uint32_t height, |
uint32_t bpp, |
uint32_t tiling, |
uint32_t *pitch) |
{ |
uint32_t tile_width, tile_height; |
uint32_t size; |
assert(width <= MAXSHORT); |
assert(height <= MAXSHORT); |
assert(bpp >= 8); |
if (kgem->gen <= 030) { |
if (tiling) { |
if (kgem->gen < 030) { |
tile_width = 128; |
tile_height = 32; |
} else { |
tile_width = 512; |
tile_height = 16; |
} |
} else { |
tile_width = 2 * bpp >> 3; |
tile_width = ALIGN(tile_width, |
kgem_pitch_alignment(kgem, flags)); |
tile_height = 2; |
} |
} else switch (tiling) { |
default: |
case I915_TILING_NONE: |
tile_width = 2 * bpp >> 3; |
tile_width = ALIGN(tile_width, |
kgem_pitch_alignment(kgem, flags)); |
tile_height = 2; |
break; |
/* XXX align to an even tile row */ |
case I915_TILING_X: |
tile_width = 512; |
tile_height = 16; |
break; |
case I915_TILING_Y: |
tile_width = 128; |
tile_height = 64; |
break; |
} |
*pitch = ALIGN(width * bpp / 8, tile_width); |
height = ALIGN(height, tile_height); |
if (kgem->gen >= 040) |
return PAGE_ALIGN(*pitch * height); |
/* If it is too wide for the blitter, don't even bother. */ |
if (tiling != I915_TILING_NONE) { |
if (*pitch > 8192) |
return 0; |
for (size = tile_width; size < *pitch; size <<= 1) |
; |
*pitch = size; |
} else { |
if (*pitch >= 32768) |
return 0; |
} |
size = *pitch * height; |
if (relaxed_fencing || tiling == I915_TILING_NONE) |
return PAGE_ALIGN(size); |
/* We need to allocate a pot fence region for a tiled buffer. */ |
if (kgem->gen < 030) |
tile_width = 512 * 1024; |
else |
tile_width = 1024 * 1024; |
while (tile_width < size) |
tile_width *= 2; |
return tile_width; |
} |
static uint32_t kgem_aligned_height(struct kgem *kgem, |
uint32_t height, uint32_t tiling) |
{ |
uint32_t tile_height; |
if (kgem->gen <= 030) { |
tile_height = tiling ? kgem->gen < 030 ? 32 : 16 : 1; |
} else switch (tiling) { |
/* XXX align to an even tile row */ |
default: |
case I915_TILING_NONE: |
tile_height = 1; |
break; |
case I915_TILING_X: |
tile_height = 16; |
break; |
case I915_TILING_Y: |
tile_height = 64; |
break; |
} |
return ALIGN(height, tile_height); |
} |
static struct drm_i915_gem_exec_object2 * |
kgem_add_handle(struct kgem *kgem, struct kgem_bo *bo) |
{ |
struct drm_i915_gem_exec_object2 *exec; |
DBG(("%s: handle=%d, index=%d\n", |
__FUNCTION__, bo->handle, kgem->nexec)); |
assert(kgem->nexec < ARRAY_SIZE(kgem->exec)); |
bo->target_handle = kgem->has_handle_lut ? kgem->nexec : bo->handle; |
exec = memset(&kgem->exec[kgem->nexec++], 0, sizeof(*exec)); |
exec->handle = bo->handle; |
exec->offset = bo->presumed_offset; |
kgem->aperture += num_pages(bo); |
return exec; |
} |
static void kgem_add_bo(struct kgem *kgem, struct kgem_bo *bo) |
{ |
bo->exec = kgem_add_handle(kgem, bo); |
bo->rq = MAKE_REQUEST(kgem->next_request, kgem->ring); |
list_move_tail(&bo->request, &kgem->next_request->buffers); |
/* XXX is it worth working around gcc here? */ |
kgem->flush |= bo->flush; |
} |
static uint32_t kgem_end_batch(struct kgem *kgem) |
{ |
kgem->batch[kgem->nbatch++] = MI_BATCH_BUFFER_END; |
if (kgem->nbatch & 1) |
kgem->batch[kgem->nbatch++] = MI_NOOP; |
return kgem->nbatch; |
} |
static void kgem_fixup_self_relocs(struct kgem *kgem, struct kgem_bo *bo) |
{ |
int n; |
assert(kgem->nreloc__self <= 256); |
if (kgem->nreloc__self == 0) |
return; |
for (n = 0; n < kgem->nreloc__self; n++) { |
int i = kgem->reloc__self[n]; |
assert(kgem->reloc[i].target_handle == ~0U); |
kgem->reloc[i].target_handle = bo->target_handle; |
kgem->reloc[i].presumed_offset = bo->presumed_offset; |
kgem->batch[kgem->reloc[i].offset/sizeof(kgem->batch[0])] = |
kgem->reloc[i].delta + bo->presumed_offset; |
} |
if (n == 256) { |
for (n = kgem->reloc__self[255]; n < kgem->nreloc; n++) { |
if (kgem->reloc[n].target_handle == ~0U) { |
kgem->reloc[n].target_handle = bo->target_handle; |
kgem->reloc[n].presumed_offset = bo->presumed_offset; |
kgem->batch[kgem->reloc[n].offset/sizeof(kgem->batch[0])] = |
kgem->reloc[n].delta + bo->presumed_offset; |
} |
} |
} |
} |
static void kgem_bo_binding_free(struct kgem *kgem, struct kgem_bo *bo) |
{ |
struct kgem_bo_binding *b; |
b = bo->binding.next; |
while (b) { |
struct kgem_bo_binding *next = b->next; |
free (b); |
b = next; |
} |
} |
static void kgem_bo_release_map(struct kgem *kgem, struct kgem_bo *bo) |
{ |
int type = IS_CPU_MAP(bo->map); |
assert(!IS_USER_MAP(bo->map)); |
DBG(("%s: releasing %s vma for handle=%d, count=%d\n", |
__FUNCTION__, type ? "CPU" : "GTT", |
bo->handle, kgem->vma[type].count)); |
VG(if (type) VALGRIND_MAKE_MEM_NOACCESS(MAP(bo->map), bytes(bo))); |
user_free(MAP(bo->map)); |
bo->map = NULL; |
if (!list_is_empty(&bo->vma)) { |
list_del(&bo->vma); |
kgem->vma[type].count--; |
} |
} |
static void kgem_bo_free(struct kgem *kgem, struct kgem_bo *bo) |
{ |
DBG(("%s: handle=%d\n", __FUNCTION__, bo->handle)); |
assert(bo->refcnt == 0); |
assert(bo->proxy == NULL); |
assert(bo->exec == NULL); |
assert(!bo->snoop || bo->rq == NULL); |
#ifdef DEBUG_MEMORY |
kgem->debug_memory.bo_allocs--; |
kgem->debug_memory.bo_bytes -= bytes(bo); |
#endif |
kgem_bo_binding_free(kgem, bo); |
if (IS_USER_MAP(bo->map)) { |
assert(bo->rq == NULL); |
assert(!__kgem_busy(kgem, bo->handle)); |
assert(MAP(bo->map) != bo || bo->io || bo->flush); |
if (!(bo->io || bo->flush)) { |
DBG(("%s: freeing snooped base\n", __FUNCTION__)); |
assert(bo != MAP(bo->map)); |
free(MAP(bo->map)); |
} |
bo->map = NULL; |
} |
if (bo->map) |
kgem_bo_release_map(kgem, bo); |
assert(list_is_empty(&bo->vma)); |
assert(bo->map == NULL); |
_list_del(&bo->list); |
_list_del(&bo->request); |
gem_close(kgem->fd, bo->handle); |
if (!bo->io) { |
*(struct kgem_bo **)bo = __kgem_freed_bo; |
__kgem_freed_bo = bo; |
} else |
free(bo); |
} |
inline static void kgem_bo_move_to_inactive(struct kgem *kgem, |
struct kgem_bo *bo) |
{ |
DBG(("%s: moving handle=%d to inactive\n", __FUNCTION__, bo->handle)); |
assert(bo->refcnt == 0); |
assert(bo->reusable); |
assert(bo->rq == NULL); |
assert(bo->exec == NULL); |
assert(bo->domain != DOMAIN_GPU); |
assert(!bo->proxy); |
assert(!bo->io); |
assert(!bo->scanout); |
assert(!bo->snoop); |
assert(!bo->flush); |
assert(!bo->needs_flush); |
assert(list_is_empty(&bo->vma)); |
assert_tiling(kgem, bo); |
ASSERT_IDLE(kgem, bo->handle); |
kgem->need_expire = true; |
if (bucket(bo) >= NUM_CACHE_BUCKETS) { |
list_move(&bo->list, &kgem->large_inactive); |
return; |
} |
assert(bo->flush == false); |
list_move(&bo->list, &kgem->inactive[bucket(bo)]); |
if (bo->map) { |
int type = IS_CPU_MAP(bo->map); |
if (bucket(bo) >= NUM_CACHE_BUCKETS || |
(!type && !__kgem_bo_is_mappable(kgem, bo))) { |
// munmap(MAP(bo->map), bytes(bo)); |
bo->map = NULL; |
} |
if (bo->map) { |
list_add(&bo->vma, &kgem->vma[type].inactive[bucket(bo)]); |
kgem->vma[type].count++; |
} |
} |
} |
static struct kgem_bo *kgem_bo_replace_io(struct kgem_bo *bo) |
{ |
struct kgem_bo *base; |
if (!bo->io) |
return bo; |
assert(!bo->snoop); |
base = malloc(sizeof(*base)); |
if (base) { |
DBG(("%s: transferring io handle=%d to bo\n", |
__FUNCTION__, bo->handle)); |
/* transfer the handle to a minimum bo */ |
memcpy(base, bo, sizeof(*base)); |
base->io = false; |
list_init(&base->list); |
list_replace(&bo->request, &base->request); |
list_replace(&bo->vma, &base->vma); |
free(bo); |
bo = base; |
} else |
bo->reusable = false; |
return bo; |
} |
inline static void kgem_bo_remove_from_inactive(struct kgem *kgem, |
struct kgem_bo *bo) |
{ |
DBG(("%s: removing handle=%d from inactive\n", __FUNCTION__, bo->handle)); |
list_del(&bo->list); |
assert(bo->rq == NULL); |
assert(bo->exec == NULL); |
if (bo->map) { |
assert(!list_is_empty(&bo->vma)); |
list_del(&bo->vma); |
kgem->vma[IS_CPU_MAP(bo->map)].count--; |
} |
} |
inline static void kgem_bo_remove_from_active(struct kgem *kgem, |
struct kgem_bo *bo) |
{ |
DBG(("%s: removing handle=%d from active\n", __FUNCTION__, bo->handle)); |
list_del(&bo->list); |
assert(bo->rq != NULL); |
if (bo->rq == (void *)kgem) |
list_del(&bo->request); |
assert(list_is_empty(&bo->vma)); |
} |
static void _kgem_bo_delete_buffer(struct kgem *kgem, struct kgem_bo *bo) |
{ |
struct kgem_buffer *io = (struct kgem_buffer *)bo->proxy; |
DBG(("%s: size=%d, offset=%d, parent used=%d\n", |
__FUNCTION__, bo->size.bytes, bo->delta, io->used)); |
if (ALIGN(bo->delta + bo->size.bytes, UPLOAD_ALIGNMENT) == io->used) |
io->used = bo->delta; |
} |
static void kgem_bo_move_to_scanout(struct kgem *kgem, struct kgem_bo *bo) |
{ |
assert(bo->refcnt == 0); |
assert(bo->scanout); |
assert(bo->delta); |
assert(!bo->flush); |
assert(!bo->snoop); |
assert(!bo->io); |
if (bo->purged) { |
DBG(("%s: discarding purged scanout - external name?\n", |
__FUNCTION__)); |
kgem_bo_free(kgem, bo); |
return; |
} |
DBG(("%s: moving %d [fb %d] to scanout cache, active? %d\n", |
__FUNCTION__, bo->handle, bo->delta, bo->rq != NULL)); |
if (bo->rq) |
list_move_tail(&bo->list, &kgem->scanout); |
else |
list_move(&bo->list, &kgem->scanout); |
} |
static void kgem_bo_move_to_snoop(struct kgem *kgem, struct kgem_bo *bo) |
{ |
assert(bo->reusable); |
assert(!bo->flush); |
assert(!bo->needs_flush); |
assert(bo->refcnt == 0); |
assert(bo->exec == NULL); |
if (num_pages(bo) > kgem->max_cpu_size >> 13) { |
DBG(("%s handle=%d discarding large CPU buffer (%d >%d pages)\n", |
__FUNCTION__, bo->handle, num_pages(bo), kgem->max_cpu_size >> 13)); |
kgem_bo_free(kgem, bo); |
return; |
} |
assert(bo->tiling == I915_TILING_NONE); |
assert(bo->rq == NULL); |
DBG(("%s: moving %d to snoop cachee\n", __FUNCTION__, bo->handle)); |
list_add(&bo->list, &kgem->snoop); |
} |
static struct kgem_bo * |
search_snoop_cache(struct kgem *kgem, unsigned int num_pages, unsigned flags) |
{ |
struct kgem_bo *bo, *first = NULL; |
DBG(("%s: num_pages=%d, flags=%x\n", __FUNCTION__, num_pages, flags)); |
if ((kgem->has_caching | kgem->has_userptr) == 0) |
return NULL; |
if (list_is_empty(&kgem->snoop)) { |
DBG(("%s: inactive and cache empty\n", __FUNCTION__)); |
if (!__kgem_throttle_retire(kgem, flags)) { |
DBG(("%s: nothing retired\n", __FUNCTION__)); |
return NULL; |
} |
} |
list_for_each_entry(bo, &kgem->snoop, list) { |
assert(bo->refcnt == 0); |
assert(bo->snoop); |
assert(!bo->scanout); |
assert(!bo->purged); |
assert(bo->proxy == NULL); |
assert(bo->tiling == I915_TILING_NONE); |
assert(bo->rq == NULL); |
assert(bo->exec == NULL); |
if (num_pages > num_pages(bo)) |
continue; |
if (num_pages(bo) > 2*num_pages) { |
if (first == NULL) |
first = bo; |
continue; |
} |
list_del(&bo->list); |
bo->pitch = 0; |
bo->delta = 0; |
DBG((" %s: found handle=%d (num_pages=%d) in snoop cache\n", |
__FUNCTION__, bo->handle, num_pages(bo))); |
return bo; |
} |
if (first) { |
list_del(&first->list); |
first->pitch = 0; |
first->delta = 0; |
DBG((" %s: found handle=%d (num_pages=%d) in snoop cache\n", |
__FUNCTION__, first->handle, num_pages(first))); |
return first; |
} |
return NULL; |
} |
void kgem_bo_undo(struct kgem *kgem, struct kgem_bo *bo) |
{ |
if (kgem->nexec != 1 || bo->exec == NULL) |
return; |
DBG(("%s: only handle in batch, discarding last operations for handle=%d\n", |
__FUNCTION__, bo->handle)); |
assert(bo->exec == &kgem->exec[0]); |
assert(kgem->exec[0].handle == bo->handle); |
assert(RQ(bo->rq) == kgem->next_request); |
bo->refcnt++; |
kgem_reset(kgem); |
bo->refcnt--; |
} |
static void __kgem_bo_destroy(struct kgem *kgem, struct kgem_bo *bo) |
{ |
DBG(("%s: handle=%d\n", __FUNCTION__, bo->handle)); |
assert(list_is_empty(&bo->list)); |
assert(bo->refcnt == 0); |
assert(!bo->purged || !bo->reusable); |
assert(bo->proxy == NULL); |
assert_tiling(kgem, bo); |
bo->binding.offset = 0; |
if (DBG_NO_CACHE) |
goto destroy; |
if (bo->snoop && !bo->flush) { |
DBG(("%s: handle=%d is snooped\n", __FUNCTION__, bo->handle)); |
assert(bo->reusable); |
assert(list_is_empty(&bo->list)); |
if (bo->exec == NULL && bo->rq && !__kgem_busy(kgem, bo->handle)) |
__kgem_bo_clear_busy(bo); |
if (bo->rq == NULL) |
kgem_bo_move_to_snoop(kgem, bo); |
return; |
} |
if (!IS_USER_MAP(bo->map)) |
bo->flush = false; |
if (bo->scanout) { |
kgem_bo_move_to_scanout(kgem, bo); |
return; |
} |
if (bo->io) |
bo = kgem_bo_replace_io(bo); |
if (!bo->reusable) { |
DBG(("%s: handle=%d, not reusable\n", |
__FUNCTION__, bo->handle)); |
goto destroy; |
} |
if (!kgem->has_llc && IS_CPU_MAP(bo->map) && bo->domain != DOMAIN_CPU) |
kgem_bo_release_map(kgem, bo); |
assert(list_is_empty(&bo->vma)); |
assert(list_is_empty(&bo->list)); |
assert(bo->flush == false); |
assert(bo->snoop == false); |
assert(bo->io == false); |
assert(bo->scanout == false); |
kgem_bo_undo(kgem, bo); |
assert(bo->refcnt == 0); |
if (bo->rq && bo->exec == NULL && !__kgem_busy(kgem, bo->handle)) |
__kgem_bo_clear_busy(bo); |
if (bo->rq) { |
struct list *cache; |
DBG(("%s: handle=%d -> active\n", __FUNCTION__, bo->handle)); |
if (bucket(bo) < NUM_CACHE_BUCKETS) |
cache = &kgem->active[bucket(bo)][bo->tiling]; |
else |
cache = &kgem->large; |
list_add(&bo->list, cache); |
return; |
} |
assert(bo->exec == NULL); |
assert(list_is_empty(&bo->request)); |
if (!IS_CPU_MAP(bo->map)) { |
if (!kgem_bo_set_purgeable(kgem, bo)) |
goto destroy; |
if (!kgem->has_llc && bo->domain == DOMAIN_CPU) |
goto destroy; |
DBG(("%s: handle=%d, purged\n", |
__FUNCTION__, bo->handle)); |
} |
kgem_bo_move_to_inactive(kgem, bo); |
return; |
destroy: |
if (!bo->exec) |
kgem_bo_free(kgem, bo); |
} |
static void kgem_bo_unref(struct kgem *kgem, struct kgem_bo *bo) |
{ |
assert(bo->refcnt); |
if (--bo->refcnt == 0) |
__kgem_bo_destroy(kgem, bo); |
} |
static void kgem_buffer_release(struct kgem *kgem, struct kgem_buffer *bo) |
{ |
while (!list_is_empty(&bo->base.vma)) { |
struct kgem_bo *cached; |
cached = list_first_entry(&bo->base.vma, struct kgem_bo, vma); |
assert(cached->proxy == &bo->base); |
list_del(&cached->vma); |
assert(*(struct kgem_bo **)cached->map == cached); |
*(struct kgem_bo **)cached->map = NULL; |
cached->map = NULL; |
kgem_bo_destroy(kgem, cached); |
} |
} |
static bool kgem_retire__buffers(struct kgem *kgem) |
{ |
bool retired = false; |
while (!list_is_empty(&kgem->active_buffers)) { |
struct kgem_buffer *bo = |
list_last_entry(&kgem->active_buffers, |
struct kgem_buffer, |
base.list); |
if (bo->base.rq) |
break; |
DBG(("%s: releasing upload cache for handle=%d? %d\n", |
__FUNCTION__, bo->base.handle, !list_is_empty(&bo->base.vma))); |
list_del(&bo->base.list); |
kgem_buffer_release(kgem, bo); |
kgem_bo_unref(kgem, &bo->base); |
retired = true; |
} |
return retired; |
} |
static bool kgem_retire__flushing(struct kgem *kgem) |
{ |
struct kgem_bo *bo, *next; |
bool retired = false; |
list_for_each_entry_safe(bo, next, &kgem->flushing, request) { |
assert(bo->rq == (void *)kgem); |
assert(bo->exec == NULL); |
if (__kgem_busy(kgem, bo->handle)) |
break; |
__kgem_bo_clear_busy(bo); |
if (bo->refcnt) |
continue; |
if (bo->snoop) { |
kgem_bo_move_to_snoop(kgem, bo); |
} else if (bo->scanout) { |
kgem_bo_move_to_scanout(kgem, bo); |
} else if ((bo = kgem_bo_replace_io(bo))->reusable && |
kgem_bo_set_purgeable(kgem, bo)) { |
kgem_bo_move_to_inactive(kgem, bo); |
retired = true; |
} else |
kgem_bo_free(kgem, bo); |
} |
#if HAS_DEBUG_FULL |
{ |
int count = 0; |
list_for_each_entry(bo, &kgem->flushing, request) |
count++; |
ErrorF("%s: %d bo on flushing list\n", __FUNCTION__, count); |
} |
#endif |
kgem->need_retire |= !list_is_empty(&kgem->flushing); |
return retired; |
} |
static bool __kgem_retire_rq(struct kgem *kgem, struct kgem_request *rq) |
{ |
bool retired = false; |
DBG(("%s: request %d complete\n", |
__FUNCTION__, rq->bo->handle)); |
while (!list_is_empty(&rq->buffers)) { |
struct kgem_bo *bo; |
bo = list_first_entry(&rq->buffers, |
struct kgem_bo, |
request); |
assert(RQ(bo->rq) == rq); |
assert(bo->exec == NULL); |
assert(bo->domain == DOMAIN_GPU || bo->domain == DOMAIN_NONE); |
list_del(&bo->request); |
if (bo->needs_flush) |
bo->needs_flush = __kgem_busy(kgem, bo->handle); |
if (bo->needs_flush) { |
DBG(("%s: moving %d to flushing\n", |
__FUNCTION__, bo->handle)); |
list_add(&bo->request, &kgem->flushing); |
bo->rq = (void *)kgem; |
continue; |
} |
bo->domain = DOMAIN_NONE; |
bo->rq = NULL; |
if (bo->refcnt) |
continue; |
if (bo->snoop) { |
kgem_bo_move_to_snoop(kgem, bo); |
} else if (bo->scanout) { |
kgem_bo_move_to_scanout(kgem, bo); |
} else if ((bo = kgem_bo_replace_io(bo))->reusable && |
kgem_bo_set_purgeable(kgem, bo)) { |
kgem_bo_move_to_inactive(kgem, bo); |
retired = true; |
} else { |
DBG(("%s: closing %d\n", |
__FUNCTION__, bo->handle)); |
kgem_bo_free(kgem, bo); |
} |
} |
assert(rq->bo->rq == NULL); |
assert(list_is_empty(&rq->bo->request)); |
if (--rq->bo->refcnt == 0) { |
if (kgem_bo_set_purgeable(kgem, rq->bo)) { |
kgem_bo_move_to_inactive(kgem, rq->bo); |
retired = true; |
} else { |
DBG(("%s: closing %d\n", |
__FUNCTION__, rq->bo->handle)); |
kgem_bo_free(kgem, rq->bo); |
} |
} |
__kgem_request_free(rq); |
return retired; |
} |
static bool kgem_retire__requests_ring(struct kgem *kgem, int ring) |
{ |
bool retired = false; |
while (!list_is_empty(&kgem->requests[ring])) { |
struct kgem_request *rq; |
rq = list_first_entry(&kgem->requests[ring], |
struct kgem_request, |
list); |
if (__kgem_busy(kgem, rq->bo->handle)) |
break; |
retired |= __kgem_retire_rq(kgem, rq); |
} |
#if HAS_DEBUG_FULL |
{ |
struct kgem_bo *bo; |
int count = 0; |
list_for_each_entry(bo, &kgem->requests[ring], request) |
count++; |
bo = NULL; |
if (!list_is_empty(&kgem->requests[ring])) |
bo = list_first_entry(&kgem->requests[ring], |
struct kgem_request, |
list)->bo; |
ErrorF("%s: ring=%d, %d outstanding requests, oldest=%d\n", |
__FUNCTION__, ring, count, bo ? bo->handle : 0); |
} |
#endif |
return retired; |
} |
static bool kgem_retire__requests(struct kgem *kgem) |
{ |
bool retired = false; |
int n; |
for (n = 0; n < ARRAY_SIZE(kgem->requests); n++) { |
retired |= kgem_retire__requests_ring(kgem, n); |
kgem->need_retire |= !list_is_empty(&kgem->requests[n]); |
} |
return retired; |
} |
bool kgem_retire(struct kgem *kgem) |
{ |
bool retired = false; |
DBG(("%s\n", __FUNCTION__)); |
kgem->need_retire = false; |
retired |= kgem_retire__flushing(kgem); |
retired |= kgem_retire__requests(kgem); |
retired |= kgem_retire__buffers(kgem); |
DBG(("%s -- retired=%d, need_retire=%d\n", |
__FUNCTION__, retired, kgem->need_retire)); |
kgem->retire(kgem); |
return retired; |
} |
bool __kgem_ring_is_idle(struct kgem *kgem, int ring) |
{ |
struct kgem_request *rq; |
assert(!list_is_empty(&kgem->requests[ring])); |
rq = list_last_entry(&kgem->requests[ring], |
struct kgem_request, list); |
if (__kgem_busy(kgem, rq->bo->handle)) { |
DBG(("%s: last requests handle=%d still busy\n", |
__FUNCTION__, rq->bo->handle)); |
return false; |
} |
DBG(("%s: ring=%d idle (handle=%d)\n", |
__FUNCTION__, ring, rq->bo->handle)); |
kgem_retire__requests_ring(kgem, ring); |
assert(list_is_empty(&kgem->requests[ring])); |
return true; |
} |
static void kgem_commit(struct kgem *kgem) |
{ |
struct kgem_request *rq = kgem->next_request; |
struct kgem_bo *bo, *next; |
list_for_each_entry_safe(bo, next, &rq->buffers, request) { |
assert(next->request.prev == &bo->request); |
DBG(("%s: release handle=%d (proxy? %d), dirty? %d flush? %d, snoop? %d -> offset=%x\n", |
__FUNCTION__, bo->handle, bo->proxy != NULL, |
bo->gpu_dirty, bo->needs_flush, bo->snoop, |
(unsigned)bo->exec->offset)); |
assert(bo->exec); |
assert(bo->proxy == NULL || bo->exec == &_kgem_dummy_exec); |
assert(RQ(bo->rq) == rq || (RQ(bo->proxy->rq) == rq)); |
bo->presumed_offset = bo->exec->offset; |
bo->exec = NULL; |
bo->target_handle = -1; |
if (!bo->refcnt && !bo->reusable) { |
assert(!bo->snoop); |
kgem_bo_free(kgem, bo); |
continue; |
} |
bo->binding.offset = 0; |
bo->domain = DOMAIN_GPU; |
bo->gpu_dirty = false; |
if (bo->proxy) { |
/* proxies are not used for domain tracking */ |
bo->exec = NULL; |
__kgem_bo_clear_busy(bo); |
} |
kgem->scanout_busy |= bo->scanout; |
} |
if (rq == &kgem->static_request) { |
struct drm_i915_gem_set_domain set_domain; |
DBG(("%s: syncing due to allocation failure\n", __FUNCTION__)); |
VG_CLEAR(set_domain); |
set_domain.handle = rq->bo->handle; |
set_domain.read_domains = I915_GEM_DOMAIN_GTT; |
set_domain.write_domain = I915_GEM_DOMAIN_GTT; |
if (drmIoctl(kgem->fd, DRM_IOCTL_I915_GEM_SET_DOMAIN, &set_domain)) { |
DBG(("%s: sync: GPU hang detected\n", __FUNCTION__)); |
kgem_throttle(kgem); |
} |
kgem_retire(kgem); |
assert(list_is_empty(&rq->buffers)); |
assert(rq->bo->map == NULL); |
gem_close(kgem->fd, rq->bo->handle); |
kgem_cleanup_cache(kgem); |
} else { |
list_add_tail(&rq->list, &kgem->requests[rq->ring]); |
kgem->need_throttle = kgem->need_retire = 1; |
} |
kgem->next_request = NULL; |
} |
static void kgem_close_list(struct kgem *kgem, struct list *head) |
{ |
while (!list_is_empty(head)) |
kgem_bo_free(kgem, list_first_entry(head, struct kgem_bo, list)); |
} |
static void kgem_close_inactive(struct kgem *kgem) |
{ |
unsigned int i; |
for (i = 0; i < ARRAY_SIZE(kgem->inactive); i++) |
kgem_close_list(kgem, &kgem->inactive[i]); |
} |
static void kgem_finish_buffers(struct kgem *kgem) |
{ |
struct kgem_buffer *bo, *next; |
list_for_each_entry_safe(bo, next, &kgem->batch_buffers, base.list) { |
DBG(("%s: buffer handle=%d, used=%d, exec?=%d, write=%d, mmapped=%s\n", |
__FUNCTION__, bo->base.handle, bo->used, bo->base.exec!=NULL, |
bo->write, bo->mmapped ? IS_CPU_MAP(bo->base.map) ? "cpu" : "gtt" : "no")); |
assert(next->base.list.prev == &bo->base.list); |
assert(bo->base.io); |
assert(bo->base.refcnt >= 1); |
if (!bo->base.exec) { |
DBG(("%s: skipping unattached handle=%d, used=%d\n", |
__FUNCTION__, bo->base.handle, bo->used)); |
continue; |
} |
if (!bo->write) { |
assert(bo->base.exec || bo->base.refcnt > 1); |
goto decouple; |
} |
if (bo->mmapped) { |
int used; |
assert(!bo->need_io); |
used = ALIGN(bo->used, PAGE_SIZE); |
if (!DBG_NO_UPLOAD_ACTIVE && |
used + PAGE_SIZE <= bytes(&bo->base) && |
(kgem->has_llc || !IS_CPU_MAP(bo->base.map) || bo->base.snoop)) { |
DBG(("%s: retaining upload buffer (%d/%d)\n", |
__FUNCTION__, bo->used, bytes(&bo->base))); |
bo->used = used; |
list_move(&bo->base.list, |
&kgem->active_buffers); |
continue; |
} |
DBG(("%s: discarding mmapped buffer, used=%d, map type=%d\n", |
__FUNCTION__, bo->used, (int)__MAP_TYPE(bo->base.map))); |
goto decouple; |
} |
if (!bo->used) { |
/* Unless we replace the handle in the execbuffer, |
* then this bo will become active. So decouple it |
* from the buffer list and track it in the normal |
* manner. |
*/ |
goto decouple; |
} |
assert(bo->need_io); |
assert(bo->base.rq == MAKE_REQUEST(kgem->next_request, kgem->ring)); |
assert(bo->base.domain != DOMAIN_GPU); |
if (bo->base.refcnt == 1 && |
bo->base.size.pages.count > 1 && |
bo->used < bytes(&bo->base) / 2) { |
struct kgem_bo *shrink; |
unsigned alloc = NUM_PAGES(bo->used); |
shrink = search_snoop_cache(kgem, alloc, |
CREATE_INACTIVE | CREATE_NO_RETIRE); |
if (shrink) { |
void *map; |
int n; |
DBG(("%s: used=%d, shrinking %d to %d, handle %d to %d\n", |
__FUNCTION__, |
bo->used, bytes(&bo->base), bytes(shrink), |
bo->base.handle, shrink->handle)); |
assert(bo->used <= bytes(shrink)); |
map = kgem_bo_map__cpu(kgem, shrink); |
if (map) { |
kgem_bo_sync__cpu(kgem, shrink); |
memcpy(map, bo->mem, bo->used); |
shrink->target_handle = |
kgem->has_handle_lut ? bo->base.target_handle : shrink->handle; |
for (n = 0; n < kgem->nreloc; n++) { |
if (kgem->reloc[n].target_handle == bo->base.target_handle) { |
kgem->reloc[n].target_handle = shrink->target_handle; |
kgem->reloc[n].presumed_offset = shrink->presumed_offset; |
kgem->batch[kgem->reloc[n].offset/sizeof(kgem->batch[0])] = |
kgem->reloc[n].delta + shrink->presumed_offset; |
} |
} |
bo->base.exec->handle = shrink->handle; |
bo->base.exec->offset = shrink->presumed_offset; |
shrink->exec = bo->base.exec; |
shrink->rq = bo->base.rq; |
list_replace(&bo->base.request, |
&shrink->request); |
list_init(&bo->base.request); |
shrink->needs_flush = bo->base.gpu_dirty; |
bo->base.exec = NULL; |
bo->base.rq = NULL; |
bo->base.gpu_dirty = false; |
bo->base.needs_flush = false; |
bo->used = 0; |
goto decouple; |
} |
__kgem_bo_destroy(kgem, shrink); |
} |
shrink = search_linear_cache(kgem, alloc, |
CREATE_INACTIVE | CREATE_NO_RETIRE); |
if (shrink) { |
int n; |
DBG(("%s: used=%d, shrinking %d to %d, handle %d to %d\n", |
__FUNCTION__, |
bo->used, bytes(&bo->base), bytes(shrink), |
bo->base.handle, shrink->handle)); |
assert(bo->used <= bytes(shrink)); |
if (gem_write(kgem->fd, shrink->handle, |
0, bo->used, bo->mem) == 0) { |
shrink->target_handle = |
kgem->has_handle_lut ? bo->base.target_handle : shrink->handle; |
for (n = 0; n < kgem->nreloc; n++) { |
if (kgem->reloc[n].target_handle == bo->base.target_handle) { |
kgem->reloc[n].target_handle = shrink->target_handle; |
kgem->reloc[n].presumed_offset = shrink->presumed_offset; |
kgem->batch[kgem->reloc[n].offset/sizeof(kgem->batch[0])] = |
kgem->reloc[n].delta + shrink->presumed_offset; |
} |
} |
bo->base.exec->handle = shrink->handle; |
bo->base.exec->offset = shrink->presumed_offset; |
shrink->exec = bo->base.exec; |
shrink->rq = bo->base.rq; |
list_replace(&bo->base.request, |
&shrink->request); |
list_init(&bo->base.request); |
shrink->needs_flush = bo->base.gpu_dirty; |
bo->base.exec = NULL; |
bo->base.rq = NULL; |
bo->base.gpu_dirty = false; |
bo->base.needs_flush = false; |
bo->used = 0; |
goto decouple; |
} |
__kgem_bo_destroy(kgem, shrink); |
} |
} |
DBG(("%s: handle=%d, uploading %d/%d\n", |
__FUNCTION__, bo->base.handle, bo->used, bytes(&bo->base))); |
ASSERT_IDLE(kgem, bo->base.handle); |
assert(bo->used <= bytes(&bo->base)); |
gem_write(kgem->fd, bo->base.handle, |
0, bo->used, bo->mem); |
bo->need_io = 0; |
decouple: |
DBG(("%s: releasing handle=%d\n", |
__FUNCTION__, bo->base.handle)); |
list_del(&bo->base.list); |
kgem_bo_unref(kgem, &bo->base); |
} |
} |
static void kgem_cleanup(struct kgem *kgem) |
{ |
int n; |
for (n = 0; n < ARRAY_SIZE(kgem->requests); n++) { |
while (!list_is_empty(&kgem->requests[n])) { |
struct kgem_request *rq; |
rq = list_first_entry(&kgem->requests[n], |
struct kgem_request, |
list); |
while (!list_is_empty(&rq->buffers)) { |
struct kgem_bo *bo; |
bo = list_first_entry(&rq->buffers, |
struct kgem_bo, |
request); |
bo->exec = NULL; |
bo->gpu_dirty = false; |
__kgem_bo_clear_busy(bo); |
if (bo->refcnt == 0) |
kgem_bo_free(kgem, bo); |
} |
__kgem_request_free(rq); |
} |
} |
kgem_close_inactive(kgem); |
} |
static int kgem_batch_write(struct kgem *kgem, uint32_t handle, uint32_t size) |
{ |
int ret; |
ASSERT_IDLE(kgem, handle); |
/* If there is no surface data, just upload the batch */ |
if (kgem->surface == kgem->batch_size) |
return gem_write(kgem->fd, handle, |
0, sizeof(uint32_t)*kgem->nbatch, |
kgem->batch); |
/* Are the batch pages conjoint with the surface pages? */ |
if (kgem->surface < kgem->nbatch + PAGE_SIZE/sizeof(uint32_t)) { |
assert(size == PAGE_ALIGN(kgem->batch_size*sizeof(uint32_t))); |
return gem_write(kgem->fd, handle, |
0, kgem->batch_size*sizeof(uint32_t), |
kgem->batch); |
} |
/* Disjoint surface/batch, upload separately */ |
ret = gem_write(kgem->fd, handle, |
0, sizeof(uint32_t)*kgem->nbatch, |
kgem->batch); |
if (ret) |
return ret; |
ret = PAGE_ALIGN(sizeof(uint32_t) * kgem->batch_size); |
ret -= sizeof(uint32_t) * kgem->surface; |
assert(size-ret >= kgem->nbatch*sizeof(uint32_t)); |
return __gem_write(kgem->fd, handle, |
size - ret, (kgem->batch_size - kgem->surface)*sizeof(uint32_t), |
kgem->batch + kgem->surface); |
} |
void kgem_reset(struct kgem *kgem) |
{ |
if (kgem->next_request) { |
struct kgem_request *rq = kgem->next_request; |
while (!list_is_empty(&rq->buffers)) { |
struct kgem_bo *bo = |
list_first_entry(&rq->buffers, |
struct kgem_bo, |
request); |
list_del(&bo->request); |
assert(RQ(bo->rq) == rq); |
bo->binding.offset = 0; |
bo->exec = NULL; |
bo->target_handle = -1; |
bo->gpu_dirty = false; |
if (bo->needs_flush && __kgem_busy(kgem, bo->handle)) { |
assert(bo->domain == DOMAIN_GPU || bo->domain == DOMAIN_NONE); |
list_add(&bo->request, &kgem->flushing); |
bo->rq = (void *)kgem; |
} else |
__kgem_bo_clear_busy(bo); |
if (bo->refcnt || bo->rq) |
continue; |
if (bo->snoop) { |
kgem_bo_move_to_snoop(kgem, bo); |
} else if (bo->scanout) { |
kgem_bo_move_to_scanout(kgem, bo); |
} else if ((bo = kgem_bo_replace_io(bo))->reusable && |
kgem_bo_set_purgeable(kgem, bo)) { |
kgem_bo_move_to_inactive(kgem, bo); |
} else { |
DBG(("%s: closing %d\n", |
__FUNCTION__, bo->handle)); |
kgem_bo_free(kgem, bo); |
} |
} |
if (rq != &kgem->static_request) { |
list_init(&rq->list); |
__kgem_request_free(rq); |
} |
} |
kgem->nfence = 0; |
kgem->nexec = 0; |
kgem->nreloc = 0; |
kgem->nreloc__self = 0; |
kgem->aperture = 0; |
kgem->aperture_fenced = 0; |
kgem->nbatch = 0; |
kgem->surface = kgem->batch_size; |
kgem->mode = KGEM_NONE; |
kgem->flush = 0; |
kgem->batch_flags = kgem->batch_flags_base; |
kgem->next_request = __kgem_request_alloc(kgem); |
kgem_sna_reset(kgem); |
} |
static int compact_batch_surface(struct kgem *kgem) |
{ |
int size, shrink, n; |
if (!kgem->has_relaxed_delta) |
return kgem->batch_size; |
/* See if we can pack the contents into one or two pages */ |
n = ALIGN(kgem->batch_size, 1024); |
size = n - kgem->surface + kgem->nbatch; |
size = ALIGN(size, 1024); |
shrink = n - size; |
if (shrink) { |
DBG(("shrinking from %d to %d\n", kgem->batch_size, size)); |
shrink *= sizeof(uint32_t); |
for (n = 0; n < kgem->nreloc; n++) { |
if (kgem->reloc[n].read_domains == I915_GEM_DOMAIN_INSTRUCTION && |
kgem->reloc[n].target_handle == ~0U) |
kgem->reloc[n].delta -= shrink; |
if (kgem->reloc[n].offset >= sizeof(uint32_t)*kgem->nbatch) |
kgem->reloc[n].offset -= shrink; |
} |
} |
return size * sizeof(uint32_t); |
} |
static struct kgem_bo * |
kgem_create_batch(struct kgem *kgem, int size) |
{ |
struct drm_i915_gem_set_domain set_domain; |
struct kgem_bo *bo; |
if (size <= 4096) { |
bo = list_first_entry(&kgem->pinned_batches[0], |
struct kgem_bo, |
list); |
if (!bo->rq) { |
out_4096: |
list_move_tail(&bo->list, &kgem->pinned_batches[0]); |
return kgem_bo_reference(bo); |
} |
if (!__kgem_busy(kgem, bo->handle)) { |
assert(RQ(bo->rq)->bo == bo); |
__kgem_retire_rq(kgem, RQ(bo->rq)); |
goto out_4096; |
} |
} |
if (size <= 16384) { |
bo = list_first_entry(&kgem->pinned_batches[1], |
struct kgem_bo, |
list); |
if (!bo->rq) { |
out_16384: |
list_move_tail(&bo->list, &kgem->pinned_batches[1]); |
return kgem_bo_reference(bo); |
} |
if (!__kgem_busy(kgem, bo->handle)) { |
assert(RQ(bo->rq)->bo == bo); |
__kgem_retire_rq(kgem, RQ(bo->rq)); |
goto out_16384; |
} |
} |
if (kgem->gen == 020 && !kgem->has_pinned_batches) { |
assert(size <= 16384); |
bo = list_first_entry(&kgem->pinned_batches[size > 4096], |
struct kgem_bo, |
list); |
list_move_tail(&bo->list, &kgem->pinned_batches[size > 4096]); |
DBG(("%s: syncing due to busy batches\n", __FUNCTION__)); |
VG_CLEAR(set_domain); |
set_domain.handle = bo->handle; |
set_domain.read_domains = I915_GEM_DOMAIN_GTT; |
set_domain.write_domain = I915_GEM_DOMAIN_GTT; |
if (drmIoctl(kgem->fd, DRM_IOCTL_I915_GEM_SET_DOMAIN, &set_domain)) { |
DBG(("%s: sync: GPU hang detected\n", __FUNCTION__)); |
kgem_throttle(kgem); |
return NULL; |
} |
kgem_retire(kgem); |
assert(bo->rq == NULL); |
return kgem_bo_reference(bo); |
} |
return kgem_create_linear(kgem, size, CREATE_NO_THROTTLE); |
} |
void _kgem_submit(struct kgem *kgem) |
{ |
struct kgem_request *rq; |
uint32_t batch_end; |
int size; |
assert(!DBG_NO_HW); |
assert(!kgem->wedged); |
assert(kgem->nbatch); |
assert(kgem->nbatch <= KGEM_BATCH_SIZE(kgem)); |
assert(kgem->nbatch <= kgem->surface); |
batch_end = kgem_end_batch(kgem); |
kgem_sna_flush(kgem); |
DBG(("batch[%d/%d, flags=%x]: %d %d %d %d, nreloc=%d, nexec=%d, nfence=%d, aperture=%d\n", |
kgem->mode, kgem->ring, kgem->batch_flags, |
batch_end, kgem->nbatch, kgem->surface, kgem->batch_size, |
kgem->nreloc, kgem->nexec, kgem->nfence, kgem->aperture)); |
assert(kgem->nbatch <= kgem->batch_size); |
assert(kgem->nbatch <= kgem->surface); |
assert(kgem->nreloc <= ARRAY_SIZE(kgem->reloc)); |
assert(kgem->nexec < ARRAY_SIZE(kgem->exec)); |
assert(kgem->nfence <= kgem->fence_max); |
kgem_finish_buffers(kgem); |
#if SHOW_BATCH |
__kgem_batch_debug(kgem, batch_end); |
#endif |
rq = kgem->next_request; |
if (kgem->surface != kgem->batch_size) |
size = compact_batch_surface(kgem); |
else |
size = kgem->nbatch * sizeof(kgem->batch[0]); |
rq->bo = kgem_create_batch(kgem, size); |
if (rq->bo) { |
uint32_t handle = rq->bo->handle; |
int i; |
assert(!rq->bo->needs_flush); |
i = kgem->nexec++; |
kgem->exec[i].handle = handle; |
kgem->exec[i].relocation_count = kgem->nreloc; |
kgem->exec[i].relocs_ptr = (uintptr_t)kgem->reloc; |
kgem->exec[i].alignment = 0; |
kgem->exec[i].offset = rq->bo->presumed_offset; |
kgem->exec[i].flags = 0; |
kgem->exec[i].rsvd1 = 0; |
kgem->exec[i].rsvd2 = 0; |
rq->bo->target_handle = kgem->has_handle_lut ? i : handle; |
rq->bo->exec = &kgem->exec[i]; |
rq->bo->rq = MAKE_REQUEST(rq, kgem->ring); /* useful sanity check */ |
list_add(&rq->bo->request, &rq->buffers); |
rq->ring = kgem->ring == KGEM_BLT; |
kgem_fixup_self_relocs(kgem, rq->bo); |
if (kgem_batch_write(kgem, handle, size) == 0) { |
struct drm_i915_gem_execbuffer2 execbuf; |
int ret, retry = 3; |
memset(&execbuf, 0, sizeof(execbuf)); |
execbuf.buffers_ptr = (uintptr_t)kgem->exec; |
execbuf.buffer_count = kgem->nexec; |
execbuf.batch_len = batch_end*sizeof(uint32_t); |
execbuf.flags = kgem->ring | kgem->batch_flags; |
if (DEBUG_DUMP) |
{ |
int fd = open("/tmp1/1/batchbuffer.bin", O_CREAT|O_WRONLY|O_BINARY); |
if (fd != -1) { |
write(fd, kgem->batch, size); |
close(fd); |
} |
else printf("SNA: failed to write batchbuffer\n"); |
asm volatile("int3"); |
} |
ret = drmIoctl(kgem->fd, |
DRM_IOCTL_I915_GEM_EXECBUFFER2, |
&execbuf); |
while (ret == -1 && errno == EBUSY && retry--) { |
__kgem_throttle(kgem); |
ret = drmIoctl(kgem->fd, |
DRM_IOCTL_I915_GEM_EXECBUFFER2, |
&execbuf); |
} |
if (DEBUG_SYNC && ret == 0) { |
struct drm_i915_gem_set_domain set_domain; |
VG_CLEAR(set_domain); |
set_domain.handle = handle; |
set_domain.read_domains = I915_GEM_DOMAIN_GTT; |
set_domain.write_domain = I915_GEM_DOMAIN_GTT; |
ret = drmIoctl(kgem->fd, DRM_IOCTL_I915_GEM_SET_DOMAIN, &set_domain); |
} |
if (ret == -1) { |
DBG(("%s: GPU hang detected [%d]\n", |
__FUNCTION__, errno)); |
kgem_throttle(kgem); |
kgem->wedged = true; |
#if 0 |
ret = errno; |
ErrorF("batch[%d/%d]: %d %d %d, nreloc=%d, nexec=%d, nfence=%d, aperture=%d: errno=%d\n", |
kgem->mode, kgem->ring, batch_end, kgem->nbatch, kgem->surface, |
kgem->nreloc, kgem->nexec, kgem->nfence, kgem->aperture, errno); |
for (i = 0; i < kgem->nexec; i++) { |
struct kgem_bo *bo, *found = NULL; |
list_for_each_entry(bo, &kgem->next_request->buffers, request) { |
if (bo->handle == kgem->exec[i].handle) { |
found = bo; |
break; |
} |
} |
ErrorF("exec[%d] = handle:%d, presumed offset: %x, size: %d, tiling %d, fenced %d, snooped %d, deleted %d\n", |
i, |
kgem->exec[i].handle, |
(int)kgem->exec[i].offset, |
found ? kgem_bo_size(found) : -1, |
found ? found->tiling : -1, |
(int)(kgem->exec[i].flags & EXEC_OBJECT_NEEDS_FENCE), |
found ? found->snoop : -1, |
found ? found->purged : -1); |
} |
for (i = 0; i < kgem->nreloc; i++) { |
ErrorF("reloc[%d] = pos:%d, target:%d, delta:%d, read:%x, write:%x, offset:%x\n", |
i, |
(int)kgem->reloc[i].offset, |
kgem->reloc[i].target_handle, |
kgem->reloc[i].delta, |
kgem->reloc[i].read_domains, |
kgem->reloc[i].write_domain, |
(int)kgem->reloc[i].presumed_offset); |
} |
if (DEBUG_SYNC) { |
int fd = open("/tmp/batchbuffer", O_WRONLY | O_CREAT | O_APPEND, 0666); |
if (fd != -1) { |
write(fd, kgem->batch, batch_end*sizeof(uint32_t)); |
close(fd); |
} |
FatalError("SNA: failed to submit batchbuffer, errno=%d\n", ret); |
} |
#endif |
} |
} |
kgem_commit(kgem); |
} |
if (kgem->wedged) |
kgem_cleanup(kgem); |
kgem_reset(kgem); |
assert(kgem->next_request != NULL); |
} |
void kgem_throttle(struct kgem *kgem) |
{ |
kgem->need_throttle = 0; |
if (kgem->wedged) |
return; |
kgem->wedged = __kgem_throttle(kgem); |
if (kgem->wedged) { |
printf("Detected a hung GPU, disabling acceleration.\n"); |
printf("When reporting this, please include i915_error_state from debugfs and the full dmesg.\n"); |
} |
} |
void kgem_purge_cache(struct kgem *kgem) |
{ |
struct kgem_bo *bo, *next; |
int i; |
for (i = 0; i < ARRAY_SIZE(kgem->inactive); i++) { |
list_for_each_entry_safe(bo, next, &kgem->inactive[i], list) { |
if (!kgem_bo_is_retained(kgem, bo)) { |
DBG(("%s: purging %d\n", |
__FUNCTION__, bo->handle)); |
kgem_bo_free(kgem, bo); |
} |
} |
} |
kgem->need_purge = false; |
} |
void kgem_clean_large_cache(struct kgem *kgem) |
{ |
while (!list_is_empty(&kgem->large_inactive)) { |
kgem_bo_free(kgem, |
list_first_entry(&kgem->large_inactive, |
struct kgem_bo, list)); |
} |
} |
bool kgem_expire_cache(struct kgem *kgem) |
{ |
time_t now, expire; |
struct kgem_bo *bo; |
unsigned int size = 0, count = 0; |
bool idle; |
unsigned int i; |
time(&now); |
while (__kgem_freed_bo) { |
bo = __kgem_freed_bo; |
__kgem_freed_bo = *(struct kgem_bo **)bo; |
free(bo); |
} |
while (__kgem_freed_request) { |
struct kgem_request *rq = __kgem_freed_request; |
__kgem_freed_request = *(struct kgem_request **)rq; |
free(rq); |
} |
kgem_clean_large_cache(kgem); |
expire = 0; |
list_for_each_entry(bo, &kgem->snoop, list) { |
if (bo->delta) { |
expire = now - MAX_INACTIVE_TIME/2; |
break; |
} |
bo->delta = now; |
} |
if (expire) { |
while (!list_is_empty(&kgem->snoop)) { |
bo = list_last_entry(&kgem->snoop, struct kgem_bo, list); |
if (bo->delta > expire) |
break; |
kgem_bo_free(kgem, bo); |
} |
} |
#ifdef DEBUG_MEMORY |
{ |
long snoop_size = 0; |
int snoop_count = 0; |
list_for_each_entry(bo, &kgem->snoop, list) |
snoop_count++, snoop_size += bytes(bo); |
ErrorF("%s: still allocated %d bo, %ld bytes, in snoop cache\n", |
__FUNCTION__, snoop_count, snoop_size); |
} |
#endif |
kgem_retire(kgem); |
if (kgem->wedged) |
kgem_cleanup(kgem); |
kgem->expire(kgem); |
if (kgem->need_purge) |
kgem_purge_cache(kgem); |
expire = 0; |
idle = !kgem->need_retire; |
for (i = 0; i < ARRAY_SIZE(kgem->inactive); i++) { |
idle &= list_is_empty(&kgem->inactive[i]); |
list_for_each_entry(bo, &kgem->inactive[i], list) { |
if (bo->delta) { |
expire = now - MAX_INACTIVE_TIME; |
break; |
} |
bo->delta = now; |
} |
} |
if (idle) { |
DBG(("%s: idle\n", __FUNCTION__)); |
kgem->need_expire = false; |
return false; |
} |
if (expire == 0) |
return true; |
idle = !kgem->need_retire; |
for (i = 0; i < ARRAY_SIZE(kgem->inactive); i++) { |
struct list preserve; |
list_init(&preserve); |
while (!list_is_empty(&kgem->inactive[i])) { |
bo = list_last_entry(&kgem->inactive[i], |
struct kgem_bo, list); |
if (bo->delta > expire) { |
idle = false; |
break; |
} |
if (bo->map && bo->delta + MAP_PRESERVE_TIME > expire) { |
idle = false; |
list_move_tail(&bo->list, &preserve); |
} else { |
count++; |
size += bytes(bo); |
kgem_bo_free(kgem, bo); |
DBG(("%s: expiring %d\n", |
__FUNCTION__, bo->handle)); |
} |
} |
if (!list_is_empty(&preserve)) { |
preserve.prev->next = kgem->inactive[i].next; |
kgem->inactive[i].next->prev = preserve.prev; |
kgem->inactive[i].next = preserve.next; |
preserve.next->prev = &kgem->inactive[i]; |
} |
} |
#ifdef DEBUG_MEMORY |
{ |
long inactive_size = 0; |
int inactive_count = 0; |
for (i = 0; i < ARRAY_SIZE(kgem->inactive); i++) |
list_for_each_entry(bo, &kgem->inactive[i], list) |
inactive_count++, inactive_size += bytes(bo); |
ErrorF("%s: still allocated %d bo, %ld bytes, in inactive cache\n", |
__FUNCTION__, inactive_count, inactive_size); |
} |
#endif |
DBG(("%s: expired %d objects, %d bytes, idle? %d\n", |
__FUNCTION__, count, size, idle)); |
kgem->need_expire = !idle; |
return !idle; |
(void)count; |
(void)size; |
} |
void kgem_cleanup_cache(struct kgem *kgem) |
{ |
unsigned int i; |
int n; |
/* sync to the most recent request */ |
for (n = 0; n < ARRAY_SIZE(kgem->requests); n++) { |
if (!list_is_empty(&kgem->requests[n])) { |
struct kgem_request *rq; |
struct drm_i915_gem_set_domain set_domain; |
rq = list_first_entry(&kgem->requests[n], |
struct kgem_request, |
list); |
DBG(("%s: sync on cleanup\n", __FUNCTION__)); |
VG_CLEAR(set_domain); |
set_domain.handle = rq->bo->handle; |
set_domain.read_domains = I915_GEM_DOMAIN_GTT; |
set_domain.write_domain = I915_GEM_DOMAIN_GTT; |
(void)drmIoctl(kgem->fd, |
DRM_IOCTL_I915_GEM_SET_DOMAIN, |
&set_domain); |
} |
} |
kgem_retire(kgem); |
kgem_cleanup(kgem); |
for (i = 0; i < ARRAY_SIZE(kgem->inactive); i++) { |
while (!list_is_empty(&kgem->inactive[i])) |
kgem_bo_free(kgem, |
list_last_entry(&kgem->inactive[i], |
struct kgem_bo, list)); |
} |
kgem_clean_large_cache(kgem); |
while (!list_is_empty(&kgem->snoop)) |
kgem_bo_free(kgem, |
list_last_entry(&kgem->snoop, |
struct kgem_bo, list)); |
while (__kgem_freed_bo) { |
struct kgem_bo *bo = __kgem_freed_bo; |
__kgem_freed_bo = *(struct kgem_bo **)bo; |
free(bo); |
} |
kgem->need_purge = false; |
kgem->need_expire = false; |
} |
static struct kgem_bo * |
search_linear_cache(struct kgem *kgem, unsigned int num_pages, unsigned flags) |
{ |
struct kgem_bo *bo, *first = NULL; |
bool use_active = (flags & CREATE_INACTIVE) == 0; |
struct list *cache; |
DBG(("%s: num_pages=%d, flags=%x, use_active? %d, use_large=%d [max=%d]\n", |
__FUNCTION__, num_pages, flags, use_active, |
num_pages >= MAX_CACHE_SIZE / PAGE_SIZE, |
MAX_CACHE_SIZE / PAGE_SIZE)); |
assert(num_pages); |
if (num_pages >= MAX_CACHE_SIZE / PAGE_SIZE) { |
DBG(("%s: searching large buffers\n", __FUNCTION__)); |
retry_large: |
cache = use_active ? &kgem->large : &kgem->large_inactive; |
list_for_each_entry_safe(bo, first, cache, list) { |
assert(bo->refcnt == 0); |
assert(bo->reusable); |
assert(!bo->scanout); |
if (num_pages > num_pages(bo)) |
goto discard; |
if (bo->tiling != I915_TILING_NONE) { |
if (use_active) |
goto discard; |
if (!gem_set_tiling(kgem->fd, bo->handle, |
I915_TILING_NONE, 0)) |
goto discard; |
bo->tiling = I915_TILING_NONE; |
bo->pitch = 0; |
} |
if (bo->purged && !kgem_bo_clear_purgeable(kgem, bo)) |
goto discard; |
list_del(&bo->list); |
if (bo->rq == (void *)kgem) |
list_del(&bo->request); |
bo->delta = 0; |
assert_tiling(kgem, bo); |
return bo; |
discard: |
if (!use_active) |
kgem_bo_free(kgem, bo); |
} |
if (use_active) { |
use_active = false; |
goto retry_large; |
} |
if (__kgem_throttle_retire(kgem, flags)) |
goto retry_large; |
return NULL; |
} |
if (!use_active && list_is_empty(inactive(kgem, num_pages))) { |
DBG(("%s: inactive and cache bucket empty\n", |
__FUNCTION__)); |
if (flags & CREATE_NO_RETIRE) { |
DBG(("%s: can not retire\n", __FUNCTION__)); |
return NULL; |
} |
if (list_is_empty(active(kgem, num_pages, I915_TILING_NONE))) { |
DBG(("%s: active cache bucket empty\n", __FUNCTION__)); |
return NULL; |
} |
if (!__kgem_throttle_retire(kgem, flags)) { |
DBG(("%s: nothing retired\n", __FUNCTION__)); |
return NULL; |
} |
if (list_is_empty(inactive(kgem, num_pages))) { |
DBG(("%s: active cache bucket still empty after retire\n", |
__FUNCTION__)); |
return NULL; |
} |
} |
if (!use_active && flags & (CREATE_CPU_MAP | CREATE_GTT_MAP)) { |
int for_cpu = !!(flags & CREATE_CPU_MAP); |
DBG(("%s: searching for inactive %s map\n", |
__FUNCTION__, for_cpu ? "cpu" : "gtt")); |
cache = &kgem->vma[for_cpu].inactive[cache_bucket(num_pages)]; |
list_for_each_entry(bo, cache, vma) { |
assert(IS_CPU_MAP(bo->map) == for_cpu); |
assert(bucket(bo) == cache_bucket(num_pages)); |
assert(bo->proxy == NULL); |
assert(bo->rq == NULL); |
assert(bo->exec == NULL); |
assert(!bo->scanout); |
if (num_pages > num_pages(bo)) { |
DBG(("inactive too small: %d < %d\n", |
num_pages(bo), num_pages)); |
continue; |
} |
if (bo->purged && !kgem_bo_clear_purgeable(kgem, bo)) { |
kgem_bo_free(kgem, bo); |
break; |
} |
if (I915_TILING_NONE != bo->tiling && |
!gem_set_tiling(kgem->fd, bo->handle, |
I915_TILING_NONE, 0)) |
continue; |
kgem_bo_remove_from_inactive(kgem, bo); |
bo->tiling = I915_TILING_NONE; |
bo->pitch = 0; |
bo->delta = 0; |
DBG((" %s: found handle=%d (num_pages=%d) in linear vma cache\n", |
__FUNCTION__, bo->handle, num_pages(bo))); |
assert(use_active || bo->domain != DOMAIN_GPU); |
assert(!bo->needs_flush); |
assert_tiling(kgem, bo); |
ASSERT_MAYBE_IDLE(kgem, bo->handle, !use_active); |
return bo; |
} |
if (flags & CREATE_EXACT) |
return NULL; |
if (flags & CREATE_CPU_MAP && !kgem->has_llc) |
return NULL; |
} |
cache = use_active ? active(kgem, num_pages, I915_TILING_NONE) : inactive(kgem, num_pages); |
list_for_each_entry(bo, cache, list) { |
assert(bo->refcnt == 0); |
assert(bo->reusable); |
assert(!!bo->rq == !!use_active); |
assert(bo->proxy == NULL); |
assert(!bo->scanout); |
if (num_pages > num_pages(bo)) |
continue; |
if (use_active && |
kgem->gen <= 040 && |
bo->tiling != I915_TILING_NONE) |
continue; |
if (bo->purged && !kgem_bo_clear_purgeable(kgem, bo)) { |
kgem_bo_free(kgem, bo); |
break; |
} |
if (I915_TILING_NONE != bo->tiling) { |
if (flags & (CREATE_CPU_MAP | CREATE_GTT_MAP)) |
continue; |
if (first) |
continue; |
if (!gem_set_tiling(kgem->fd, bo->handle, |
I915_TILING_NONE, 0)) |
continue; |
bo->tiling = I915_TILING_NONE; |
bo->pitch = 0; |
} |
if (bo->map) { |
if (flags & (CREATE_CPU_MAP | CREATE_GTT_MAP)) { |
int for_cpu = !!(flags & CREATE_CPU_MAP); |
if (IS_CPU_MAP(bo->map) != for_cpu) { |
if (first != NULL) |
break; |
first = bo; |
continue; |
} |
} else { |
if (first != NULL) |
break; |
first = bo; |
continue; |
} |
} else { |
if (flags & (CREATE_CPU_MAP | CREATE_GTT_MAP)) { |
if (first != NULL) |
break; |
first = bo; |
continue; |
} |
} |
if (use_active) |
kgem_bo_remove_from_active(kgem, bo); |
else |
kgem_bo_remove_from_inactive(kgem, bo); |
assert(bo->tiling == I915_TILING_NONE); |
bo->pitch = 0; |
bo->delta = 0; |
DBG((" %s: found handle=%d (num_pages=%d) in linear %s cache\n", |
__FUNCTION__, bo->handle, num_pages(bo), |
use_active ? "active" : "inactive")); |
assert(list_is_empty(&bo->list)); |
assert(use_active || bo->domain != DOMAIN_GPU); |
assert(!bo->needs_flush || use_active); |
assert_tiling(kgem, bo); |
ASSERT_MAYBE_IDLE(kgem, bo->handle, !use_active); |
return bo; |
} |
if (first) { |
assert(first->tiling == I915_TILING_NONE); |
if (use_active) |
kgem_bo_remove_from_active(kgem, first); |
else |
kgem_bo_remove_from_inactive(kgem, first); |
first->pitch = 0; |
first->delta = 0; |
DBG((" %s: found handle=%d (near-miss) (num_pages=%d) in linear %s cache\n", |
__FUNCTION__, first->handle, num_pages(first), |
use_active ? "active" : "inactive")); |
assert(list_is_empty(&first->list)); |
assert(use_active || first->domain != DOMAIN_GPU); |
assert(!first->needs_flush || use_active); |
ASSERT_MAYBE_IDLE(kgem, first->handle, !use_active); |
return first; |
} |
return NULL; |
} |
struct kgem_bo *kgem_create_linear(struct kgem *kgem, int size, unsigned flags) |
{ |
struct kgem_bo *bo; |
uint32_t handle; |
DBG(("%s(%d)\n", __FUNCTION__, size)); |
assert(size); |
if (flags & CREATE_GTT_MAP && kgem->has_llc) { |
flags &= ~CREATE_GTT_MAP; |
flags |= CREATE_CPU_MAP; |
} |
size = NUM_PAGES(size); |
bo = search_linear_cache(kgem, size, CREATE_INACTIVE | flags); |
if (bo) { |
assert(bo->domain != DOMAIN_GPU); |
ASSERT_IDLE(kgem, bo->handle); |
bo->refcnt = 1; |
return bo; |
} |
if (flags & CREATE_CACHED) |
return NULL; |
handle = gem_create(kgem->fd, size); |
if (handle == 0) |
return NULL; |
DBG(("%s: new handle=%d, num_pages=%d\n", __FUNCTION__, handle, size)); |
bo = __kgem_bo_alloc(handle, size); |
if (bo == NULL) { |
gem_close(kgem->fd, handle); |
return NULL; |
} |
debug_alloc__bo(kgem, bo); |
return bo; |
} |
inline int kgem_bo_fenced_size(struct kgem *kgem, struct kgem_bo *bo) |
{ |
unsigned int size; |
assert(bo->tiling); |
assert_tiling(kgem, bo); |
assert(kgem->gen < 040); |
if (kgem->gen < 030) |
size = 512 * 1024; |
else |
size = 1024 * 1024; |
while (size < bytes(bo)) |
size *= 2; |
return size; |
} |
struct kgem_bo *kgem_create_2d(struct kgem *kgem, |
int width, |
int height, |
int bpp, |
int tiling, |
uint32_t flags) |
{ |
struct list *cache; |
struct kgem_bo *bo; |
uint32_t pitch, tiled_height, size; |
uint32_t handle; |
int i, bucket, retry; |
bool exact = flags & (CREATE_EXACT | CREATE_SCANOUT); |
if (tiling < 0) |
exact = true, tiling = -tiling; |
DBG(("%s(%dx%d, bpp=%d, tiling=%d, exact=%d, inactive=%d, cpu-mapping=%d, gtt-mapping=%d, scanout?=%d, prime?=%d, temp?=%d)\n", __FUNCTION__, |
width, height, bpp, tiling, exact, |
!!(flags & CREATE_INACTIVE), |
!!(flags & CREATE_CPU_MAP), |
!!(flags & CREATE_GTT_MAP), |
!!(flags & CREATE_SCANOUT), |
!!(flags & CREATE_PRIME), |
!!(flags & CREATE_TEMPORARY))); |
size = kgem_surface_size(kgem, kgem->has_relaxed_fencing, flags, |
width, height, bpp, tiling, &pitch); |
assert(size && size <= kgem->max_object_size); |
size /= PAGE_SIZE; |
bucket = cache_bucket(size); |
if (flags & CREATE_SCANOUT) { |
struct kgem_bo *last = NULL; |
list_for_each_entry_reverse(bo, &kgem->scanout, list) { |
assert(bo->scanout); |
assert(bo->delta); |
assert(!bo->flush); |
assert_tiling(kgem, bo); |
if (size > num_pages(bo) || num_pages(bo) > 2*size) |
continue; |
if (bo->tiling != tiling || |
(tiling != I915_TILING_NONE && bo->pitch != pitch)) { |
if (!gem_set_tiling(kgem->fd, bo->handle, |
tiling, pitch)) |
continue; |
bo->tiling = tiling; |
bo->pitch = pitch; |
} |
if (flags & CREATE_INACTIVE && bo->rq) { |
last = bo; |
continue; |
} |
list_del(&bo->list); |
bo->unique_id = kgem_get_unique_id(kgem); |
DBG((" 1:from scanout: pitch=%d, tiling=%d, handle=%d, id=%d\n", |
bo->pitch, bo->tiling, bo->handle, bo->unique_id)); |
assert(bo->pitch*kgem_aligned_height(kgem, height, bo->tiling) <= kgem_bo_size(bo)); |
assert_tiling(kgem, bo); |
bo->refcnt = 1; |
return bo; |
} |
if (last) { |
list_del(&last->list); |
last->unique_id = kgem_get_unique_id(kgem); |
DBG((" 1:from scanout: pitch=%d, tiling=%d, handle=%d, id=%d\n", |
last->pitch, last->tiling, last->handle, last->unique_id)); |
assert(last->pitch*kgem_aligned_height(kgem, height, last->tiling) <= kgem_bo_size(last)); |
assert_tiling(kgem, last); |
last->refcnt = 1; |
return last; |
} |
bo = NULL; //__kgem_bo_create_as_display(kgem, size, tiling, pitch); |
if (bo) |
return bo; |
} |
if (bucket >= NUM_CACHE_BUCKETS) { |
DBG(("%s: large bo num pages=%d, bucket=%d\n", |
__FUNCTION__, size, bucket)); |
if (flags & CREATE_INACTIVE) |
goto large_inactive; |
tiled_height = kgem_aligned_height(kgem, height, tiling); |
list_for_each_entry(bo, &kgem->large, list) { |
assert(!bo->purged); |
assert(!bo->scanout); |
assert(bo->refcnt == 0); |
assert(bo->reusable); |
assert_tiling(kgem, bo); |
if (kgem->gen < 040) { |
if (bo->pitch < pitch) { |
DBG(("tiled and pitch too small: tiling=%d, (want %d), pitch=%d, need %d\n", |
bo->tiling, tiling, |
bo->pitch, pitch)); |
continue; |
} |
if (bo->pitch * tiled_height > bytes(bo)) |
continue; |
} else { |
if (num_pages(bo) < size) |
continue; |
if (bo->pitch != pitch || bo->tiling != tiling) { |
if (!gem_set_tiling(kgem->fd, bo->handle, |
tiling, pitch)) |
continue; |
bo->pitch = pitch; |
bo->tiling = tiling; |
} |
} |
kgem_bo_remove_from_active(kgem, bo); |
bo->unique_id = kgem_get_unique_id(kgem); |
bo->delta = 0; |
DBG((" 1:from active: pitch=%d, tiling=%d, handle=%d, id=%d\n", |
bo->pitch, bo->tiling, bo->handle, bo->unique_id)); |
assert(bo->pitch*kgem_aligned_height(kgem, height, bo->tiling) <= kgem_bo_size(bo)); |
assert_tiling(kgem, bo); |
bo->refcnt = 1; |
bo->flush = true; |
return bo; |
} |
large_inactive: |
__kgem_throttle_retire(kgem, flags); |
list_for_each_entry(bo, &kgem->large_inactive, list) { |
assert(bo->refcnt == 0); |
assert(bo->reusable); |
assert(!bo->scanout); |
assert_tiling(kgem, bo); |
if (size > num_pages(bo)) |
continue; |
if (bo->tiling != tiling || |
(tiling != I915_TILING_NONE && bo->pitch != pitch)) { |
if (!gem_set_tiling(kgem->fd, bo->handle, |
tiling, pitch)) |
continue; |
bo->tiling = tiling; |
bo->pitch = pitch; |
} |
if (bo->purged && !kgem_bo_clear_purgeable(kgem, bo)) { |
kgem_bo_free(kgem, bo); |
break; |
} |
list_del(&bo->list); |
assert(bo->domain != DOMAIN_GPU); |
bo->unique_id = kgem_get_unique_id(kgem); |
bo->pitch = pitch; |
bo->delta = 0; |
DBG((" 1:from large inactive: pitch=%d, tiling=%d, handle=%d, id=%d\n", |
bo->pitch, bo->tiling, bo->handle, bo->unique_id)); |
assert(bo->pitch*kgem_aligned_height(kgem, height, bo->tiling) <= kgem_bo_size(bo)); |
assert_tiling(kgem, bo); |
bo->refcnt = 1; |
return bo; |
} |
goto create; |
} |
if (flags & (CREATE_CPU_MAP | CREATE_GTT_MAP)) { |
int for_cpu = !!(flags & CREATE_CPU_MAP); |
if (kgem->has_llc && tiling == I915_TILING_NONE) |
for_cpu = 1; |
/* We presume that we will need to upload to this bo, |
* and so would prefer to have an active VMA. |
*/ |
cache = &kgem->vma[for_cpu].inactive[bucket]; |
do { |
list_for_each_entry(bo, cache, vma) { |
assert(bucket(bo) == bucket); |
assert(bo->refcnt == 0); |
assert(!bo->scanout); |
assert(bo->map); |
assert(IS_CPU_MAP(bo->map) == for_cpu); |
assert(bo->rq == NULL); |
assert(list_is_empty(&bo->request)); |
assert(bo->flush == false); |
assert_tiling(kgem, bo); |
if (size > num_pages(bo)) { |
DBG(("inactive too small: %d < %d\n", |
num_pages(bo), size)); |
continue; |
} |
if (bo->tiling != tiling || |
(tiling != I915_TILING_NONE && bo->pitch != pitch)) { |
DBG(("inactive vma with wrong tiling: %d < %d\n", |
bo->tiling, tiling)); |
continue; |
} |
if (bo->purged && !kgem_bo_clear_purgeable(kgem, bo)) { |
kgem_bo_free(kgem, bo); |
break; |
} |
assert(bo->tiling == tiling); |
bo->pitch = pitch; |
bo->delta = 0; |
bo->unique_id = kgem_get_unique_id(kgem); |
bo->domain = DOMAIN_NONE; |
kgem_bo_remove_from_inactive(kgem, bo); |
DBG((" from inactive vma: pitch=%d, tiling=%d: handle=%d, id=%d\n", |
bo->pitch, bo->tiling, bo->handle, bo->unique_id)); |
assert(bo->reusable); |
assert(bo->domain != DOMAIN_GPU); |
ASSERT_IDLE(kgem, bo->handle); |
assert(bo->pitch*kgem_aligned_height(kgem, height, bo->tiling) <= kgem_bo_size(bo)); |
assert_tiling(kgem, bo); |
bo->refcnt = 1; |
return bo; |
} |
} while (!list_is_empty(cache) && |
__kgem_throttle_retire(kgem, flags)); |
if (flags & CREATE_CPU_MAP && !kgem->has_llc) { |
if (list_is_empty(&kgem->active[bucket][tiling]) && |
list_is_empty(&kgem->inactive[bucket])) |
flags &= ~CREATE_CACHED; |
goto create; |
} |
} |
if (flags & CREATE_INACTIVE) |
goto skip_active_search; |
/* Best active match */ |
retry = NUM_CACHE_BUCKETS - bucket; |
if (retry > 3 && (flags & CREATE_TEMPORARY) == 0) |
retry = 3; |
search_again: |
assert(bucket < NUM_CACHE_BUCKETS); |
cache = &kgem->active[bucket][tiling]; |
if (tiling) { |
tiled_height = kgem_aligned_height(kgem, height, tiling); |
list_for_each_entry(bo, cache, list) { |
assert(!bo->purged); |
assert(bo->refcnt == 0); |
assert(bucket(bo) == bucket); |
assert(bo->reusable); |
assert(bo->tiling == tiling); |
assert(bo->flush == false); |
assert(!bo->scanout); |
assert_tiling(kgem, bo); |
if (kgem->gen < 040) { |
if (bo->pitch < pitch) { |
DBG(("tiled and pitch too small: tiling=%d, (want %d), pitch=%d, need %d\n", |
bo->tiling, tiling, |
bo->pitch, pitch)); |
continue; |
} |
if (bo->pitch * tiled_height > bytes(bo)) |
continue; |
} else { |
if (num_pages(bo) < size) |
continue; |
if (bo->pitch != pitch) { |
if (!gem_set_tiling(kgem->fd, |
bo->handle, |
tiling, pitch)) |
continue; |
bo->pitch = pitch; |
} |
} |
kgem_bo_remove_from_active(kgem, bo); |
bo->unique_id = kgem_get_unique_id(kgem); |
bo->delta = 0; |
DBG((" 1:from active: pitch=%d, tiling=%d, handle=%d, id=%d\n", |
bo->pitch, bo->tiling, bo->handle, bo->unique_id)); |
assert(bo->pitch*kgem_aligned_height(kgem, height, bo->tiling) <= kgem_bo_size(bo)); |
assert_tiling(kgem, bo); |
bo->refcnt = 1; |
return bo; |
} |
} else { |
list_for_each_entry(bo, cache, list) { |
assert(bucket(bo) == bucket); |
assert(!bo->purged); |
assert(bo->refcnt == 0); |
assert(bo->reusable); |
assert(!bo->scanout); |
assert(bo->tiling == tiling); |
assert(bo->flush == false); |
assert_tiling(kgem, bo); |
if (num_pages(bo) < size) |
continue; |
kgem_bo_remove_from_active(kgem, bo); |
bo->pitch = pitch; |
bo->unique_id = kgem_get_unique_id(kgem); |
bo->delta = 0; |
DBG((" 1:from active: pitch=%d, tiling=%d, handle=%d, id=%d\n", |
bo->pitch, bo->tiling, bo->handle, bo->unique_id)); |
assert(bo->pitch*kgem_aligned_height(kgem, height, bo->tiling) <= kgem_bo_size(bo)); |
assert_tiling(kgem, bo); |
bo->refcnt = 1; |
return bo; |
} |
} |
if (--retry && exact) { |
if (kgem->gen >= 040) { |
for (i = I915_TILING_NONE; i <= I915_TILING_Y; i++) { |
if (i == tiling) |
continue; |
cache = &kgem->active[bucket][i]; |
list_for_each_entry(bo, cache, list) { |
assert(!bo->purged); |
assert(bo->refcnt == 0); |
assert(bo->reusable); |
assert(!bo->scanout); |
assert(bo->flush == false); |
assert_tiling(kgem, bo); |
if (num_pages(bo) < size) |
continue; |
if (!gem_set_tiling(kgem->fd, |
bo->handle, |
tiling, pitch)) |
continue; |
kgem_bo_remove_from_active(kgem, bo); |
bo->unique_id = kgem_get_unique_id(kgem); |
bo->pitch = pitch; |
bo->tiling = tiling; |
bo->delta = 0; |
DBG((" 1:from active: pitch=%d, tiling=%d, handle=%d, id=%d\n", |
bo->pitch, bo->tiling, bo->handle, bo->unique_id)); |
assert(bo->pitch*kgem_aligned_height(kgem, height, bo->tiling) <= kgem_bo_size(bo)); |
assert_tiling(kgem, bo); |
bo->refcnt = 1; |
return bo; |
} |
} |
} |
bucket++; |
goto search_again; |
} |
if (!exact) { /* allow an active near-miss? */ |
i = tiling; |
while (--i >= 0) { |
tiled_height = kgem_surface_size(kgem, kgem->has_relaxed_fencing, flags, |
width, height, bpp, tiling, &pitch); |
cache = active(kgem, tiled_height / PAGE_SIZE, i); |
tiled_height = kgem_aligned_height(kgem, height, i); |
list_for_each_entry(bo, cache, list) { |
assert(!bo->purged); |
assert(bo->refcnt == 0); |
assert(bo->reusable); |
assert(!bo->scanout); |
assert(bo->flush == false); |
assert_tiling(kgem, bo); |
if (bo->tiling) { |
if (bo->pitch < pitch) { |
DBG(("tiled and pitch too small: tiling=%d, (want %d), pitch=%d, need %d\n", |
bo->tiling, tiling, |
bo->pitch, pitch)); |
continue; |
} |
} else |
bo->pitch = pitch; |
if (bo->pitch * tiled_height > bytes(bo)) |
continue; |
kgem_bo_remove_from_active(kgem, bo); |
bo->unique_id = kgem_get_unique_id(kgem); |
bo->delta = 0; |
DBG((" 1:from active: pitch=%d, tiling=%d, handle=%d, id=%d\n", |
bo->pitch, bo->tiling, bo->handle, bo->unique_id)); |
assert(bo->pitch*kgem_aligned_height(kgem, height, bo->tiling) <= kgem_bo_size(bo)); |
assert_tiling(kgem, bo); |
bo->refcnt = 1; |
return bo; |
} |
} |
} |
skip_active_search: |
bucket = cache_bucket(size); |
retry = NUM_CACHE_BUCKETS - bucket; |
if (retry > 3) |
retry = 3; |
search_inactive: |
/* Now just look for a close match and prefer any currently active */ |
assert(bucket < NUM_CACHE_BUCKETS); |
cache = &kgem->inactive[bucket]; |
list_for_each_entry(bo, cache, list) { |
assert(bucket(bo) == bucket); |
assert(bo->reusable); |
assert(!bo->scanout); |
assert(bo->flush == false); |
assert_tiling(kgem, bo); |
if (size > num_pages(bo)) { |
DBG(("inactive too small: %d < %d\n", |
num_pages(bo), size)); |
continue; |
} |
if (bo->tiling != tiling || |
(tiling != I915_TILING_NONE && bo->pitch != pitch)) { |
if (!gem_set_tiling(kgem->fd, bo->handle, |
tiling, pitch)) |
continue; |
if (bo->map) |
kgem_bo_release_map(kgem, bo); |
} |
if (bo->purged && !kgem_bo_clear_purgeable(kgem, bo)) { |
kgem_bo_free(kgem, bo); |
break; |
} |
kgem_bo_remove_from_inactive(kgem, bo); |
bo->pitch = pitch; |
bo->tiling = tiling; |
bo->delta = 0; |
bo->unique_id = kgem_get_unique_id(kgem); |
assert(bo->pitch); |
DBG((" from inactive: pitch=%d, tiling=%d: handle=%d, id=%d\n", |
bo->pitch, bo->tiling, bo->handle, bo->unique_id)); |
assert(bo->refcnt == 0); |
assert(bo->reusable); |
assert((flags & CREATE_INACTIVE) == 0 || bo->domain != DOMAIN_GPU); |
ASSERT_MAYBE_IDLE(kgem, bo->handle, flags & CREATE_INACTIVE); |
assert(bo->pitch*kgem_aligned_height(kgem, height, bo->tiling) <= kgem_bo_size(bo)); |
assert_tiling(kgem, bo); |
bo->refcnt = 1; |
return bo; |
} |
if (flags & CREATE_INACTIVE && |
!list_is_empty(&kgem->active[bucket][tiling]) && |
__kgem_throttle_retire(kgem, flags)) { |
flags &= ~CREATE_INACTIVE; |
goto search_inactive; |
} |
if (--retry) { |
bucket++; |
flags &= ~CREATE_INACTIVE; |
goto search_inactive; |
} |
create: |
if (flags & CREATE_CACHED) |
return NULL; |
if (bucket >= NUM_CACHE_BUCKETS) |
size = ALIGN(size, 1024); |
handle = gem_create(kgem->fd, size); |
if (handle == 0) |
return NULL; |
bo = __kgem_bo_alloc(handle, size); |
if (!bo) { |
gem_close(kgem->fd, handle); |
return NULL; |
} |
if (bucket >= NUM_CACHE_BUCKETS) { |
DBG(("%s: marking large bo for automatic flushing\n", |
__FUNCTION__)); |
bo->flush = true; |
} |
bo->unique_id = kgem_get_unique_id(kgem); |
if (tiling == I915_TILING_NONE || |
gem_set_tiling(kgem->fd, handle, tiling, pitch)) { |
bo->tiling = tiling; |
bo->pitch = pitch; |
} else { |
if (flags & CREATE_EXACT) { |
if (bo->pitch != pitch || bo->tiling != tiling) { |
kgem_bo_free(kgem, bo); |
return NULL; |
} |
} |
} |
assert(bytes(bo) >= bo->pitch * kgem_aligned_height(kgem, height, bo->tiling)); |
assert_tiling(kgem, bo); |
debug_alloc__bo(kgem, bo); |
DBG((" new pitch=%d, tiling=%d, handle=%d, id=%d, num_pages=%d [%d], bucket=%d\n", |
bo->pitch, bo->tiling, bo->handle, bo->unique_id, |
size, num_pages(bo), bucket(bo))); |
return bo; |
} |
#if 0 |
struct kgem_bo *kgem_create_cpu_2d(struct kgem *kgem, |
int width, |
int height, |
int bpp, |
uint32_t flags) |
{ |
struct kgem_bo *bo; |
int stride, size; |
if (DBG_NO_CPU) |
return NULL; |
DBG(("%s(%dx%d, bpp=%d)\n", __FUNCTION__, width, height, bpp)); |
if (kgem->has_llc) { |
bo = kgem_create_2d(kgem, width, height, bpp, |
I915_TILING_NONE, flags); |
if (bo == NULL) |
return bo; |
assert(bo->tiling == I915_TILING_NONE); |
assert_tiling(kgem, bo); |
if (kgem_bo_map__cpu(kgem, bo) == NULL) { |
kgem_bo_destroy(kgem, bo); |
return NULL; |
} |
return bo; |
} |
assert(width > 0 && height > 0); |
stride = ALIGN(width, 2) * bpp >> 3; |
stride = ALIGN(stride, 4); |
size = stride * ALIGN(height, 2); |
assert(size >= PAGE_SIZE); |
DBG(("%s: %dx%d, %d bpp, stride=%d\n", |
__FUNCTION__, width, height, bpp, stride)); |
bo = search_snoop_cache(kgem, NUM_PAGES(size), 0); |
if (bo) { |
assert(bo->tiling == I915_TILING_NONE); |
assert_tiling(kgem, bo); |
assert(bo->snoop); |
bo->refcnt = 1; |
bo->pitch = stride; |
bo->unique_id = kgem_get_unique_id(kgem); |
return bo; |
} |
if (kgem->has_caching) { |
bo = kgem_create_linear(kgem, size, flags); |
if (bo == NULL) |
return NULL; |
assert(bo->tiling == I915_TILING_NONE); |
assert_tiling(kgem, bo); |
if (!gem_set_caching(kgem->fd, bo->handle, SNOOPED)) { |
kgem_bo_destroy(kgem, bo); |
return NULL; |
} |
bo->snoop = true; |
if (kgem_bo_map__cpu(kgem, bo) == NULL) { |
kgem_bo_destroy(kgem, bo); |
return NULL; |
} |
bo->pitch = stride; |
bo->unique_id = kgem_get_unique_id(kgem); |
return bo; |
} |
if (kgem->has_userptr) { |
void *ptr; |
/* XXX */ |
//if (posix_memalign(&ptr, 64, ALIGN(size, 64))) |
if (posix_memalign(&ptr, PAGE_SIZE, ALIGN(size, PAGE_SIZE))) |
return NULL; |
bo = kgem_create_map(kgem, ptr, size, false); |
if (bo == NULL) { |
free(ptr); |
return NULL; |
} |
bo->pitch = stride; |
bo->unique_id = kgem_get_unique_id(kgem); |
return bo; |
} |
return NULL; |
} |
#endif |
void _kgem_bo_destroy(struct kgem *kgem, struct kgem_bo *bo) |
{ |
DBG(("%s: handle=%d, proxy? %d\n", |
__FUNCTION__, bo->handle, bo->proxy != NULL)); |
if (bo->proxy) { |
_list_del(&bo->vma); |
_list_del(&bo->request); |
if (bo->io && bo->exec == NULL) |
_kgem_bo_delete_buffer(kgem, bo); |
kgem_bo_unref(kgem, bo->proxy); |
kgem_bo_binding_free(kgem, bo); |
free(bo); |
return; |
} |
__kgem_bo_destroy(kgem, bo); |
} |
static void __kgem_flush(struct kgem *kgem, struct kgem_bo *bo) |
{ |
assert(bo->rq); |
assert(bo->exec == NULL); |
assert(bo->needs_flush); |
/* The kernel will emit a flush *and* update its own flushing lists. */ |
if (!__kgem_busy(kgem, bo->handle)) |
__kgem_bo_clear_busy(bo); |
DBG(("%s: handle=%d, busy?=%d\n", |
__FUNCTION__, bo->handle, bo->rq != NULL)); |
} |
void kgem_scanout_flush(struct kgem *kgem, struct kgem_bo *bo) |
{ |
kgem_bo_submit(kgem, bo); |
if (!bo->needs_flush) |
return; |
/* If the kernel fails to emit the flush, then it will be forced when |
* we assume direct access. And as the usual failure is EIO, we do |
* not actually care. |
*/ |
assert(bo->exec == NULL); |
if (bo->rq) |
__kgem_flush(kgem, bo); |
/* Whatever actually happens, we can regard the GTT write domain |
* as being flushed. |
*/ |
bo->gtt_dirty = false; |
bo->needs_flush = false; |
bo->domain = DOMAIN_NONE; |
} |
inline static bool needs_semaphore(struct kgem *kgem, struct kgem_bo *bo) |
{ |
return kgem->nreloc && bo->rq && RQ_RING(bo->rq) != kgem->ring; |
} |
bool kgem_check_bo(struct kgem *kgem, ...) |
{ |
va_list ap; |
struct kgem_bo *bo; |
int num_exec = 0; |
int num_pages = 0; |
bool flush = false; |
va_start(ap, kgem); |
while ((bo = va_arg(ap, struct kgem_bo *))) { |
while (bo->proxy) |
bo = bo->proxy; |
if (bo->exec) |
continue; |
if (needs_semaphore(kgem, bo)) |
return false; |
num_pages += num_pages(bo); |
num_exec++; |
flush |= bo->flush; |
} |
va_end(ap); |
DBG(("%s: num_pages=+%d, num_exec=+%d\n", |
__FUNCTION__, num_pages, num_exec)); |
if (!num_pages) |
return true; |
if (kgem_flush(kgem, flush)) |
return false; |
if (kgem->aperture > kgem->aperture_low && |
kgem_ring_is_idle(kgem, kgem->ring)) { |
DBG(("%s: current aperture usage (%d) is greater than low water mark (%d)\n", |
__FUNCTION__, kgem->aperture, kgem->aperture_low)); |
return false; |
} |
if (num_pages + kgem->aperture > kgem->aperture_high) { |
DBG(("%s: final aperture usage (%d) is greater than high water mark (%d)\n", |
__FUNCTION__, num_pages + kgem->aperture, kgem->aperture_high)); |
return false; |
} |
if (kgem->nexec + num_exec >= KGEM_EXEC_SIZE(kgem)) { |
DBG(("%s: out of exec slots (%d + %d / %d)\n", __FUNCTION__, |
kgem->nexec, num_exec, KGEM_EXEC_SIZE(kgem))); |
return false; |
} |
return true; |
} |
uint32_t kgem_add_reloc(struct kgem *kgem, |
uint32_t pos, |
struct kgem_bo *bo, |
uint32_t read_write_domain, |
uint32_t delta) |
{ |
int index; |
DBG(("%s: handle=%d, pos=%d, delta=%d, domains=%08x\n", |
__FUNCTION__, bo ? bo->handle : 0, pos, delta, read_write_domain)); |
assert((read_write_domain & 0x7fff) == 0 || bo != NULL); |
if( bo != NULL && bo->handle == -2) |
{ |
if (bo->exec == NULL) |
kgem_add_bo(kgem, bo); |
if (read_write_domain & 0x7fff && !bo->gpu_dirty) { |
__kgem_bo_mark_dirty(bo); |
} |
return 0; |
}; |
index = kgem->nreloc++; |
assert(index < ARRAY_SIZE(kgem->reloc)); |
kgem->reloc[index].offset = pos * sizeof(kgem->batch[0]); |
if (bo) { |
assert(bo->refcnt); |
while (bo->proxy) { |
DBG(("%s: adding proxy [delta=%d] for handle=%d\n", |
__FUNCTION__, bo->delta, bo->handle)); |
delta += bo->delta; |
assert(bo->handle == bo->proxy->handle); |
/* need to release the cache upon batch submit */ |
if (bo->exec == NULL) { |
list_move_tail(&bo->request, |
&kgem->next_request->buffers); |
bo->rq = MAKE_REQUEST(kgem->next_request, |
kgem->ring); |
bo->exec = &_kgem_dummy_exec; |
} |
if (read_write_domain & 0x7fff && !bo->gpu_dirty) |
__kgem_bo_mark_dirty(bo); |
bo = bo->proxy; |
assert(bo->refcnt); |
} |
assert(bo->refcnt); |
if (bo->exec == NULL) |
kgem_add_bo(kgem, bo); |
assert(bo->rq == MAKE_REQUEST(kgem->next_request, kgem->ring)); |
assert(RQ_RING(bo->rq) == kgem->ring); |
if (kgem->gen < 040 && read_write_domain & KGEM_RELOC_FENCED) { |
if (bo->tiling && |
(bo->exec->flags & EXEC_OBJECT_NEEDS_FENCE) == 0) { |
assert(kgem->nfence < kgem->fence_max); |
kgem->aperture_fenced += |
kgem_bo_fenced_size(kgem, bo); |
kgem->nfence++; |
} |
bo->exec->flags |= EXEC_OBJECT_NEEDS_FENCE; |
} |
kgem->reloc[index].delta = delta; |
kgem->reloc[index].target_handle = bo->target_handle; |
kgem->reloc[index].presumed_offset = bo->presumed_offset; |
if (read_write_domain & 0x7fff && !bo->gpu_dirty) { |
assert(!bo->snoop || kgem->can_blt_cpu); |
__kgem_bo_mark_dirty(bo); |
} |
delta += bo->presumed_offset; |
} else { |
kgem->reloc[index].delta = delta; |
kgem->reloc[index].target_handle = ~0U; |
kgem->reloc[index].presumed_offset = 0; |
if (kgem->nreloc__self < 256) |
kgem->reloc__self[kgem->nreloc__self++] = index; |
} |
kgem->reloc[index].read_domains = read_write_domain >> 16; |
kgem->reloc[index].write_domain = read_write_domain & 0x7fff; |
return delta; |
} |
static void kgem_trim_vma_cache(struct kgem *kgem, int type, int bucket) |
{ |
int i, j; |
DBG(("%s: type=%d, count=%d (bucket: %d)\n", |
__FUNCTION__, type, kgem->vma[type].count, bucket)); |
if (kgem->vma[type].count <= 0) |
return; |
if (kgem->need_purge) |
kgem_purge_cache(kgem); |
/* vma are limited on a per-process basis to around 64k. |
* This includes all malloc arenas as well as other file |
* mappings. In order to be fair and not hog the cache, |
* and more importantly not to exhaust that limit and to |
* start failing mappings, we keep our own number of open |
* vma to within a conservative value. |
*/ |
i = 0; |
while (kgem->vma[type].count > 0) { |
struct kgem_bo *bo = NULL; |
for (j = 0; |
bo == NULL && j < ARRAY_SIZE(kgem->vma[type].inactive); |
j++) { |
struct list *head = &kgem->vma[type].inactive[i++%ARRAY_SIZE(kgem->vma[type].inactive)]; |
if (!list_is_empty(head)) |
bo = list_last_entry(head, struct kgem_bo, vma); |
} |
if (bo == NULL) |
break; |
DBG(("%s: discarding inactive %s vma cache for %d\n", |
__FUNCTION__, |
IS_CPU_MAP(bo->map) ? "CPU" : "GTT", bo->handle)); |
assert(IS_CPU_MAP(bo->map) == type); |
assert(bo->map); |
assert(bo->rq == NULL); |
VG(if (type) VALGRIND_MAKE_MEM_NOACCESS(MAP(bo->map), bytes(bo))); |
// munmap(MAP(bo->map), bytes(bo)); |
bo->map = NULL; |
list_del(&bo->vma); |
kgem->vma[type].count--; |
if (!bo->purged && !kgem_bo_set_purgeable(kgem, bo)) { |
DBG(("%s: freeing unpurgeable old mapping\n", |
__FUNCTION__)); |
kgem_bo_free(kgem, bo); |
} |
} |
} |
void *kgem_bo_map__async(struct kgem *kgem, struct kgem_bo *bo) |
{ |
void *ptr; |
DBG(("%s: handle=%d, offset=%d, tiling=%d, map=%p, domain=%d\n", __FUNCTION__, |
bo->handle, bo->presumed_offset, bo->tiling, bo->map, bo->domain)); |
assert(bo->proxy == NULL); |
assert(list_is_empty(&bo->list)); |
assert(!IS_USER_MAP(bo->map)); |
assert_tiling(kgem, bo); |
if (bo->tiling == I915_TILING_NONE && !bo->scanout && kgem->has_llc) { |
DBG(("%s: converting request for GTT map into CPU map\n", |
__FUNCTION__)); |
return kgem_bo_map__cpu(kgem, bo); |
} |
if (IS_CPU_MAP(bo->map)) |
kgem_bo_release_map(kgem, bo); |
ptr = bo->map; |
if (ptr == NULL) { |
assert(kgem_bo_size(bo) <= kgem->aperture_mappable / 2); |
kgem_trim_vma_cache(kgem, MAP_GTT, bucket(bo)); |
ptr = __kgem_bo_map__gtt(kgem, bo); |
if (ptr == NULL) |
return NULL; |
/* Cache this mapping to avoid the overhead of an |
* excruciatingly slow GTT pagefault. This is more an |
* issue with compositing managers which need to frequently |
* flush CPU damage to their GPU bo. |
*/ |
bo->map = ptr; |
DBG(("%s: caching GTT vma for %d\n", __FUNCTION__, bo->handle)); |
} |
return ptr; |
} |
void *kgem_bo_map(struct kgem *kgem, struct kgem_bo *bo) |
{ |
void *ptr; |
DBG(("%s: handle=%d, offset=%d, tiling=%d, map=%p, domain=%d\n", __FUNCTION__, |
bo->handle, bo->presumed_offset, bo->tiling, bo->map, bo->domain)); |
assert(bo->proxy == NULL); |
assert(list_is_empty(&bo->list)); |
assert(!IS_USER_MAP(bo->map)); |
assert(bo->exec == NULL); |
assert_tiling(kgem, bo); |
if (bo->tiling == I915_TILING_NONE && !bo->scanout && |
(kgem->has_llc || bo->domain == DOMAIN_CPU)) { |
DBG(("%s: converting request for GTT map into CPU map\n", |
__FUNCTION__)); |
ptr = kgem_bo_map__cpu(kgem, bo); |
if (ptr) |
kgem_bo_sync__cpu(kgem, bo); |
return ptr; |
} |
if (IS_CPU_MAP(bo->map)) |
kgem_bo_release_map(kgem, bo); |
ptr = bo->map; |
if (ptr == NULL) { |
assert(kgem_bo_size(bo) <= kgem->aperture_mappable / 2); |
assert(kgem->gen != 021 || bo->tiling != I915_TILING_Y); |
kgem_trim_vma_cache(kgem, MAP_GTT, bucket(bo)); |
ptr = __kgem_bo_map__gtt(kgem, bo); |
if (ptr == NULL) |
return NULL; |
/* Cache this mapping to avoid the overhead of an |
* excruciatingly slow GTT pagefault. This is more an |
* issue with compositing managers which need to frequently |
* flush CPU damage to their GPU bo. |
*/ |
bo->map = ptr; |
DBG(("%s: caching GTT vma for %d\n", __FUNCTION__, bo->handle)); |
} |
if (bo->domain != DOMAIN_GTT || FORCE_MMAP_SYNC & (1 << DOMAIN_GTT)) { |
struct drm_i915_gem_set_domain set_domain; |
DBG(("%s: sync: needs_flush? %d, domain? %d, busy? %d\n", __FUNCTION__, |
bo->needs_flush, bo->domain, __kgem_busy(kgem, bo->handle))); |
/* XXX use PROT_READ to avoid the write flush? */ |
VG_CLEAR(set_domain); |
set_domain.handle = bo->handle; |
set_domain.read_domains = I915_GEM_DOMAIN_GTT; |
set_domain.write_domain = I915_GEM_DOMAIN_GTT; |
if (drmIoctl(kgem->fd, DRM_IOCTL_I915_GEM_SET_DOMAIN, &set_domain) == 0) { |
kgem_bo_retire(kgem, bo); |
bo->domain = DOMAIN_GTT; |
bo->gtt_dirty = true; |
} |
} |
return ptr; |
} |
void *kgem_bo_map__gtt(struct kgem *kgem, struct kgem_bo *bo) |
{ |
void *ptr; |
DBG(("%s: handle=%d, offset=%d, tiling=%d, map=%p, domain=%d\n", __FUNCTION__, |
bo->handle, bo->presumed_offset, bo->tiling, bo->map, bo->domain)); |
assert(bo->exec == NULL); |
assert(list_is_empty(&bo->list)); |
assert(!IS_USER_MAP(bo->map)); |
assert_tiling(kgem, bo); |
if (IS_CPU_MAP(bo->map)) |
kgem_bo_release_map(kgem, bo); |
ptr = bo->map; |
if (ptr == NULL) { |
assert(bytes(bo) <= kgem->aperture_mappable / 4); |
kgem_trim_vma_cache(kgem, MAP_GTT, bucket(bo)); |
ptr = __kgem_bo_map__gtt(kgem, bo); |
if (ptr == NULL) |
return NULL; |
/* Cache this mapping to avoid the overhead of an |
* excruciatingly slow GTT pagefault. This is more an |
* issue with compositing managers which need to frequently |
* flush CPU damage to their GPU bo. |
*/ |
bo->map = ptr; |
DBG(("%s: caching GTT vma for %d\n", __FUNCTION__, bo->handle)); |
} |
return ptr; |
} |
void *kgem_bo_map__debug(struct kgem *kgem, struct kgem_bo *bo) |
{ |
if (bo->map) |
return MAP(bo->map); |
kgem_trim_vma_cache(kgem, MAP_GTT, bucket(bo)); |
return bo->map = __kgem_bo_map__gtt(kgem, bo); |
} |
void *kgem_bo_map__cpu(struct kgem *kgem, struct kgem_bo *bo) |
{ |
struct drm_i915_gem_mmap mmap_arg; |
DBG(("%s(handle=%d, size=%d, mapped? %d)\n", |
__FUNCTION__, bo->handle, bytes(bo), (int)__MAP_TYPE(bo->map))); |
assert(!bo->purged); |
assert(list_is_empty(&bo->list)); |
assert(bo->proxy == NULL); |
if (IS_CPU_MAP(bo->map)) |
return MAP(bo->map); |
if (bo->map) |
kgem_bo_release_map(kgem, bo); |
kgem_trim_vma_cache(kgem, MAP_CPU, bucket(bo)); |
retry: |
VG_CLEAR(mmap_arg); |
mmap_arg.handle = bo->handle; |
mmap_arg.offset = 0; |
mmap_arg.size = bytes(bo); |
if (drmIoctl(kgem->fd, DRM_IOCTL_I915_GEM_MMAP, &mmap_arg)) { |
if (__kgem_throttle_retire(kgem, 0)) |
goto retry; |
if (kgem->need_expire) { |
kgem_cleanup_cache(kgem); |
goto retry; |
} |
ErrorF("%s: failed to mmap handle=%d, %d bytes, into CPU domain\n", |
__FUNCTION__, bo->handle, bytes(bo)); |
return NULL; |
} |
VG(VALGRIND_MAKE_MEM_DEFINED(mmap_arg.addr_ptr, bytes(bo))); |
DBG(("%s: caching CPU vma for %d\n", __FUNCTION__, bo->handle)); |
bo->map = MAKE_CPU_MAP(mmap_arg.addr_ptr); |
return (void *)(uintptr_t)mmap_arg.addr_ptr; |
} |
void *__kgem_bo_map__cpu(struct kgem *kgem, struct kgem_bo *bo) |
{ |
struct drm_i915_gem_mmap mmap_arg; |
DBG(("%s(handle=%d, size=%d, mapped? %d)\n", |
__FUNCTION__, bo->handle, bytes(bo), (int)__MAP_TYPE(bo->map))); |
assert(bo->refcnt); |
assert(!bo->purged); |
assert(list_is_empty(&bo->list)); |
assert(bo->proxy == NULL); |
if (IS_CPU_MAP(bo->map)) |
return MAP(bo->map); |
retry: |
VG_CLEAR(mmap_arg); |
mmap_arg.handle = bo->handle; |
mmap_arg.offset = 0; |
mmap_arg.size = bytes(bo); |
if (drmIoctl(kgem->fd, DRM_IOCTL_I915_GEM_MMAP, &mmap_arg)) { |
int err = errno; |
assert(err != EINVAL); |
if (__kgem_throttle_retire(kgem, 0)) |
goto retry; |
if (kgem->need_expire) { |
kgem_cleanup_cache(kgem); |
goto retry; |
} |
ErrorF("%s: failed to mmap handle=%d, %d bytes, into CPU domain: %d\n", |
__FUNCTION__, bo->handle, bytes(bo), err); |
return NULL; |
} |
VG(VALGRIND_MAKE_MEM_DEFINED(mmap_arg.addr_ptr, bytes(bo))); |
if (bo->map && bo->domain == DOMAIN_CPU) { |
DBG(("%s: discarding GTT vma for %d\n", __FUNCTION__, bo->handle)); |
kgem_bo_release_map(kgem, bo); |
} |
if (bo->map == NULL) { |
DBG(("%s: caching CPU vma for %d\n", __FUNCTION__, bo->handle)); |
bo->map = MAKE_CPU_MAP(mmap_arg.addr_ptr); |
} |
return (void *)(uintptr_t)mmap_arg.addr_ptr; |
} |
void kgem_bo_sync__cpu(struct kgem *kgem, struct kgem_bo *bo) |
{ |
DBG(("%s: handle=%d\n", __FUNCTION__, bo->handle)); |
assert(!bo->scanout); |
kgem_bo_submit(kgem, bo); |
/* SHM pixmaps use proxies for subpage offsets */ |
assert(!bo->purged); |
while (bo->proxy) |
bo = bo->proxy; |
assert(!bo->purged); |
if (bo->domain != DOMAIN_CPU || FORCE_MMAP_SYNC & (1 << DOMAIN_CPU)) { |
struct drm_i915_gem_set_domain set_domain; |
DBG(("%s: SYNC: handle=%d, needs_flush? %d, domain? %d, busy? %d\n", |
__FUNCTION__, bo->handle, |
bo->needs_flush, bo->domain, |
__kgem_busy(kgem, bo->handle))); |
VG_CLEAR(set_domain); |
set_domain.handle = bo->handle; |
set_domain.read_domains = I915_GEM_DOMAIN_CPU; |
set_domain.write_domain = I915_GEM_DOMAIN_CPU; |
if (drmIoctl(kgem->fd, DRM_IOCTL_I915_GEM_SET_DOMAIN, &set_domain) == 0) { |
kgem_bo_retire(kgem, bo); |
bo->domain = DOMAIN_CPU; |
} |
} |
} |
void kgem_clear_dirty(struct kgem *kgem) |
{ |
struct list * const buffers = &kgem->next_request->buffers; |
struct kgem_bo *bo; |
list_for_each_entry(bo, buffers, request) { |
if (!bo->gpu_dirty) |
break; |
bo->gpu_dirty = false; |
} |
} |
struct kgem_bo *kgem_create_proxy(struct kgem *kgem, |
struct kgem_bo *target, |
int offset, int length) |
{ |
struct kgem_bo *bo; |
DBG(("%s: target handle=%d [proxy? %d], offset=%d, length=%d, io=%d\n", |
__FUNCTION__, target->handle, target->proxy ? target->proxy->delta : -1, |
offset, length, target->io)); |
bo = __kgem_bo_alloc(target->handle, length); |
if (bo == NULL) |
return NULL; |
bo->unique_id = kgem_get_unique_id(kgem); |
bo->reusable = false; |
bo->size.bytes = length; |
bo->io = target->io && target->proxy == NULL; |
bo->gpu_dirty = target->gpu_dirty; |
bo->tiling = target->tiling; |
bo->pitch = target->pitch; |
bo->flush = target->flush; |
bo->snoop = target->snoop; |
assert(!bo->scanout); |
bo->proxy = kgem_bo_reference(target); |
bo->delta = offset; |
if (target->exec) { |
list_move_tail(&bo->request, &kgem->next_request->buffers); |
bo->exec = &_kgem_dummy_exec; |
} |
bo->rq = target->rq; |
return bo; |
} |
#if 0 |
static struct kgem_buffer * |
buffer_alloc(void) |
{ |
struct kgem_buffer *bo; |
bo = malloc(sizeof(*bo)); |
if (bo == NULL) |
return NULL; |
bo->mem = NULL; |
bo->need_io = false; |
bo->mmapped = true; |
return bo; |
} |
static struct kgem_buffer * |
buffer_alloc_with_data(int num_pages) |
{ |
struct kgem_buffer *bo; |
bo = malloc(sizeof(*bo) + 2*UPLOAD_ALIGNMENT + num_pages * PAGE_SIZE); |
if (bo == NULL) |
return NULL; |
bo->mem = (void *)ALIGN((uintptr_t)bo + sizeof(*bo), UPLOAD_ALIGNMENT); |
bo->mmapped = false; |
return bo; |
} |
static inline bool |
use_snoopable_buffer(struct kgem *kgem, uint32_t flags) |
{ |
if ((flags & KGEM_BUFFER_WRITE) == 0) |
return kgem->gen >= 030; |
return true; |
} |
static void |
init_buffer_from_bo(struct kgem_buffer *bo, struct kgem_bo *old) |
{ |
DBG(("%s: reusing handle=%d for buffer\n", |
__FUNCTION__, old->handle)); |
assert(old->proxy == NULL); |
memcpy(&bo->base, old, sizeof(*old)); |
if (old->rq) |
list_replace(&old->request, &bo->base.request); |
else |
list_init(&bo->base.request); |
list_replace(&old->vma, &bo->base.vma); |
list_init(&bo->base.list); |
free(old); |
assert(bo->base.tiling == I915_TILING_NONE); |
bo->base.refcnt = 1; |
} |
static struct kgem_buffer * |
search_snoopable_buffer(struct kgem *kgem, unsigned alloc) |
{ |
struct kgem_buffer *bo; |
struct kgem_bo *old; |
old = search_snoop_cache(kgem, alloc, 0); |
if (old) { |
if (!old->io) { |
bo = buffer_alloc(); |
if (bo == NULL) |
return NULL; |
init_buffer_from_bo(bo, old); |
} else { |
bo = (struct kgem_buffer *)old; |
bo->base.refcnt = 1; |
} |
DBG(("%s: created CPU handle=%d for buffer, size %d\n", |
__FUNCTION__, bo->base.handle, num_pages(&bo->base))); |
assert(bo->base.snoop); |
assert(bo->base.tiling == I915_TILING_NONE); |
assert(num_pages(&bo->base) >= alloc); |
assert(bo->mmapped == true); |
assert(bo->need_io == false); |
bo->mem = kgem_bo_map__cpu(kgem, &bo->base); |
if (bo->mem == NULL) { |
bo->base.refcnt = 0; |
kgem_bo_free(kgem, &bo->base); |
bo = NULL; |
} |
return bo; |
} |
return NULL; |
} |
static struct kgem_buffer * |
create_snoopable_buffer(struct kgem *kgem, unsigned alloc) |
{ |
struct kgem_buffer *bo; |
uint32_t handle; |
if (kgem->has_llc) { |
struct kgem_bo *old; |
bo = buffer_alloc(); |
if (bo == NULL) |
return NULL; |
old = search_linear_cache(kgem, alloc, |
CREATE_INACTIVE | CREATE_CPU_MAP | CREATE_EXACT); |
if (old) { |
init_buffer_from_bo(bo, old); |
} else { |
handle = gem_create(kgem->fd, alloc); |
if (handle == 0) { |
free(bo); |
return NULL; |
} |
debug_alloc(kgem, alloc); |
__kgem_bo_init(&bo->base, handle, alloc); |
DBG(("%s: created CPU (LLC) handle=%d for buffer, size %d\n", |
__FUNCTION__, bo->base.handle, alloc)); |
} |
assert(bo->base.refcnt == 1); |
assert(bo->mmapped == true); |
assert(bo->need_io == false); |
bo->mem = kgem_bo_map__cpu(kgem, &bo->base); |
if (bo->mem != NULL) |
return bo; |
bo->base.refcnt = 0; /* for valgrind */ |
kgem_bo_free(kgem, &bo->base); |
} |
if (kgem->has_caching) { |
struct kgem_bo *old; |
bo = buffer_alloc(); |
if (bo == NULL) |
return NULL; |
old = search_linear_cache(kgem, alloc, |
CREATE_INACTIVE | CREATE_CPU_MAP | CREATE_EXACT); |
if (old) { |
init_buffer_from_bo(bo, old); |
} else { |
handle = gem_create(kgem->fd, alloc); |
if (handle == 0) { |
free(bo); |
return NULL; |
} |
debug_alloc(kgem, alloc); |
__kgem_bo_init(&bo->base, handle, alloc); |
DBG(("%s: created CPU handle=%d for buffer, size %d\n", |
__FUNCTION__, bo->base.handle, alloc)); |
} |
assert(bo->base.refcnt == 1); |
assert(bo->mmapped == true); |
assert(bo->need_io == false); |
if (!gem_set_caching(kgem->fd, bo->base.handle, SNOOPED)) |
goto free_caching; |
bo->base.snoop = true; |
bo->mem = kgem_bo_map__cpu(kgem, &bo->base); |
if (bo->mem == NULL) |
goto free_caching; |
return bo; |
free_caching: |
bo->base.refcnt = 0; /* for valgrind */ |
kgem_bo_free(kgem, &bo->base); |
} |
if (kgem->has_userptr) { |
bo = buffer_alloc(); |
if (bo == NULL) |
return NULL; |
//if (posix_memalign(&ptr, 64, ALIGN(size, 64))) |
if (posix_memalign(&bo->mem, PAGE_SIZE, alloc * PAGE_SIZE)) { |
free(bo); |
return NULL; |
} |
handle = gem_userptr(kgem->fd, bo->mem, alloc * PAGE_SIZE, false); |
if (handle == 0) { |
free(bo->mem); |
free(bo); |
return NULL; |
} |
debug_alloc(kgem, alloc); |
__kgem_bo_init(&bo->base, handle, alloc); |
DBG(("%s: created snoop handle=%d for buffer\n", |
__FUNCTION__, bo->base.handle)); |
assert(bo->mmapped == true); |
assert(bo->need_io == false); |
bo->base.refcnt = 1; |
bo->base.snoop = true; |
bo->base.map = MAKE_USER_MAP(bo->mem); |
return bo; |
} |
return NULL; |
} |
struct kgem_bo *kgem_create_buffer(struct kgem *kgem, |
uint32_t size, uint32_t flags, |
void **ret) |
{ |
struct kgem_buffer *bo; |
unsigned offset, alloc; |
struct kgem_bo *old; |
DBG(("%s: size=%d, flags=%x [write?=%d, inplace?=%d, last?=%d]\n", |
__FUNCTION__, size, flags, |
!!(flags & KGEM_BUFFER_WRITE), |
!!(flags & KGEM_BUFFER_INPLACE), |
!!(flags & KGEM_BUFFER_LAST))); |
assert(size); |
/* we should never be asked to create anything TOO large */ |
assert(size <= kgem->max_object_size); |
#if !DBG_NO_UPLOAD_CACHE |
list_for_each_entry(bo, &kgem->batch_buffers, base.list) { |
assert(bo->base.io); |
assert(bo->base.refcnt >= 1); |
/* We can reuse any write buffer which we can fit */ |
if (flags == KGEM_BUFFER_LAST && |
bo->write == KGEM_BUFFER_WRITE && |
bo->base.refcnt == 1 && !bo->mmapped && |
size <= bytes(&bo->base)) { |
DBG(("%s: reusing write buffer for read of %d bytes? used=%d, total=%d\n", |
__FUNCTION__, size, bo->used, bytes(&bo->base))); |
gem_write(kgem->fd, bo->base.handle, |
0, bo->used, bo->mem); |
kgem_buffer_release(kgem, bo); |
bo->need_io = 0; |
bo->write = 0; |
offset = 0; |
bo->used = size; |
goto done; |
} |
if (flags & KGEM_BUFFER_WRITE) { |
if ((bo->write & KGEM_BUFFER_WRITE) == 0 || |
(((bo->write & ~flags) & KGEM_BUFFER_INPLACE) && |
!bo->base.snoop)) { |
DBG(("%s: skip write %x buffer, need %x\n", |
__FUNCTION__, bo->write, flags)); |
continue; |
} |
assert(bo->mmapped || bo->need_io); |
} else { |
if (bo->write & KGEM_BUFFER_WRITE) { |
DBG(("%s: skip write %x buffer, need %x\n", |
__FUNCTION__, bo->write, flags)); |
continue; |
} |
} |
if (bo->used + size <= bytes(&bo->base)) { |
DBG(("%s: reusing buffer? used=%d + size=%d, total=%d\n", |
__FUNCTION__, bo->used, size, bytes(&bo->base))); |
offset = bo->used; |
bo->used += size; |
goto done; |
} |
} |
if (flags & KGEM_BUFFER_WRITE) { |
list_for_each_entry(bo, &kgem->active_buffers, base.list) { |
assert(bo->base.io); |
assert(bo->base.refcnt >= 1); |
assert(bo->mmapped); |
assert(!IS_CPU_MAP(bo->base.map) || kgem->has_llc || bo->base.snoop); |
if (!kgem->has_llc && (bo->write & ~flags) & KGEM_BUFFER_INPLACE) { |
DBG(("%s: skip write %x buffer, need %x\n", |
__FUNCTION__, bo->write, flags)); |
continue; |
} |
if (bo->used + size <= bytes(&bo->base)) { |
DBG(("%s: reusing buffer? used=%d + size=%d, total=%d\n", |
__FUNCTION__, bo->used, size, bytes(&bo->base))); |
offset = bo->used; |
bo->used += size; |
list_move(&bo->base.list, &kgem->batch_buffers); |
goto done; |
} |
} |
} |
#endif |
#if !DBG_NO_MAP_UPLOAD |
/* Be a little more generous and hope to hold fewer mmappings */ |
alloc = ALIGN(2*size, kgem->buffer_size); |
if (alloc > MAX_CACHE_SIZE) |
alloc = ALIGN(size, kgem->buffer_size); |
if (alloc > MAX_CACHE_SIZE) |
alloc = PAGE_ALIGN(size); |
assert(alloc); |
if (alloc > kgem->aperture_mappable / 4) |
flags &= ~KGEM_BUFFER_INPLACE; |
alloc /= PAGE_SIZE; |
if (kgem->has_llc && |
(flags & KGEM_BUFFER_WRITE_INPLACE) != KGEM_BUFFER_WRITE_INPLACE) { |
bo = buffer_alloc(); |
if (bo == NULL) |
goto skip_llc; |
old = NULL; |
if ((flags & KGEM_BUFFER_WRITE) == 0) |
old = search_linear_cache(kgem, alloc, CREATE_CPU_MAP); |
if (old == NULL) |
old = search_linear_cache(kgem, alloc, CREATE_INACTIVE | CREATE_CPU_MAP); |
if (old == NULL) |
old = search_linear_cache(kgem, NUM_PAGES(size), CREATE_INACTIVE | CREATE_CPU_MAP); |
if (old) { |
DBG(("%s: found LLC handle=%d for buffer\n", |
__FUNCTION__, old->handle)); |
init_buffer_from_bo(bo, old); |
} else { |
uint32_t handle = gem_create(kgem->fd, alloc); |
if (handle == 0) { |
free(bo); |
goto skip_llc; |
} |
__kgem_bo_init(&bo->base, handle, alloc); |
DBG(("%s: created LLC handle=%d for buffer\n", |
__FUNCTION__, bo->base.handle)); |
debug_alloc(kgem, alloc); |
} |
assert(bo->mmapped); |
assert(!bo->need_io); |
bo->mem = kgem_bo_map__cpu(kgem, &bo->base); |
if (bo->mem) { |
if (flags & KGEM_BUFFER_WRITE) |
kgem_bo_sync__cpu(kgem, &bo->base); |
flags &= ~KGEM_BUFFER_INPLACE; |
goto init; |
} else { |
bo->base.refcnt = 0; /* for valgrind */ |
kgem_bo_free(kgem, &bo->base); |
} |
} |
skip_llc: |
if ((flags & KGEM_BUFFER_WRITE_INPLACE) == KGEM_BUFFER_WRITE_INPLACE) { |
/* The issue with using a GTT upload buffer is that we may |
* cause eviction-stalls in order to free up some GTT space. |
* An is-mappable? ioctl could help us detect when we are |
* about to block, or some per-page magic in the kernel. |
* |
* XXX This is especially noticeable on memory constrained |
* devices like gen2 or with relatively slow gpu like i3. |
*/ |
DBG(("%s: searching for an inactive GTT map for upload\n", |
__FUNCTION__)); |
old = search_linear_cache(kgem, alloc, |
CREATE_EXACT | CREATE_INACTIVE | CREATE_GTT_MAP); |
#if HAVE_I915_GEM_BUFFER_INFO |
if (old) { |
struct drm_i915_gem_buffer_info info; |
/* An example of such a non-blocking ioctl might work */ |
VG_CLEAR(info); |
info.handle = handle; |
if (drmIoctl(kgem->fd, |
DRM_IOCTL_I915_GEM_BUFFER_INFO, |
&fino) == 0) { |
old->presumed_offset = info.addr; |
if ((info.flags & I915_GEM_MAPPABLE) == 0) { |
kgem_bo_move_to_inactive(kgem, old); |
old = NULL; |
} |
} |
} |
#endif |
if (old == NULL) |
old = search_linear_cache(kgem, NUM_PAGES(size), |
CREATE_EXACT | CREATE_INACTIVE | CREATE_GTT_MAP); |
if (old == NULL) { |
old = search_linear_cache(kgem, alloc, CREATE_INACTIVE); |
if (old && !__kgem_bo_is_mappable(kgem, old)) { |
_kgem_bo_destroy(kgem, old); |
old = NULL; |
} |
} |
if (old) { |
DBG(("%s: reusing handle=%d for buffer\n", |
__FUNCTION__, old->handle)); |
assert(__kgem_bo_is_mappable(kgem, old)); |
assert(!old->snoop); |
assert(old->rq == NULL); |
bo = buffer_alloc(); |
if (bo == NULL) |
return NULL; |
init_buffer_from_bo(bo, old); |
assert(num_pages(&bo->base) >= NUM_PAGES(size)); |
assert(bo->mmapped); |
assert(bo->base.refcnt == 1); |
bo->mem = kgem_bo_map(kgem, &bo->base); |
if (bo->mem) { |
if (IS_CPU_MAP(bo->base.map)) |
flags &= ~KGEM_BUFFER_INPLACE; |
goto init; |
} else { |
bo->base.refcnt = 0; |
kgem_bo_free(kgem, &bo->base); |
} |
} |
} |
#else |
flags &= ~KGEM_BUFFER_INPLACE; |
#endif |
/* Be more parsimonious with pwrite/pread/cacheable buffers */ |
if ((flags & KGEM_BUFFER_INPLACE) == 0) |
alloc = NUM_PAGES(size); |
if (use_snoopable_buffer(kgem, flags)) { |
bo = search_snoopable_buffer(kgem, alloc); |
if (bo) { |
if (flags & KGEM_BUFFER_WRITE) |
kgem_bo_sync__cpu(kgem, &bo->base); |
flags &= ~KGEM_BUFFER_INPLACE; |
goto init; |
} |
if ((flags & KGEM_BUFFER_INPLACE) == 0) { |
bo = create_snoopable_buffer(kgem, alloc); |
if (bo) |
goto init; |
} |
} |
flags &= ~KGEM_BUFFER_INPLACE; |
old = NULL; |
if ((flags & KGEM_BUFFER_WRITE) == 0) |
old = search_linear_cache(kgem, alloc, 0); |
if (old == NULL) |
old = search_linear_cache(kgem, alloc, CREATE_INACTIVE); |
if (old) { |
DBG(("%s: reusing ordinary handle %d for io\n", |
__FUNCTION__, old->handle)); |
bo = buffer_alloc_with_data(num_pages(old)); |
if (bo == NULL) |
return NULL; |
init_buffer_from_bo(bo, old); |
bo->need_io = flags & KGEM_BUFFER_WRITE; |
} else { |
unsigned hint; |
if (use_snoopable_buffer(kgem, flags)) { |
bo = create_snoopable_buffer(kgem, alloc); |
if (bo) |
goto init; |
} |
bo = buffer_alloc(); |
if (bo == NULL) |
return NULL; |
hint = CREATE_INACTIVE; |
if (flags & KGEM_BUFFER_WRITE) |
hint |= CREATE_CPU_MAP; |
old = search_linear_cache(kgem, alloc, hint); |
if (old) { |
DBG(("%s: reusing handle=%d for buffer\n", |
__FUNCTION__, old->handle)); |
init_buffer_from_bo(bo, old); |
} else { |
uint32_t handle = gem_create(kgem->fd, alloc); |
if (handle == 0) { |
free(bo); |
return NULL; |
} |
DBG(("%s: created handle=%d for buffer\n", |
__FUNCTION__, handle)); |
__kgem_bo_init(&bo->base, handle, alloc); |
debug_alloc(kgem, alloc * PAGE_SIZE); |
} |
assert(bo->mmapped); |
assert(!bo->need_io); |
assert(bo->base.refcnt == 1); |
if (flags & KGEM_BUFFER_WRITE) { |
bo->mem = kgem_bo_map__cpu(kgem, &bo->base); |
if (bo->mem != NULL) { |
kgem_bo_sync__cpu(kgem, &bo->base); |
goto init; |
} |
} |
DBG(("%s: failing back to new pwrite buffer\n", __FUNCTION__)); |
old = &bo->base; |
bo = buffer_alloc_with_data(num_pages(old)); |
if (bo == NULL) { |
old->refcnt= 0; |
kgem_bo_free(kgem, old); |
return NULL; |
} |
init_buffer_from_bo(bo, old); |
assert(bo->mem); |
assert(!bo->mmapped); |
assert(bo->base.refcnt == 1); |
bo->need_io = flags & KGEM_BUFFER_WRITE; |
} |
init: |
bo->base.io = true; |
assert(bo->base.refcnt == 1); |
assert(num_pages(&bo->base) >= NUM_PAGES(size)); |
assert(!bo->need_io || !bo->base.needs_flush); |
assert(!bo->need_io || bo->base.domain != DOMAIN_GPU); |
assert(bo->mem); |
assert(!bo->mmapped || bo->base.map != NULL); |
bo->used = size; |
bo->write = flags & KGEM_BUFFER_WRITE_INPLACE; |
offset = 0; |
assert(list_is_empty(&bo->base.list)); |
list_add(&bo->base.list, &kgem->batch_buffers); |
DBG(("%s(pages=%d [%d]) new handle=%d, used=%d, write=%d\n", |
__FUNCTION__, num_pages(&bo->base), alloc, bo->base.handle, bo->used, bo->write)); |
done: |
bo->used = ALIGN(bo->used, UPLOAD_ALIGNMENT); |
assert(bo->mem); |
*ret = (char *)bo->mem + offset; |
return kgem_create_proxy(kgem, &bo->base, offset, size); |
} |
bool kgem_buffer_is_inplace(struct kgem_bo *_bo) |
{ |
struct kgem_buffer *bo = (struct kgem_buffer *)_bo->proxy; |
return bo->write & KGEM_BUFFER_WRITE_INPLACE; |
} |
struct kgem_bo *kgem_create_buffer_2d(struct kgem *kgem, |
int width, int height, int bpp, |
uint32_t flags, |
void **ret) |
{ |
struct kgem_bo *bo; |
int stride; |
assert(width > 0 && height > 0); |
assert(ret != NULL); |
stride = ALIGN(width, 2) * bpp >> 3; |
stride = ALIGN(stride, 4); |
DBG(("%s: %dx%d, %d bpp, stride=%d\n", |
__FUNCTION__, width, height, bpp, stride)); |
bo = kgem_create_buffer(kgem, stride * ALIGN(height, 2), flags, ret); |
if (bo == NULL) { |
DBG(("%s: allocation failure for upload buffer\n", |
__FUNCTION__)); |
return NULL; |
} |
assert(*ret != NULL); |
assert(bo->proxy != NULL); |
if (height & 1) { |
struct kgem_buffer *io = (struct kgem_buffer *)bo->proxy; |
int min; |
assert(io->used); |
/* Having padded this surface to ensure that accesses to |
* the last pair of rows is valid, remove the padding so |
* that it can be allocated to other pixmaps. |
*/ |
min = bo->delta + height * stride; |
min = ALIGN(min, UPLOAD_ALIGNMENT); |
if (io->used != min) { |
DBG(("%s: trimming buffer from %d to %d\n", |
__FUNCTION__, io->used, min)); |
io->used = min; |
} |
bo->size.bytes -= stride; |
} |
bo->map = MAKE_CPU_MAP(*ret); |
bo->pitch = stride; |
bo->unique_id = kgem_get_unique_id(kgem); |
return bo; |
} |
struct kgem_bo *kgem_upload_source_image(struct kgem *kgem, |
const void *data, |
const BoxRec *box, |
int stride, int bpp) |
{ |
int width = box->x2 - box->x1; |
int height = box->y2 - box->y1; |
struct kgem_bo *bo; |
void *dst; |
if (!kgem_can_create_2d(kgem, width, height, bpp)) |
return NULL; |
DBG(("%s : (%d, %d), (%d, %d), stride=%d, bpp=%d\n", |
__FUNCTION__, box->x1, box->y1, box->x2, box->y2, stride, bpp)); |
assert(data); |
assert(width > 0); |
assert(height > 0); |
assert(stride); |
assert(bpp); |
bo = kgem_create_buffer_2d(kgem, |
width, height, bpp, |
KGEM_BUFFER_WRITE_INPLACE, &dst); |
if (bo) |
memcpy_blt(data, dst, bpp, |
stride, bo->pitch, |
box->x1, box->y1, |
0, 0, |
width, height); |
return bo; |
} |
void kgem_proxy_bo_attach(struct kgem_bo *bo, |
struct kgem_bo **ptr) |
{ |
DBG(("%s: handle=%d\n", __FUNCTION__, bo->handle)); |
assert(bo->map == NULL || IS_CPU_MAP(bo->map)); |
assert(bo->proxy); |
list_add(&bo->vma, &bo->proxy->vma); |
bo->map = ptr; |
*ptr = kgem_bo_reference(bo); |
} |
void kgem_buffer_read_sync(struct kgem *kgem, struct kgem_bo *_bo) |
{ |
struct kgem_buffer *bo; |
uint32_t offset = _bo->delta, length = _bo->size.bytes; |
/* We expect the caller to have already submitted the batch */ |
assert(_bo->io); |
assert(_bo->exec == NULL); |
assert(_bo->rq == NULL); |
assert(_bo->proxy); |
_bo = _bo->proxy; |
assert(_bo->proxy == NULL); |
assert(_bo->exec == NULL); |
bo = (struct kgem_buffer *)_bo; |
DBG(("%s(offset=%d, length=%d, snooped=%d)\n", __FUNCTION__, |
offset, length, bo->base.snoop)); |
if (bo->mmapped) { |
struct drm_i915_gem_set_domain set_domain; |
DBG(("%s: sync: needs_flush? %d, domain? %d, busy? %d\n", |
__FUNCTION__, |
bo->base.needs_flush, |
bo->base.domain, |
__kgem_busy(kgem, bo->base.handle))); |
assert(!IS_CPU_MAP(bo->base.map) || bo->base.snoop || kgem->has_llc); |
VG_CLEAR(set_domain); |
set_domain.handle = bo->base.handle; |
set_domain.write_domain = 0; |
set_domain.read_domains = |
IS_CPU_MAP(bo->base.map) ? I915_GEM_DOMAIN_CPU : I915_GEM_DOMAIN_GTT; |
if (drmIoctl(kgem->fd, |
DRM_IOCTL_I915_GEM_SET_DOMAIN, &set_domain)) |
return; |
} else { |
if (gem_read(kgem->fd, |
bo->base.handle, (char *)bo->mem+offset, |
offset, length)) |
return; |
} |
kgem_bo_retire(kgem, &bo->base); |
bo->base.domain = DOMAIN_NONE; |
} |
#endif |
uint32_t kgem_bo_get_binding(struct kgem_bo *bo, uint32_t format) |
{ |
struct kgem_bo_binding *b; |
for (b = &bo->binding; b && b->offset; b = b->next) |
if (format == b->format) |
return b->offset; |
return 0; |
} |
void kgem_bo_set_binding(struct kgem_bo *bo, uint32_t format, uint16_t offset) |
{ |
struct kgem_bo_binding *b; |
for (b = &bo->binding; b; b = b->next) { |
if (b->offset) |
continue; |
b->offset = offset; |
b->format = format; |
if (b->next) |
b->next->offset = 0; |
return; |
} |
b = malloc(sizeof(*b)); |
if (b) { |
b->next = bo->binding.next; |
b->format = format; |
b->offset = offset; |
bo->binding.next = b; |
} |
} |
int kgem_init_fb(struct kgem *kgem, struct sna_fb *fb) |
{ |
struct kgem_bo *bo; |
size_t size; |
int ret; |
ret = drmIoctl(kgem->fd, SRV_FBINFO, fb); |
if( ret != 0 ) |
return 0; |
size = fb->pitch * fb->height / PAGE_SIZE; |
bo = __kgem_bo_alloc(-2, size); |
if (!bo) { |
return 0; |
} |
bo->domain = DOMAIN_GTT; |
bo->unique_id = kgem_get_unique_id(kgem); |
bo->pitch = fb->pitch; |
bo->tiling = I915_TILING_X; |
bo->scanout = 1; |
fb->fb_bo = bo; |
// printf("fb width %d height %d pitch %d bo %p\n", |
// fb->width, fb->height, fb->pitch, fb->fb_bo); |
return 1; |
}; |
int kgem_update_fb(struct kgem *kgem, struct sna_fb *fb) |
{ |
struct kgem_bo *bo; |
size_t size; |
int ret; |
bo = fb->fb_bo; |
ret = drmIoctl(kgem->fd, SRV_FBINFO, fb); |
if( ret != 0 ) |
return 0; |
fb->fb_bo = bo; |
size = fb->pitch * fb->height / PAGE_SIZE; |
if((size != bo->size.pages.count) || |
(fb->pitch != bo->pitch)) |
{ |
bo->size.pages.count = size; |
bo->pitch = fb->pitch; |
printf("fb width %d height %d pitch %d bo %p\n", |
fb->width, fb->height, fb->pitch, fb->fb_bo); |
return 1; |
} |
return 0; |
}; |
void sna_bo_destroy(struct kgem *kgem, struct kgem_bo *bo) |
{ |
kgem_bo_destroy(kgem, bo); |
kgem_bo_free(kgem, bo); |
} |
void kgem_close_batches(struct kgem *kgem) |
{ |
int n; |
for (n = 0; n < ARRAY_SIZE(kgem->pinned_batches); n++) { |
while (!list_is_empty(&kgem->pinned_batches[n])) { |
kgem_bo_destroy(kgem, |
list_first_entry(&kgem->pinned_batches[n], |
struct kgem_bo, list)); |
} |
} |
}; |
struct kgem_bo *kgem_bo_from_handle(struct kgem *kgem, int handle, |
int pitch, int height) |
{ |
struct kgem_bo *bo; |
int size; |
size = pitch * height / PAGE_SIZE; |
bo = __kgem_bo_alloc(handle, size); |
if(bo == NULL) |
return NULL; |
bo->domain = DOMAIN_GTT; |
bo->unique_id = kgem_get_unique_id(kgem); |
bo->pitch = pitch; |
bo->tiling = I915_TILING_X; |
bo->scanout = 0; |
return bo; |
} |
/drivers/video/Intel-2D/sna/kgem.h |
---|
0,0 → 1,761 |
/* |
* Copyright (c) 2011 Intel Corporation |
* |
* Permission is hereby granted, free of charge, to any person obtaining a |
* copy of this software and associated documentation files (the "Software"), |
* to deal in the Software without restriction, including without limitation |
* the rights to use, copy, modify, merge, publish, distribute, sublicense, |
* and/or sell copies of the Software, and to permit persons to whom the |
* Software is furnished to do so, subject to the following conditions: |
* |
* The above copyright notice and this permission notice (including the next |
* paragraph) shall be included in all copies or substantial portions of the |
* Software. |
* |
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL |
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, |
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE |
* SOFTWARE. |
* |
* Authors: |
* Chris Wilson <chris@chris-wilson.co.uk> |
* |
*/ |
#ifndef KGEM_H |
#define KGEM_H |
#include <stdint.h> |
#include <stdio.h> |
#include <stdbool.h> |
#include <stdarg.h> |
#include <i915_drm.h> |
#include "compiler.h" |
#include "intel_list.h" |
#include <limits.h> |
#if !defined(MAXSHORT) || !defined(MINSHORT) || \ |
!defined(MAXINT) || !defined(MININT) |
/* |
* Some implementations #define these through <math.h>, so preclude |
* #include'ing it later. |
*/ |
#include <math.h> |
#undef MAXSHORT |
#define MAXSHORT SHRT_MAX |
#undef MINSHORT |
#define MINSHORT SHRT_MIN |
#undef MAXINT |
#define MAXINT INT_MAX |
#undef MININT |
#define MININT INT_MIN |
#endif |
struct kgem_bo { |
struct kgem_request *rq; |
#define RQ(rq) ((struct kgem_request *)((uintptr_t)(rq) & ~3)) |
#define RQ_RING(rq) ((uintptr_t)(rq) & 3) |
#define RQ_IS_BLT(rq) (RQ_RING(rq) == KGEM_BLT) |
struct drm_i915_gem_exec_object2 *exec; |
struct kgem_bo *proxy; |
struct list list; |
struct list request; |
struct list vma; |
void *map; |
#define IS_CPU_MAP(ptr) ((uintptr_t)(ptr) & 1) |
#define IS_GTT_MAP(ptr) (ptr && ((uintptr_t)(ptr) & 1) == 0) |
#define MAP(ptr) ((void*)((uintptr_t)(ptr) & ~3)) |
struct kgem_bo_binding { |
struct kgem_bo_binding *next; |
uint32_t format; |
uint16_t offset; |
} binding; |
uint32_t unique_id; |
uint32_t refcnt; |
uint32_t handle; |
uint32_t target_handle; |
uint32_t presumed_offset; |
uint32_t delta; |
union { |
struct { |
uint32_t count:27; |
#define PAGE_SIZE 4096 |
uint32_t bucket:5; |
#define NUM_CACHE_BUCKETS 16 |
#define MAX_CACHE_SIZE (1 << (NUM_CACHE_BUCKETS+12)) |
} pages; |
uint32_t bytes; |
} size; |
uint32_t pitch : 18; /* max 128k */ |
uint32_t tiling : 2; |
uint32_t reusable : 1; |
uint32_t gpu_dirty : 1; |
uint32_t gtt_dirty : 1; |
uint32_t domain : 2; |
uint32_t needs_flush : 1; |
uint32_t snoop : 1; |
uint32_t io : 1; |
uint32_t flush : 1; |
uint32_t scanout : 1; |
uint32_t purged : 1; |
}; |
#define DOMAIN_NONE 0 |
#define DOMAIN_CPU 1 |
#define DOMAIN_GTT 2 |
#define DOMAIN_GPU 3 |
struct kgem_request { |
struct list list; |
struct kgem_bo *bo; |
struct list buffers; |
int ring; |
}; |
enum { |
MAP_GTT = 0, |
MAP_CPU, |
NUM_MAP_TYPES, |
}; |
struct kgem { |
int fd; |
int wedged; |
unsigned gen; |
uint32_t unique_id; |
enum kgem_mode { |
/* order matches I915_EXEC_RING ordering */ |
KGEM_NONE = 0, |
KGEM_RENDER, |
KGEM_BSD, |
KGEM_BLT, |
} mode, ring; |
struct list flushing; |
struct list large; |
struct list large_inactive; |
struct list active[NUM_CACHE_BUCKETS][3]; |
struct list inactive[NUM_CACHE_BUCKETS]; |
struct list pinned_batches[2]; |
struct list snoop; |
struct list scanout; |
struct list batch_buffers, active_buffers; |
struct list requests[2]; |
struct kgem_request *next_request; |
struct kgem_request static_request; |
struct { |
struct list inactive[NUM_CACHE_BUCKETS]; |
int16_t count; |
} vma[NUM_MAP_TYPES]; |
uint32_t batch_flags; |
uint32_t batch_flags_base; |
#define I915_EXEC_SECURE (1<<9) |
#define LOCAL_EXEC_OBJECT_WRITE (1<<2) |
uint16_t nbatch; |
uint16_t surface; |
uint16_t nexec; |
uint16_t nreloc; |
uint16_t nreloc__self; |
uint16_t nfence; |
uint16_t batch_size; |
uint16_t min_alignment; |
uint32_t flush:1; |
uint32_t need_expire:1; |
uint32_t need_purge:1; |
uint32_t need_retire:1; |
uint32_t need_throttle:1; |
uint32_t scanout_busy:1; |
uint32_t busy:1; |
uint32_t has_create2 :1; |
uint32_t has_userptr :1; |
uint32_t has_blt :1; |
uint32_t has_relaxed_fencing :1; |
uint32_t has_relaxed_delta :1; |
uint32_t has_semaphores :1; |
uint32_t has_secure_batches :1; |
uint32_t has_pinned_batches :1; |
uint32_t has_caching :1; |
uint32_t has_llc :1; |
uint32_t has_wt :1; |
uint32_t has_no_reloc :1; |
uint32_t has_handle_lut :1; |
uint32_t can_blt_cpu :1; |
uint16_t fence_max; |
uint16_t half_cpu_cache_pages; |
uint32_t aperture_total, aperture_high, aperture_low, aperture_mappable; |
uint32_t aperture, aperture_fenced; |
uint32_t max_upload_tile_size, max_copy_tile_size; |
uint32_t max_gpu_size, max_cpu_size; |
uint32_t large_object_size, max_object_size; |
uint32_t buffer_size; |
void (*context_switch)(struct kgem *kgem, int new_mode); |
void (*retire)(struct kgem *kgem); |
void (*expire)(struct kgem *kgem); |
#if 0 |
void (*memcpy_to_tiled_x)(const void *src, void *dst, int bpp, |
int32_t src_stride, int32_t dst_stride, |
int16_t src_x, int16_t src_y, |
int16_t dst_x, int16_t dst_y, |
uint16_t width, uint16_t height); |
void (*memcpy_from_tiled_x)(const void *src, void *dst, int bpp, |
int32_t src_stride, int32_t dst_stride, |
int16_t src_x, int16_t src_y, |
int16_t dst_x, int16_t dst_y, |
uint16_t width, uint16_t height); |
#endif |
uint16_t reloc__self[256]; |
uint32_t batch[64*1024-8] page_aligned; |
struct drm_i915_gem_exec_object2 exec[384] page_aligned; |
struct drm_i915_gem_relocation_entry reloc[8192] page_aligned; |
#ifdef DEBUG_MEMORY |
struct { |
int bo_allocs; |
size_t bo_bytes; |
} debug_memory; |
#endif |
}; |
#define KGEM_MAX_DEFERRED_VBO 16 |
#define KGEM_BATCH_RESERVED 1 |
#define KGEM_RELOC_RESERVED (KGEM_MAX_DEFERRED_VBO) |
#define KGEM_EXEC_RESERVED (1+KGEM_MAX_DEFERRED_VBO) |
#ifndef ARRAY_SIZE |
#define ARRAY_SIZE(a) (sizeof(a)/sizeof((a)[0])) |
#endif |
#define KGEM_BATCH_SIZE(K) ((K)->batch_size-KGEM_BATCH_RESERVED) |
#define KGEM_EXEC_SIZE(K) (int)(ARRAY_SIZE((K)->exec)-KGEM_EXEC_RESERVED) |
#define KGEM_RELOC_SIZE(K) (int)(ARRAY_SIZE((K)->reloc)-KGEM_RELOC_RESERVED) |
void kgem_init(struct kgem *kgem, int fd, struct pci_device *dev, unsigned gen); |
void kgem_reset(struct kgem *kgem); |
struct kgem_bo *kgem_create_map(struct kgem *kgem, |
void *ptr, uint32_t size, |
bool read_only); |
struct kgem_bo *kgem_create_for_name(struct kgem *kgem, uint32_t name); |
struct kgem_bo *kgem_create_for_prime(struct kgem *kgem, int name, uint32_t size); |
int kgem_bo_export_to_prime(struct kgem *kgem, struct kgem_bo *bo); |
struct kgem_bo *kgem_create_linear(struct kgem *kgem, int size, unsigned flags); |
struct kgem_bo *kgem_create_proxy(struct kgem *kgem, |
struct kgem_bo *target, |
int offset, int length); |
void kgem_proxy_bo_attach(struct kgem_bo *bo, struct kgem_bo **ptr); |
int kgem_choose_tiling(struct kgem *kgem, |
int tiling, int width, int height, int bpp); |
unsigned kgem_can_create_2d(struct kgem *kgem, int width, int height, int depth); |
#define KGEM_CAN_CREATE_GPU 0x1 |
#define KGEM_CAN_CREATE_CPU 0x2 |
#define KGEM_CAN_CREATE_LARGE 0x4 |
#define KGEM_CAN_CREATE_GTT 0x8 |
uint32_t kgem_get_unique_id(struct kgem *kgem); |
struct kgem_bo * |
kgem_replace_bo(struct kgem *kgem, |
struct kgem_bo *src, |
uint32_t width, |
uint32_t height, |
uint32_t pitch, |
uint32_t bpp); |
enum { |
CREATE_EXACT = 0x1, |
CREATE_INACTIVE = 0x2, |
CREATE_CPU_MAP = 0x4, |
CREATE_GTT_MAP = 0x8, |
CREATE_SCANOUT = 0x10, |
CREATE_PRIME = 0x20, |
CREATE_TEMPORARY = 0x40, |
CREATE_CACHED = 0x80, |
CREATE_NO_RETIRE = 0x100, |
CREATE_NO_THROTTLE = 0x200, |
}; |
struct kgem_bo *kgem_create_2d(struct kgem *kgem, |
int width, |
int height, |
int bpp, |
int tiling, |
uint32_t flags); |
struct kgem_bo *kgem_create_cpu_2d(struct kgem *kgem, |
int width, |
int height, |
int bpp, |
uint32_t flags); |
uint32_t kgem_bo_get_binding(struct kgem_bo *bo, uint32_t format); |
void kgem_bo_set_binding(struct kgem_bo *bo, uint32_t format, uint16_t offset); |
bool kgem_retire(struct kgem *kgem); |
bool __kgem_ring_is_idle(struct kgem *kgem, int ring); |
static inline bool kgem_ring_is_idle(struct kgem *kgem, int ring) |
{ |
ring = ring == KGEM_BLT; |
if (list_is_empty(&kgem->requests[ring])) |
return true; |
return __kgem_ring_is_idle(kgem, ring); |
} |
static inline bool kgem_is_idle(struct kgem *kgem) |
{ |
if (!kgem->need_retire) |
return true; |
return kgem_ring_is_idle(kgem, kgem->ring); |
} |
static inline bool __kgem_ring_empty(struct kgem *kgem) |
{ |
return list_is_empty(&kgem->requests[kgem->ring == KGEM_BLT]); |
} |
void _kgem_submit(struct kgem *kgem); |
static inline void kgem_submit(struct kgem *kgem) |
{ |
if (kgem->nbatch) |
_kgem_submit(kgem); |
} |
static inline bool kgem_flush(struct kgem *kgem, bool flush) |
{ |
if (kgem->nreloc == 0) |
return false; |
return (kgem->flush ^ flush) && kgem_ring_is_idle(kgem, kgem->ring); |
} |
static inline void kgem_bo_submit(struct kgem *kgem, struct kgem_bo *bo) |
{ |
if (bo->exec) |
_kgem_submit(kgem); |
} |
void kgem_scanout_flush(struct kgem *kgem, struct kgem_bo *bo); |
static inline struct kgem_bo *kgem_bo_reference(struct kgem_bo *bo) |
{ |
assert(bo->refcnt); |
bo->refcnt++; |
return bo; |
} |
void _kgem_bo_destroy(struct kgem *kgem, struct kgem_bo *bo); |
static inline void kgem_bo_destroy(struct kgem *kgem, struct kgem_bo *bo) |
{ |
assert(bo->refcnt); |
if (--bo->refcnt == 0) |
_kgem_bo_destroy(kgem, bo); |
} |
void kgem_clear_dirty(struct kgem *kgem); |
static inline void kgem_set_mode(struct kgem *kgem, |
enum kgem_mode mode, |
struct kgem_bo *bo) |
{ |
assert(!kgem->wedged); |
#if DEBUG_FLUSH_BATCH |
kgem_submit(kgem); |
#endif |
if (kgem->nreloc && bo->exec == NULL && kgem_ring_is_idle(kgem, kgem->ring)) |
_kgem_submit(kgem); |
if (kgem->mode == mode) |
return; |
// kgem->context_switch(kgem, mode); |
kgem->mode = mode; |
} |
static inline void _kgem_set_mode(struct kgem *kgem, enum kgem_mode mode) |
{ |
assert(kgem->mode == KGEM_NONE); |
assert(kgem->nbatch == 0); |
assert(!kgem->wedged); |
// kgem->context_switch(kgem, mode); |
kgem->mode = mode; |
} |
static inline bool kgem_check_batch(struct kgem *kgem, int num_dwords) |
{ |
assert(num_dwords > 0); |
assert(kgem->nbatch < kgem->surface); |
assert(kgem->surface <= kgem->batch_size); |
return likely(kgem->nbatch + num_dwords + KGEM_BATCH_RESERVED <= kgem->surface); |
} |
static inline bool kgem_check_reloc(struct kgem *kgem, int n) |
{ |
assert(kgem->nreloc <= KGEM_RELOC_SIZE(kgem)); |
return likely(kgem->nreloc + n <= KGEM_RELOC_SIZE(kgem)); |
} |
static inline bool kgem_check_exec(struct kgem *kgem, int n) |
{ |
assert(kgem->nexec <= KGEM_EXEC_SIZE(kgem)); |
return likely(kgem->nexec + n <= KGEM_EXEC_SIZE(kgem)); |
} |
static inline bool kgem_check_reloc_and_exec(struct kgem *kgem, int n) |
{ |
return kgem_check_reloc(kgem, n) && kgem_check_exec(kgem, n); |
} |
static inline bool kgem_check_batch_with_surfaces(struct kgem *kgem, |
int num_dwords, |
int num_surfaces) |
{ |
return (int)(kgem->nbatch + num_dwords + KGEM_BATCH_RESERVED) <= (int)(kgem->surface - num_surfaces*8) && |
kgem_check_reloc(kgem, num_surfaces) && |
kgem_check_exec(kgem, num_surfaces); |
} |
static inline uint32_t *kgem_get_batch(struct kgem *kgem) |
{ |
if (kgem->nreloc) { |
unsigned mode = kgem->mode; |
_kgem_submit(kgem); |
_kgem_set_mode(kgem, mode); |
} |
return kgem->batch + kgem->nbatch; |
} |
bool kgem_check_bo(struct kgem *kgem, ...) __attribute__((sentinel(0))); |
bool kgem_check_bo_fenced(struct kgem *kgem, struct kgem_bo *bo); |
bool kgem_check_many_bo_fenced(struct kgem *kgem, ...) __attribute__((sentinel(0))); |
#define KGEM_RELOC_FENCED 0x8000 |
uint32_t kgem_add_reloc(struct kgem *kgem, |
uint32_t pos, |
struct kgem_bo *bo, |
uint32_t read_write_domains, |
uint32_t delta); |
void *kgem_bo_map(struct kgem *kgem, struct kgem_bo *bo); |
void *kgem_bo_map__async(struct kgem *kgem, struct kgem_bo *bo); |
void *kgem_bo_map__gtt(struct kgem *kgem, struct kgem_bo *bo); |
void kgem_bo_sync__gtt(struct kgem *kgem, struct kgem_bo *bo); |
void *kgem_bo_map__debug(struct kgem *kgem, struct kgem_bo *bo); |
void *kgem_bo_map__cpu(struct kgem *kgem, struct kgem_bo *bo); |
void kgem_bo_sync__cpu(struct kgem *kgem, struct kgem_bo *bo); |
void kgem_bo_sync__cpu_full(struct kgem *kgem, struct kgem_bo *bo, bool write); |
void *__kgem_bo_map__cpu(struct kgem *kgem, struct kgem_bo *bo); |
void __kgem_bo_unmap__cpu(struct kgem *kgem, struct kgem_bo *bo, void *ptr); |
uint32_t kgem_bo_flink(struct kgem *kgem, struct kgem_bo *bo); |
bool kgem_bo_write(struct kgem *kgem, struct kgem_bo *bo, |
const void *data, int length); |
int kgem_bo_fenced_size(struct kgem *kgem, struct kgem_bo *bo); |
void kgem_get_tile_size(struct kgem *kgem, int tiling, |
int *tile_width, int *tile_height, int *tile_size); |
static inline int __kgem_buffer_size(struct kgem_bo *bo) |
{ |
assert(bo->proxy != NULL); |
return bo->size.bytes; |
} |
static inline int __kgem_bo_size(struct kgem_bo *bo) |
{ |
assert(bo->proxy == NULL); |
return PAGE_SIZE * bo->size.pages.count; |
} |
static inline int kgem_bo_size(struct kgem_bo *bo) |
{ |
if (bo->proxy) |
return __kgem_buffer_size(bo); |
else |
return __kgem_bo_size(bo); |
} |
/* |
static inline bool kgem_bo_blt_pitch_is_ok(struct kgem *kgem, |
struct kgem_bo *bo) |
{ |
int pitch = bo->pitch; |
if (kgem->gen >= 040 && bo->tiling) |
pitch /= 4; |
if (pitch > MAXSHORT) { |
DBG(("%s: can not blt to handle=%d, adjusted pitch=%d\n", |
__FUNCTION__, bo->handle, pitch)); |
return false; |
} |
return true; |
} |
static inline bool kgem_bo_can_blt(struct kgem *kgem, |
struct kgem_bo *bo) |
{ |
if (bo->tiling == I915_TILING_Y) { |
DBG(("%s: can not blt to handle=%d, tiling=Y\n", |
__FUNCTION__, bo->handle)); |
return false; |
} |
return kgem_bo_blt_pitch_is_ok(kgem, bo); |
} |
*/ |
static inline bool __kgem_bo_is_mappable(struct kgem *kgem, |
struct kgem_bo *bo) |
{ |
if (bo->domain == DOMAIN_GTT) |
return true; |
if (kgem->gen < 040 && bo->tiling && |
bo->presumed_offset & (kgem_bo_fenced_size(kgem, bo) - 1)) |
return false; |
if (kgem->gen == 021 && bo->tiling == I915_TILING_Y) |
return false; |
if (kgem->has_llc && bo->tiling == I915_TILING_NONE) |
return true; |
if (!bo->presumed_offset) |
return kgem_bo_size(bo) <= kgem->aperture_mappable / 4; |
return bo->presumed_offset + kgem_bo_size(bo) <= kgem->aperture_mappable; |
} |
static inline bool kgem_bo_is_mappable(struct kgem *kgem, |
struct kgem_bo *bo) |
{ |
DBG(("%s: domain=%d, offset: %d size: %d\n", |
__FUNCTION__, bo->domain, bo->presumed_offset, kgem_bo_size(bo))); |
assert(bo->refcnt); |
return __kgem_bo_is_mappable(kgem, bo); |
} |
static inline bool kgem_bo_mapped(struct kgem *kgem, struct kgem_bo *bo) |
{ |
DBG(("%s: map=%p, tiling=%d, domain=%d\n", |
__FUNCTION__, bo->map, bo->tiling, bo->domain)); |
assert(bo->refcnt); |
if (bo->map == NULL) |
return bo->tiling == I915_TILING_NONE && bo->domain == DOMAIN_CPU; |
return IS_CPU_MAP(bo->map) == !bo->tiling; |
} |
static inline bool kgem_bo_can_map(struct kgem *kgem, struct kgem_bo *bo) |
{ |
if (kgem_bo_mapped(kgem, bo)) |
return true; |
if (!bo->tiling && (kgem->has_llc || bo->domain == DOMAIN_CPU)) |
return true; |
if (kgem->gen == 021 && bo->tiling == I915_TILING_Y) |
return false; |
return kgem_bo_size(bo) <= kgem->aperture_mappable / 4; |
} |
static inline bool kgem_bo_can_map__cpu(struct kgem *kgem, |
struct kgem_bo *bo, |
bool write) |
{ |
if (bo->purged || (bo->scanout && write)) |
return false; |
if (kgem->has_llc) |
return true; |
if (bo->domain != DOMAIN_CPU) |
return false; |
return !write || bo->exec == NULL; |
} |
static inline bool kgem_bo_is_snoop(struct kgem_bo *bo) |
{ |
assert(bo->refcnt); |
while (bo->proxy) |
bo = bo->proxy; |
return bo->snoop; |
} |
void kgem_bo_undo(struct kgem *kgem, struct kgem_bo *bo); |
bool __kgem_busy(struct kgem *kgem, int handle); |
static inline void kgem_bo_mark_busy(struct kgem_bo *bo, int ring) |
{ |
bo->rq = (struct kgem_request *)((uintptr_t)bo->rq | ring); |
} |
inline static void __kgem_bo_clear_busy(struct kgem_bo *bo) |
{ |
bo->rq = NULL; |
list_del(&bo->request); |
bo->domain = DOMAIN_NONE; |
bo->needs_flush = false; |
bo->gtt_dirty = false; |
} |
static inline bool kgem_bo_is_busy(struct kgem_bo *bo) |
{ |
DBG(("%s: handle=%d, domain: %d exec? %d, rq? %d\n", __FUNCTION__, |
bo->handle, bo->domain, bo->exec != NULL, bo->rq != NULL)); |
assert(bo->refcnt); |
return bo->rq; |
} |
static inline bool __kgem_bo_is_busy(struct kgem *kgem, struct kgem_bo *bo) |
{ |
DBG(("%s: handle=%d, domain: %d exec? %d, rq? %d\n", __FUNCTION__, |
bo->handle, bo->domain, bo->exec != NULL, bo->rq != NULL)); |
assert(bo->refcnt); |
if (bo->exec) |
return true; |
if (kgem_flush(kgem, bo->flush)) |
kgem_submit(kgem); |
if (bo->rq && !__kgem_busy(kgem, bo->handle)) |
__kgem_bo_clear_busy(bo); |
return kgem_bo_is_busy(bo); |
} |
static inline bool kgem_bo_is_render(struct kgem_bo *bo) |
{ |
DBG(("%s: handle=%d, rq? %d [%d]\n", __FUNCTION__, |
bo->handle, bo->rq != NULL, (int)RQ_RING(bo->rq))); |
assert(bo->refcnt); |
return bo->rq && RQ_RING(bo->rq) == I915_EXEC_RENDER; |
} |
static inline void kgem_bo_mark_unreusable(struct kgem_bo *bo) |
{ |
while (bo->proxy) { |
bo->flush = true; |
bo = bo->proxy; |
} |
bo->flush = true; |
bo->reusable = false; |
} |
static inline bool kgem_bo_is_dirty(struct kgem_bo *bo) |
{ |
if (bo == NULL) |
return false; |
assert(bo->refcnt); |
return bo->gpu_dirty; |
} |
static inline void kgem_bo_unclean(struct kgem *kgem, struct kgem_bo *bo) |
{ |
/* The bo is outside of our control, so presume it is written to */ |
bo->needs_flush = true; |
if (bo->rq == NULL) |
bo->rq = (void *)kgem; |
if (bo->domain != DOMAIN_GPU) |
bo->domain = DOMAIN_NONE; |
} |
static inline void __kgem_bo_mark_dirty(struct kgem_bo *bo) |
{ |
DBG(("%s: handle=%d (proxy? %d)\n", __FUNCTION__, |
bo->handle, bo->proxy != NULL)); |
bo->exec->flags |= LOCAL_EXEC_OBJECT_WRITE; |
bo->needs_flush = bo->gpu_dirty = true; |
list_move(&bo->request, &RQ(bo->rq)->buffers); |
} |
static inline void kgem_bo_mark_dirty(struct kgem_bo *bo) |
{ |
assert(bo->refcnt); |
do { |
assert(bo->exec); |
assert(bo->rq); |
if (bo->gpu_dirty) |
return; |
__kgem_bo_mark_dirty(bo); |
} while ((bo = bo->proxy)); |
} |
#define KGEM_BUFFER_WRITE 0x1 |
#define KGEM_BUFFER_INPLACE 0x2 |
#define KGEM_BUFFER_LAST 0x4 |
#define KGEM_BUFFER_WRITE_INPLACE (KGEM_BUFFER_WRITE | KGEM_BUFFER_INPLACE) |
struct kgem_bo *kgem_create_buffer(struct kgem *kgem, |
uint32_t size, uint32_t flags, |
void **ret); |
struct kgem_bo *kgem_create_buffer_2d(struct kgem *kgem, |
int width, int height, int bpp, |
uint32_t flags, |
void **ret); |
bool kgem_buffer_is_inplace(struct kgem_bo *bo); |
void kgem_buffer_read_sync(struct kgem *kgem, struct kgem_bo *bo); |
void kgem_throttle(struct kgem *kgem); |
#define MAX_INACTIVE_TIME 10 |
bool kgem_expire_cache(struct kgem *kgem); |
void kgem_purge_cache(struct kgem *kgem); |
void kgem_cleanup_cache(struct kgem *kgem); |
void kgem_clean_scanout_cache(struct kgem *kgem); |
void kgem_clean_large_cache(struct kgem *kgem); |
#if HAS_DEBUG_FULL |
void __kgem_batch_debug(struct kgem *kgem, uint32_t nbatch); |
#else |
static inline void __kgem_batch_debug(struct kgem *kgem, uint32_t nbatch) |
{ |
(void)kgem; |
(void)nbatch; |
} |
#endif |
#endif /* KGEM_H */ |
/drivers/video/Intel-2D/sna/kgem_debug.c |
---|
0,0 → 1,424 |
/* |
* Copyright © 2007-2011 Intel Corporation |
* |
* Permission is hereby granted, free of charge, to any person obtaining a |
* copy of this software and associated documentation files (the "Software"), |
* to deal in the Software without restriction, including without limitation |
* the rights to use, copy, modify, merge, publish, distribute, sublicense, |
* and/or sell copies of the Software, and to permit persons to whom the |
* Software is furnished to do so, subject to the following conditions: |
* |
* The above copyright notice and this permission notice (including the next |
* paragraph) shall be included in all copies or substantial portions of the |
* Software. |
* |
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL |
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, |
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE |
* SOFTWARE. |
* |
* Authors: |
* Eric Anholt <eric@anholt.net> |
* |
*/ |
#ifdef HAVE_CONFIG_H |
#include "config.h" |
#endif |
//#include <sys/mman.h> |
#include <assert.h> |
#include "sna.h" |
#include "sna_reg.h" |
#include "kgem_debug.h" |
struct drm_i915_gem_relocation_entry * |
kgem_debug_get_reloc_entry(struct kgem *kgem, uint32_t offset) |
{ |
int i; |
offset *= sizeof(uint32_t); |
for (i = 0; i < kgem->nreloc; i++) |
if (kgem->reloc[i].offset == offset) |
return kgem->reloc+i; |
assert(!"valid relocation entry, unknown batch offset"); |
return NULL; |
} |
struct kgem_bo * |
kgem_debug_get_bo_for_reloc_entry(struct kgem *kgem, |
struct drm_i915_gem_relocation_entry *reloc) |
{ |
struct kgem_bo *bo; |
if (reloc == NULL) |
return NULL; |
list_for_each_entry(bo, &kgem->next_request->buffers, request) |
if (bo->target_handle == reloc->target_handle && bo->proxy == NULL) |
break; |
assert(&bo->request != &kgem->next_request->buffers); |
return bo; |
} |
static int kgem_debug_handle_is_fenced(struct kgem *kgem, uint32_t handle) |
{ |
int i; |
if (kgem->has_handle_lut) |
return kgem->exec[handle].flags & EXEC_OBJECT_NEEDS_FENCE; |
for (i = 0; i < kgem->nexec; i++) |
if (kgem->exec[i].handle == handle) |
return kgem->exec[i].flags & EXEC_OBJECT_NEEDS_FENCE; |
return 0; |
} |
static int kgem_debug_handle_tiling(struct kgem *kgem, uint32_t handle) |
{ |
struct kgem_bo *bo; |
list_for_each_entry(bo, &kgem->next_request->buffers, request) |
if (bo->target_handle == handle) |
return bo->tiling; |
return 0; |
} |
void |
kgem_debug_print(const uint32_t *data, |
uint32_t offset, unsigned int index, |
const char *fmt, ...) |
{ |
va_list va; |
char buf[240]; |
int len; |
len = snprintf(buf, sizeof(buf), |
"0x%08x: 0x%08x: %s", |
(offset + index) * 4, |
data[index], |
index == 0 ? "" : " "); |
va_start(va, fmt); |
vsnprintf(buf + len, sizeof(buf) - len, fmt, va); |
va_end(va); |
ErrorF("%s", buf); |
delay(1); |
} |
static int |
decode_nop(struct kgem *kgem, uint32_t offset) |
{ |
uint32_t *data = kgem->batch + offset; |
kgem_debug_print(data, offset, 0, "UNKNOWN\n"); |
assert(0); |
return 1; |
} |
static int |
decode_mi(struct kgem *kgem, uint32_t offset) |
{ |
static const struct { |
uint32_t opcode; |
int len_mask; |
int min_len; |
int max_len; |
const char *name; |
} opcodes[] = { |
{ 0x08, 0, 1, 1, "MI_ARB_ON_OFF" }, |
{ 0x0a, 0, 1, 1, "MI_BATCH_BUFFER_END" }, |
{ 0x30, 0x3f, 3, 3, "MI_BATCH_BUFFER" }, |
{ 0x31, 0x3f, 2, 2, "MI_BATCH_BUFFER_START" }, |
{ 0x14, 0x3f, 3, 3, "MI_DISPLAY_BUFFER_INFO" }, |
{ 0x04, 0, 1, 1, "MI_FLUSH" }, |
{ 0x22, 0x1f, 3, 3, "MI_LOAD_REGISTER_IMM" }, |
{ 0x13, 0x3f, 2, 2, "MI_LOAD_SCAN_LINES_EXCL" }, |
{ 0x12, 0x3f, 2, 2, "MI_LOAD_SCAN_LINES_INCL" }, |
{ 0x00, 0, 1, 1, "MI_NOOP" }, |
{ 0x11, 0x3f, 2, 2, "MI_OVERLAY_FLIP" }, |
{ 0x07, 0, 1, 1, "MI_REPORT_HEAD" }, |
{ 0x18, 0x3f, 2, 2, "MI_SET_CONTEXT" }, |
{ 0x20, 0x3f, 3, 4, "MI_STORE_DATA_IMM" }, |
{ 0x21, 0x3f, 3, 4, "MI_STORE_DATA_INDEX" }, |
{ 0x24, 0x3f, 3, 3, "MI_STORE_REGISTER_MEM" }, |
{ 0x02, 0, 1, 1, "MI_USER_INTERRUPT" }, |
{ 0x03, 0, 1, 1, "MI_WAIT_FOR_EVENT" }, |
{ 0x16, 0x7f, 3, 3, "MI_SEMAPHORE_MBOX" }, |
{ 0x26, 0x1f, 3, 4, "MI_FLUSH_DW" }, |
{ 0x0b, 0, 1, 1, "MI_SUSPEND_FLUSH" }, |
}; |
uint32_t *data = kgem->batch + offset; |
int op; |
for (op = 0; op < ARRAY_SIZE(opcodes); op++) { |
if ((data[0] & 0x1f800000) >> 23 == opcodes[op].opcode) { |
unsigned int len = 1, i; |
kgem_debug_print(data, offset, 0, "%s\n", opcodes[op].name); |
if (opcodes[op].max_len > 1) { |
len = (data[0] & opcodes[op].len_mask) + 2; |
if (len < opcodes[op].min_len || |
len > opcodes[op].max_len) |
{ |
ErrorF("Bad length (%d) in %s, [%d, %d]\n", |
len, opcodes[op].name, |
opcodes[op].min_len, |
opcodes[op].max_len); |
assert(0); |
} |
} |
for (i = 1; i < len; i++) |
kgem_debug_print(data, offset, i, "dword %d\n", i); |
return len; |
} |
} |
kgem_debug_print(data, offset, 0, "MI UNKNOWN\n"); |
assert(0); |
return 1; |
} |
static int |
decode_2d(struct kgem *kgem, uint32_t offset) |
{ |
static const struct { |
uint32_t opcode; |
int min_len; |
int max_len; |
const char *name; |
} opcodes[] = { |
{ 0x40, 5, 5, "COLOR_BLT" }, |
{ 0x43, 6, 6, "SRC_COPY_BLT" }, |
{ 0x01, 8, 8, "XY_SETUP_BLT" }, |
{ 0x11, 9, 9, "XY_SETUP_MONO_PATTERN_SL_BLT" }, |
{ 0x03, 3, 3, "XY_SETUP_CLIP_BLT" }, |
{ 0x24, 2, 2, "XY_PIXEL_BLT" }, |
{ 0x25, 3, 3, "XY_SCANLINES_BLT" }, |
{ 0x26, 4, 4, "Y_TEXT_BLT" }, |
{ 0x31, 5, 134, "XY_TEXT_IMMEDIATE_BLT" }, |
{ 0x50, 6, 6, "XY_COLOR_BLT" }, |
{ 0x51, 6, 6, "XY_PAT_BLT" }, |
{ 0x76, 8, 8, "XY_PAT_CHROMA_BLT" }, |
{ 0x72, 7, 135, "XY_PAT_BLT_IMMEDIATE" }, |
{ 0x77, 9, 137, "XY_PAT_CHROMA_BLT_IMMEDIATE" }, |
{ 0x52, 9, 9, "XY_MONO_PAT_BLT" }, |
{ 0x59, 7, 7, "XY_MONO_PAT_FIXED_BLT" }, |
{ 0x53, 8, 8, "XY_SRC_COPY_BLT" }, |
{ 0x54, 8, 8, "XY_MONO_SRC_COPY_BLT" }, |
{ 0x71, 9, 137, "XY_MONO_SRC_COPY_IMMEDIATE_BLT" }, |
{ 0x55, 9, 9, "XY_FULL_BLT" }, |
{ 0x55, 9, 137, "XY_FULL_IMMEDIATE_PATTERN_BLT" }, |
{ 0x56, 9, 9, "XY_FULL_MONO_SRC_BLT" }, |
{ 0x75, 10, 138, "XY_FULL_MONO_SRC_IMMEDIATE_PATTERN_BLT" }, |
{ 0x57, 12, 12, "XY_FULL_MONO_PATTERN_BLT" }, |
{ 0x58, 12, 12, "XY_FULL_MONO_PATTERN_MONO_SRC_BLT" }, |
}; |
unsigned int op, len; |
const char *format = NULL; |
uint32_t *data = kgem->batch + offset; |
struct drm_i915_gem_relocation_entry *reloc; |
/* Special case the two most common ops that we detail in full */ |
switch ((data[0] & 0x1fc00000) >> 22) { |
case 0x50: |
kgem_debug_print(data, offset, 0, |
"XY_COLOR_BLT (rgb %sabled, alpha %sabled, dst tile %d)\n", |
(data[0] & (1 << 20)) ? "en" : "dis", |
(data[0] & (1 << 21)) ? "en" : "dis", |
(data[0] >> 11) & 1); |
len = (data[0] & 0x000000ff) + 2; |
assert(len == 6); |
switch ((data[1] >> 24) & 0x3) { |
case 0: |
format="8"; |
break; |
case 1: |
format="565"; |
break; |
case 2: |
format="1555"; |
break; |
case 3: |
format="8888"; |
break; |
} |
kgem_debug_print(data, offset, 1, "format %s, rop %x, pitch %d, " |
"clipping %sabled\n", format, |
(data[1] >> 16) & 0xff, |
(short)(data[1] & 0xffff), |
data[1] & (1 << 30) ? "en" : "dis"); |
kgem_debug_print(data, offset, 2, "(%d,%d)\n", |
data[2] & 0xffff, data[2] >> 16); |
kgem_debug_print(data, offset, 3, "(%d,%d)\n", |
data[3] & 0xffff, data[3] >> 16); |
reloc = kgem_debug_get_reloc_entry(kgem, offset+4); |
kgem_debug_print(data, offset, 4, "dst offset 0x%08x [handle=%d, delta=%d, read=%x, write=%x (fenced? %d, tiling? %d)]\n", |
data[4], |
reloc->target_handle, reloc->delta, |
reloc->read_domains, reloc->write_domain, |
kgem_debug_handle_is_fenced(kgem, reloc->target_handle), |
kgem_debug_handle_tiling(kgem, reloc->target_handle)); |
kgem_debug_print(data, offset, 5, "color\n"); |
assert(kgem->gen >= 040 || |
kgem_debug_handle_is_fenced(kgem, reloc->target_handle)); |
return len; |
case 0x53: |
kgem_debug_print(data, offset, 0, |
"XY_SRC_COPY_BLT (rgb %sabled, alpha %sabled, " |
"src tile %d, dst tile %d)\n", |
(data[0] & (1 << 20)) ? "en" : "dis", |
(data[0] & (1 << 21)) ? "en" : "dis", |
(data[0] >> 15) & 1, |
(data[0] >> 11) & 1); |
len = (data[0] & 0x000000ff) + 2; |
assert(len == 8); |
switch ((data[1] >> 24) & 0x3) { |
case 0: |
format="8"; |
break; |
case 1: |
format="565"; |
break; |
case 2: |
format="1555"; |
break; |
case 3: |
format="8888"; |
break; |
} |
kgem_debug_print(data, offset, 1, "format %s, rop %x, dst pitch %d, " |
"clipping %sabled\n", format, |
(data[1] >> 16) & 0xff, |
(short)(data[1] & 0xffff), |
data[1] & (1 << 30) ? "en" : "dis"); |
kgem_debug_print(data, offset, 2, "dst (%d,%d)\n", |
data[2] & 0xffff, data[2] >> 16); |
kgem_debug_print(data, offset, 3, "dst (%d,%d)\n", |
data[3] & 0xffff, data[3] >> 16); |
reloc = kgem_debug_get_reloc_entry(kgem, offset+4); |
assert(reloc); |
kgem_debug_print(data, offset, 4, "dst offset 0x%08x [handle=%d, delta=%d, read=%x, write=%x, (fenced? %d, tiling? %d)]\n", |
data[4], |
reloc->target_handle, reloc->delta, |
reloc->read_domains, reloc->write_domain, |
kgem_debug_handle_is_fenced(kgem, reloc->target_handle), |
kgem_debug_handle_tiling(kgem, reloc->target_handle)); |
assert(kgem->gen >= 040 || |
kgem_debug_handle_is_fenced(kgem, reloc->target_handle)); |
kgem_debug_print(data, offset, 5, "src (%d,%d)\n", |
data[5] & 0xffff, data[5] >> 16); |
kgem_debug_print(data, offset, 6, "src pitch %d\n", |
(short)(data[6] & 0xffff)); |
reloc = kgem_debug_get_reloc_entry(kgem, offset+7); |
assert(reloc); |
kgem_debug_print(data, offset, 7, "src offset 0x%08x [handle=%d, delta=%d, read=%x, write=%x (fenced? %d, tiling? %d)]\n", |
data[7], |
reloc->target_handle, reloc->delta, |
reloc->read_domains, reloc->write_domain, |
kgem_debug_handle_is_fenced(kgem, reloc->target_handle), |
kgem_debug_handle_tiling(kgem, reloc->target_handle)); |
assert(kgem->gen >= 040 || |
kgem_debug_handle_is_fenced(kgem, reloc->target_handle)); |
return len; |
} |
for (op = 0; op < ARRAY_SIZE(opcodes); op++) { |
if ((data[0] & 0x1fc00000) >> 22 == opcodes[op].opcode) { |
unsigned int i; |
len = 1; |
kgem_debug_print(data, offset, 0, "%s\n", opcodes[op].name); |
if (opcodes[op].max_len > 1) { |
len = (data[0] & 0x000000ff) + 2; |
assert(len >= opcodes[op].min_len && |
len <= opcodes[op].max_len); |
} |
for (i = 1; i < len; i++) |
kgem_debug_print(data, offset, i, "dword %d\n", i); |
return len; |
} |
} |
kgem_debug_print(data, offset, 0, "2D UNKNOWN\n"); |
assert(0); |
return 1; |
} |
static int (*decode_3d(int gen))(struct kgem*, uint32_t) |
{ |
if (gen >= 0100) { |
} else if (gen >= 070) { |
return kgem_gen7_decode_3d; |
} else if (gen >= 060) { |
return kgem_gen6_decode_3d; |
} else if (gen >= 050) { |
return kgem_gen5_decode_3d; |
} else if (gen >= 040) { |
return kgem_gen4_decode_3d; |
} else if (gen >= 030) { |
return kgem_gen3_decode_3d; |
} |
assert(0); |
} |
static void (*finish_state(int gen))(struct kgem*) |
{ |
if (gen >= 0100) { |
} else if (gen >= 070) { |
return kgem_gen7_finish_state; |
} else if (gen >= 060) { |
return kgem_gen6_finish_state; |
} else if (gen >= 050) { |
return kgem_gen5_finish_state; |
} else if (gen >= 040) { |
return kgem_gen4_finish_state; |
} else if (gen >= 030) { |
return kgem_gen3_finish_state; |
} |
assert(0); |
} |
void __kgem_batch_debug(struct kgem *kgem, uint32_t nbatch) |
{ |
int (*const decode[])(struct kgem *, uint32_t) = { |
decode_mi, |
decode_nop, |
decode_2d, |
decode_3d(kgem->gen), |
}; |
uint32_t offset = 0; |
while (offset < nbatch) { |
int class = (kgem->batch[offset] & 0xe0000000) >> 29; |
assert(class < ARRAY_SIZE(decode)); |
offset += decode[class](kgem, offset); |
} |
finish_state(kgem->gen)(kgem); |
} |
/drivers/video/Intel-2D/sna/kgem_debug.h |
---|
0,0 → 1,34 |
#ifndef KGEM_DEBUG_H |
#define KGEM_DEBUG_H |
void |
kgem_debug_print(const uint32_t *data, |
uint32_t offset, unsigned int index, |
const char *fmt, ...); |
struct drm_i915_gem_relocation_entry * |
kgem_debug_get_reloc_entry(struct kgem *kgem, uint32_t offset); |
struct kgem_bo * |
kgem_debug_get_bo_for_reloc_entry(struct kgem *kgem, |
struct drm_i915_gem_relocation_entry *reloc); |
int kgem_gen7_decode_3d(struct kgem *kgem, uint32_t offset); |
void kgem_gen7_finish_state(struct kgem *kgem); |
int kgem_gen6_decode_3d(struct kgem *kgem, uint32_t offset); |
void kgem_gen6_finish_state(struct kgem *kgem); |
int kgem_gen5_decode_3d(struct kgem *kgem, uint32_t offset); |
void kgem_gen5_finish_state(struct kgem *kgem); |
int kgem_gen4_decode_3d(struct kgem *kgem, uint32_t offset); |
void kgem_gen4_finish_state(struct kgem *kgem); |
int kgem_gen3_decode_3d(struct kgem *kgem, uint32_t offset); |
void kgem_gen3_finish_state(struct kgem *kgem); |
int kgem_gen2_decode_3d(struct kgem *kgem, uint32_t offset); |
void kgem_gen2_finish_state(struct kgem *kgem); |
#endif |
/drivers/video/Intel-2D/sna/kgem_debug_gen3.c |
---|
0,0 → 1,1599 |
/* |
* Copyright © 2007-2011 Intel Corporation |
* |
* Permission is hereby granted, free of charge, to any person obtaining a |
* copy of this software and associated documentation files (the "Software"), |
* to deal in the Software without restriction, including without limitation |
* the rights to use, copy, modify, merge, publish, distribute, sublicense, |
* and/or sell copies of the Software, and to permit persons to whom the |
* Software is furnished to do so, subject to the following conditions: |
* |
* The above copyright notice and this permission notice (including the next |
* paragraph) shall be included in all copies or substantial portions of the |
* Software. |
* |
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL |
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, |
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE |
* SOFTWARE. |
* |
* Authors: |
* Eric Anholt <eric@anholt.net> |
* Chris Wilson <chris@chris-wilson.co.uk> |
* |
*/ |
#ifdef HAVE_CONFIG_H |
#include "config.h" |
#endif |
#include <assert.h> |
#include "sna.h" |
#include "sna_reg.h" |
#include "gen3_render.h" |
#include "kgem_debug.h" |
enum type { |
T_FLOAT32, |
T_FLOAT16, |
}; |
static struct state { |
struct vertex_buffer { |
int handle; |
void *base; |
const char *ptr; |
int pitch; |
struct kgem_bo *current; |
} vb; |
struct vertex_elements { |
int offset; |
bool valid; |
enum type type; |
int size; |
uint8_t swizzle[4]; |
} ve[33]; |
int num_ve; |
} state; |
static float int_as_float(int i) |
{ |
union { |
float f; |
int i; |
} x; |
x.i = i; |
return x.f; |
} |
static void gen3_update_vertex_buffer_addr(struct kgem *kgem, |
uint32_t offset) |
{ |
uint32_t handle; |
struct kgem_bo *bo = NULL; |
void *base, *ptr; |
int i; |
offset *= sizeof(uint32_t); |
for (i = 0; i < kgem->nreloc; i++) |
if (kgem->reloc[i].offset == offset) |
break; |
assert(i < kgem->nreloc); |
handle = kgem->reloc[i].target_handle; |
if (handle == 0) { |
base = kgem->batch; |
} else { |
list_for_each_entry(bo, &kgem->next_request->buffers, request) |
if (bo->handle == handle) |
break; |
assert(&bo->request != &kgem->next_request->buffers); |
base = kgem_bo_map__debug(kgem, bo); |
} |
ptr = (char *)base + kgem->reloc[i].delta; |
state.vb.current = bo; |
state.vb.base = base; |
state.vb.ptr = ptr; |
} |
static void gen3_update_vertex_buffer_pitch(struct kgem *kgem, |
uint32_t offset) |
{ |
state.vb.pitch = kgem->batch[offset] >> 16 & 0x3f; |
state.vb.pitch *= sizeof(uint32_t); |
} |
static void gen3_update_vertex_elements(struct kgem *kgem, uint32_t data) |
{ |
state.ve[1].valid = 1; |
switch ((data >> 6) & 7) { |
case 1: |
state.ve[1].type = T_FLOAT32; |
state.ve[1].size = 3; |
state.ve[1].swizzle[0] = 1; |
state.ve[1].swizzle[1] = 1; |
state.ve[1].swizzle[2] = 1; |
state.ve[1].swizzle[3] = 3; |
break; |
case 2: |
state.ve[1].type = T_FLOAT32; |
state.ve[1].size = 4; |
state.ve[1].swizzle[0] = 1; |
state.ve[1].swizzle[1] = 1; |
state.ve[1].swizzle[2] = 1; |
state.ve[1].swizzle[3] = 1; |
break; |
case 3: |
state.ve[1].type = T_FLOAT32; |
state.ve[1].size = 2; |
state.ve[1].swizzle[0] = 1; |
state.ve[1].swizzle[1] = 1; |
state.ve[1].swizzle[2] = 2; |
state.ve[1].swizzle[3] = 3; |
break; |
case 4: |
state.ve[1].type = T_FLOAT32; |
state.ve[1].size = 3; |
state.ve[1].swizzle[0] = 1; |
state.ve[1].swizzle[1] = 1; |
state.ve[1].swizzle[2] = 3; |
state.ve[1].swizzle[3] = 1; |
break; |
} |
state.ve[2].valid = 0; |
state.ve[3].valid = 0; |
} |
static void gen3_update_vertex_texcoords(struct kgem *kgem, uint32_t data) |
{ |
int id; |
for (id = 0; id < 8; id++) { |
uint32_t fmt = (data >> (id*4)) & 0xf; |
int width; |
state.ve[id+4].valid = fmt != 0xf; |
width = 0; |
switch (fmt) { |
case 0: |
state.ve[id+4].type = T_FLOAT32; |
width = state.ve[id+4].size = 2; |
break; |
case 1: |
state.ve[id+4].type = T_FLOAT32; |
width = state.ve[id+4].size = 3; |
break; |
case 2: |
state.ve[id+4].type = T_FLOAT32; |
width = state.ve[id+4].size = 4; |
break; |
case 3: |
state.ve[id+4].type = T_FLOAT32; |
width = state.ve[id+4].size = 1; |
break; |
case 4: |
state.ve[id+4].type = T_FLOAT16; |
width = state.ve[id+4].size = 2; |
break; |
case 5: |
state.ve[id+4].type = T_FLOAT16; |
width = state.ve[id+4].size = 4; |
break; |
} |
state.ve[id+4].swizzle[0] = width > 0 ? 1 : 2; |
state.ve[id+4].swizzle[1] = width > 1 ? 1 : 2; |
state.ve[id+4].swizzle[2] = width > 2 ? 1 : 2; |
state.ve[id+4].swizzle[3] = width > 3 ? 1 : 2; |
} |
} |
static void gen3_update_vertex_elements_offsets(struct kgem *kgem) |
{ |
int i, offset; |
for (i = offset = 0; i < ARRAY_SIZE(state.ve); i++) { |
if (!state.ve[i].valid) |
continue; |
state.ve[i].offset = offset; |
offset += 4 * state.ve[i].size; |
state.num_ve = i; |
} |
} |
static void vertices_float32_out(const struct vertex_elements *ve, const float *f, int max) |
{ |
int c; |
ErrorF("("); |
for (c = 0; c < max; c++) { |
switch (ve->swizzle[c]) { |
case 0: ErrorF("#"); break; |
case 1: ErrorF("%f", f[c]); break; |
case 2: ErrorF("0.0"); break; |
case 3: ErrorF("1.0"); break; |
case 4: ErrorF("0x1"); break; |
case 5: break; |
default: ErrorF("?"); |
} |
if (c < max-1) |
ErrorF(", "); |
} |
ErrorF(")"); |
} |
static void ve_out(const struct vertex_elements *ve, const void *ptr) |
{ |
switch (ve->type) { |
case T_FLOAT32: |
vertices_float32_out(ve, ptr, ve->size); |
break; |
case T_FLOAT16: |
//vertices_float16_out(ve, ptr, ve->size); |
break; |
} |
} |
static void indirect_vertex_out(struct kgem *kgem, uint32_t v) |
{ |
const struct vertex_buffer *vb = &state.vb; |
int i = 1; |
do { |
const struct vertex_elements *ve = &state.ve[i]; |
const void *ptr = vb->ptr + v * vb->pitch + ve->offset; |
if (!ve->valid) |
continue; |
ve_out(ve, ptr); |
while (++i <= state.num_ve && !state.ve[i].valid) |
; |
if (i <= state.num_ve) |
ErrorF(", "); |
} while (i <= state.num_ve); |
} |
static int inline_vertex_out(struct kgem *kgem, void *base) |
{ |
const struct vertex_buffer *vb = &state.vb; |
int i = 1; |
do { |
const struct vertex_elements *ve = &state.ve[i]; |
const void *ptr = (char *)base + ve->offset; |
if (!ve->valid) |
continue; |
ve_out(ve, ptr); |
while (++i <= state.num_ve && !state.ve[i].valid) |
; |
if (i <= state.num_ve) |
ErrorF(", "); |
} while (i <= state.num_ve); |
return vb->pitch; |
} |
static int |
gen3_decode_3d_1c(struct kgem *kgem, uint32_t offset) |
{ |
uint32_t *data = kgem->batch + offset; |
uint32_t opcode; |
opcode = (data[0] & 0x00f80000) >> 19; |
switch (opcode) { |
case 0x11: |
kgem_debug_print(data, offset, 0, "3DSTATE_DEPTH_SUBRECTANGLE_DISABLE\n"); |
return 1; |
case 0x10: |
kgem_debug_print(data, offset, 0, "3DSTATE_SCISSOR_ENABLE %s\n", |
data[0]&1?"enabled":"disabled"); |
return 1; |
case 0x01: |
kgem_debug_print(data, offset, 0, "3DSTATE_MAP_COORD_SET_I830\n"); |
return 1; |
case 0x0a: |
kgem_debug_print(data, offset, 0, "3DSTATE_MAP_CUBE_I830\n"); |
return 1; |
case 0x05: |
kgem_debug_print(data, offset, 0, "3DSTATE_MAP_TEX_STREAM_I830\n"); |
return 1; |
} |
kgem_debug_print(data, offset, 0, "3D UNKNOWN: 3d_1c opcode = 0x%x\n", |
opcode); |
assert(0); |
return 1; |
} |
/** Sets the string dstname to describe the destination of the PS instruction */ |
static void |
gen3_get_instruction_dst(uint32_t *data, int i, char *dstname, int do_mask) |
{ |
uint32_t a0 = data[i]; |
int dst_nr = (a0 >> 14) & 0xf; |
char dstmask[8]; |
const char *sat; |
if (do_mask) { |
if (((a0 >> 10) & 0xf) == 0xf) { |
dstmask[0] = 0; |
} else { |
int dstmask_index = 0; |
dstmask[dstmask_index++] = '.'; |
if (a0 & (1 << 10)) |
dstmask[dstmask_index++] = 'x'; |
if (a0 & (1 << 11)) |
dstmask[dstmask_index++] = 'y'; |
if (a0 & (1 << 12)) |
dstmask[dstmask_index++] = 'z'; |
if (a0 & (1 << 13)) |
dstmask[dstmask_index++] = 'w'; |
dstmask[dstmask_index++] = 0; |
} |
if (a0 & (1 << 22)) |
sat = ".sat"; |
else |
sat = ""; |
} else { |
dstmask[0] = 0; |
sat = ""; |
} |
switch ((a0 >> 19) & 0x7) { |
case 0: |
assert(dst_nr <= 15); |
sprintf(dstname, "R%d%s%s", dst_nr, dstmask, sat); |
break; |
case 4: |
assert(dst_nr == 0); |
sprintf(dstname, "oC%s%s", dstmask, sat); |
break; |
case 5: |
assert(dst_nr == 0); |
sprintf(dstname, "oD%s%s", dstmask, sat); |
break; |
case 6: |
assert(dst_nr <= 3); |
sprintf(dstname, "U%d%s%s", dst_nr, dstmask, sat); |
break; |
default: |
sprintf(dstname, "RESERVED"); |
break; |
} |
} |
static const char * |
gen3_get_channel_swizzle(uint32_t select) |
{ |
switch (select & 0x7) { |
case 0: |
return (select & 8) ? "-x" : "x"; |
case 1: |
return (select & 8) ? "-y" : "y"; |
case 2: |
return (select & 8) ? "-z" : "z"; |
case 3: |
return (select & 8) ? "-w" : "w"; |
case 4: |
return (select & 8) ? "-0" : "0"; |
case 5: |
return (select & 8) ? "-1" : "1"; |
default: |
return (select & 8) ? "-bad" : "bad"; |
} |
} |
static void |
gen3_get_instruction_src_name(uint32_t src_type, uint32_t src_nr, char *name) |
{ |
switch (src_type) { |
case 0: |
sprintf(name, "R%d", src_nr); |
assert(src_nr <= 15); |
break; |
case 1: |
if (src_nr < 8) |
sprintf(name, "T%d", src_nr); |
else if (src_nr == 8) |
sprintf(name, "DIFFUSE"); |
else if (src_nr == 9) |
sprintf(name, "SPECULAR"); |
else if (src_nr == 10) |
sprintf(name, "FOG"); |
else { |
assert(0); |
sprintf(name, "RESERVED"); |
} |
break; |
case 2: |
sprintf(name, "C%d", src_nr); |
assert(src_nr <= 31); |
break; |
case 4: |
sprintf(name, "oC"); |
assert(src_nr == 0); |
break; |
case 5: |
sprintf(name, "oD"); |
assert(src_nr == 0); |
break; |
case 6: |
sprintf(name, "U%d", src_nr); |
assert(src_nr <= 3); |
break; |
default: |
sprintf(name, "RESERVED"); |
assert(0); |
break; |
} |
} |
static void |
gen3_get_instruction_src0(uint32_t *data, int i, char *srcname) |
{ |
uint32_t a0 = data[i]; |
uint32_t a1 = data[i + 1]; |
int src_nr = (a0 >> 2) & 0x1f; |
const char *swizzle_x = gen3_get_channel_swizzle((a1 >> 28) & 0xf); |
const char *swizzle_y = gen3_get_channel_swizzle((a1 >> 24) & 0xf); |
const char *swizzle_z = gen3_get_channel_swizzle((a1 >> 20) & 0xf); |
const char *swizzle_w = gen3_get_channel_swizzle((a1 >> 16) & 0xf); |
char swizzle[100]; |
gen3_get_instruction_src_name((a0 >> 7) & 0x7, src_nr, srcname); |
sprintf(swizzle, ".%s%s%s%s", swizzle_x, swizzle_y, swizzle_z, swizzle_w); |
if (strcmp(swizzle, ".xyzw") != 0) |
strcat(srcname, swizzle); |
} |
static void |
gen3_get_instruction_src1(uint32_t *data, int i, char *srcname) |
{ |
uint32_t a1 = data[i + 1]; |
uint32_t a2 = data[i + 2]; |
int src_nr = (a1 >> 8) & 0x1f; |
const char *swizzle_x = gen3_get_channel_swizzle((a1 >> 4) & 0xf); |
const char *swizzle_y = gen3_get_channel_swizzle((a1 >> 0) & 0xf); |
const char *swizzle_z = gen3_get_channel_swizzle((a2 >> 28) & 0xf); |
const char *swizzle_w = gen3_get_channel_swizzle((a2 >> 24) & 0xf); |
char swizzle[100]; |
gen3_get_instruction_src_name((a1 >> 13) & 0x7, src_nr, srcname); |
sprintf(swizzle, ".%s%s%s%s", swizzle_x, swizzle_y, swizzle_z, swizzle_w); |
if (strcmp(swizzle, ".xyzw") != 0) |
strcat(srcname, swizzle); |
} |
static void |
gen3_get_instruction_src2(uint32_t *data, int i, char *srcname) |
{ |
uint32_t a2 = data[i + 2]; |
int src_nr = (a2 >> 16) & 0x1f; |
const char *swizzle_x = gen3_get_channel_swizzle((a2 >> 12) & 0xf); |
const char *swizzle_y = gen3_get_channel_swizzle((a2 >> 8) & 0xf); |
const char *swizzle_z = gen3_get_channel_swizzle((a2 >> 4) & 0xf); |
const char *swizzle_w = gen3_get_channel_swizzle((a2 >> 0) & 0xf); |
char swizzle[100]; |
gen3_get_instruction_src_name((a2 >> 21) & 0x7, src_nr, srcname); |
sprintf(swizzle, ".%s%s%s%s", swizzle_x, swizzle_y, swizzle_z, swizzle_w); |
if (strcmp(swizzle, ".xyzw") != 0) |
strcat(srcname, swizzle); |
} |
static void |
gen3_get_instruction_addr(uint32_t src_type, uint32_t src_nr, char *name) |
{ |
switch (src_type) { |
case 0: |
sprintf(name, "R%d", src_nr); |
assert(src_nr <= 15); |
break; |
case 1: |
if (src_nr < 8) |
sprintf(name, "T%d", src_nr); |
else if (src_nr == 8) |
sprintf(name, "DIFFUSE"); |
else if (src_nr == 9) |
sprintf(name, "SPECULAR"); |
else if (src_nr == 10) |
sprintf(name, "FOG"); |
else { |
assert(0); |
sprintf(name, "RESERVED"); |
} |
break; |
case 4: |
sprintf(name, "oC"); |
assert(src_nr == 0); |
break; |
case 5: |
sprintf(name, "oD"); |
assert(src_nr == 0); |
break; |
default: |
assert(0); |
sprintf(name, "RESERVED"); |
break; |
} |
} |
static void |
gen3_decode_alu1(uint32_t *data, uint32_t offset, |
int i, char *instr_prefix, const char *op_name) |
{ |
char dst[100], src0[100]; |
gen3_get_instruction_dst(data, i, dst, 1); |
gen3_get_instruction_src0(data, i, src0); |
kgem_debug_print(data, offset, i++, "%s: %s %s, %s\n", instr_prefix, |
op_name, dst, src0); |
kgem_debug_print(data, offset, i++, "%s\n", instr_prefix); |
kgem_debug_print(data, offset, i++, "%s\n", instr_prefix); |
} |
static void |
gen3_decode_alu2(uint32_t *data, uint32_t offset, |
int i, char *instr_prefix, const char *op_name) |
{ |
char dst[100], src0[100], src1[100]; |
gen3_get_instruction_dst(data, i, dst, 1); |
gen3_get_instruction_src0(data, i, src0); |
gen3_get_instruction_src1(data, i, src1); |
kgem_debug_print(data, offset, i++, "%s: %s %s, %s, %s\n", instr_prefix, |
op_name, dst, src0, src1); |
kgem_debug_print(data, offset, i++, "%s\n", instr_prefix); |
kgem_debug_print(data, offset, i++, "%s\n", instr_prefix); |
} |
static void |
gen3_decode_alu3(uint32_t *data, uint32_t offset, |
int i, char *instr_prefix, const char *op_name) |
{ |
char dst[100], src0[100], src1[100], src2[100]; |
gen3_get_instruction_dst(data, i, dst, 1); |
gen3_get_instruction_src0(data, i, src0); |
gen3_get_instruction_src1(data, i, src1); |
gen3_get_instruction_src2(data, i, src2); |
kgem_debug_print(data, offset, i++, "%s: %s %s, %s, %s, %s\n", instr_prefix, |
op_name, dst, src0, src1, src2); |
kgem_debug_print(data, offset, i++, "%s\n", instr_prefix); |
kgem_debug_print(data, offset, i++, "%s\n", instr_prefix); |
} |
static void |
gen3_decode_tex(uint32_t *data, uint32_t offset, int i, char *instr_prefix, |
const char *tex_name) |
{ |
uint32_t t0 = data[i]; |
uint32_t t1 = data[i + 1]; |
char dst_name[100]; |
char addr_name[100]; |
int sampler_nr; |
gen3_get_instruction_dst(data, i, dst_name, 0); |
gen3_get_instruction_addr((t1 >> 24) & 0x7, |
(t1 >> 17) & 0xf, |
addr_name); |
sampler_nr = t0 & 0xf; |
kgem_debug_print(data, offset, i++, "%s: %s %s, S%d, %s\n", instr_prefix, |
tex_name, dst_name, sampler_nr, addr_name); |
kgem_debug_print(data, offset, i++, "%s\n", instr_prefix); |
kgem_debug_print(data, offset, i++, "%s\n", instr_prefix); |
} |
static void |
gen3_decode_dcl(uint32_t *data, uint32_t offset, int i, char *instr_prefix) |
{ |
uint32_t d0 = data[i]; |
const char *sampletype; |
int dcl_nr = (d0 >> 14) & 0xf; |
const char *dcl_x = d0 & (1 << 10) ? "x" : ""; |
const char *dcl_y = d0 & (1 << 11) ? "y" : ""; |
const char *dcl_z = d0 & (1 << 12) ? "z" : ""; |
const char *dcl_w = d0 & (1 << 13) ? "w" : ""; |
char dcl_mask[10]; |
switch ((d0 >> 19) & 0x3) { |
case 1: |
sprintf(dcl_mask, ".%s%s%s%s", dcl_x, dcl_y, dcl_z, dcl_w); |
assert (strcmp(dcl_mask, ".")); |
assert(dcl_nr <= 10); |
if (dcl_nr < 8) { |
if (strcmp(dcl_mask, ".x") != 0 && |
strcmp(dcl_mask, ".xy") != 0 && |
strcmp(dcl_mask, ".xz") != 0 && |
strcmp(dcl_mask, ".w") != 0 && |
strcmp(dcl_mask, ".xyzw") != 0) { |
assert(0); |
} |
kgem_debug_print(data, offset, i++, "%s: DCL T%d%s\n", instr_prefix, |
dcl_nr, dcl_mask); |
} else { |
if (strcmp(dcl_mask, ".xz") == 0) |
assert(0); |
else if (strcmp(dcl_mask, ".xw") == 0) |
assert(0); |
else if (strcmp(dcl_mask, ".xzw") == 0) |
assert(0); |
if (dcl_nr == 8) { |
kgem_debug_print(data, offset, i++, "%s: DCL DIFFUSE%s\n", instr_prefix, |
dcl_mask); |
} else if (dcl_nr == 9) { |
kgem_debug_print(data, offset, i++, "%s: DCL SPECULAR%s\n", instr_prefix, |
dcl_mask); |
} else if (dcl_nr == 10) { |
kgem_debug_print(data, offset, i++, "%s: DCL FOG%s\n", instr_prefix, |
dcl_mask); |
} |
} |
kgem_debug_print(data, offset, i++, "%s\n", instr_prefix); |
kgem_debug_print(data, offset, i++, "%s\n", instr_prefix); |
break; |
case 3: |
switch ((d0 >> 22) & 0x3) { |
case 0: |
sampletype = "2D"; |
break; |
case 1: |
sampletype = "CUBE"; |
break; |
case 2: |
sampletype = "3D"; |
break; |
default: |
sampletype = "RESERVED"; |
break; |
} |
assert(dcl_nr <= 15); |
kgem_debug_print(data, offset, i++, "%s: DCL S%d %s\n", instr_prefix, |
dcl_nr, sampletype); |
kgem_debug_print(data, offset, i++, "%s\n", instr_prefix); |
kgem_debug_print(data, offset, i++, "%s\n", instr_prefix); |
break; |
default: |
kgem_debug_print(data, offset, i++, "%s: DCL RESERVED%d\n", instr_prefix, dcl_nr); |
kgem_debug_print(data, offset, i++, "%s\n", instr_prefix); |
kgem_debug_print(data, offset, i++, "%s\n", instr_prefix); |
} |
} |
static void |
gen3_decode_instruction(uint32_t *data, uint32_t offset, |
int i, char *instr_prefix) |
{ |
switch ((data[i] >> 24) & 0x1f) { |
case 0x0: |
kgem_debug_print(data, offset, i++, "%s: NOP\n", instr_prefix); |
kgem_debug_print(data, offset, i++, "%s\n", instr_prefix); |
kgem_debug_print(data, offset, i++, "%s\n", instr_prefix); |
break; |
case 0x01: |
gen3_decode_alu2(data, offset, i, instr_prefix, "ADD"); |
break; |
case 0x02: |
gen3_decode_alu1(data, offset, i, instr_prefix, "MOV"); |
break; |
case 0x03: |
gen3_decode_alu2(data, offset, i, instr_prefix, "MUL"); |
break; |
case 0x04: |
gen3_decode_alu3(data, offset, i, instr_prefix, "MAD"); |
break; |
case 0x05: |
gen3_decode_alu3(data, offset, i, instr_prefix, "DP2ADD"); |
break; |
case 0x06: |
gen3_decode_alu2(data, offset, i, instr_prefix, "DP3"); |
break; |
case 0x07: |
gen3_decode_alu2(data, offset, i, instr_prefix, "DP4"); |
break; |
case 0x08: |
gen3_decode_alu1(data, offset, i, instr_prefix, "FRC"); |
break; |
case 0x09: |
gen3_decode_alu1(data, offset, i, instr_prefix, "RCP"); |
break; |
case 0x0a: |
gen3_decode_alu1(data, offset, i, instr_prefix, "RSQ"); |
break; |
case 0x0b: |
gen3_decode_alu1(data, offset, i, instr_prefix, "EXP"); |
break; |
case 0x0c: |
gen3_decode_alu1(data, offset, i, instr_prefix, "LOG"); |
break; |
case 0x0d: |
gen3_decode_alu2(data, offset, i, instr_prefix, "CMP"); |
break; |
case 0x0e: |
gen3_decode_alu2(data, offset, i, instr_prefix, "MIN"); |
break; |
case 0x0f: |
gen3_decode_alu2(data, offset, i, instr_prefix, "MAX"); |
break; |
case 0x10: |
gen3_decode_alu1(data, offset, i, instr_prefix, "FLR"); |
break; |
case 0x11: |
gen3_decode_alu1(data, offset, i, instr_prefix, "MOD"); |
break; |
case 0x12: |
gen3_decode_alu1(data, offset, i, instr_prefix, "TRC"); |
break; |
case 0x13: |
gen3_decode_alu2(data, offset, i, instr_prefix, "SGE"); |
break; |
case 0x14: |
gen3_decode_alu2(data, offset, i, instr_prefix, "SLT"); |
break; |
case 0x15: |
gen3_decode_tex(data, offset, i, instr_prefix, "TEXLD"); |
break; |
case 0x16: |
gen3_decode_tex(data, offset, i, instr_prefix, "TEXLDP"); |
break; |
case 0x17: |
gen3_decode_tex(data, offset, i, instr_prefix, "TEXLDB"); |
break; |
case 0x19: |
gen3_decode_dcl(data, offset, i, instr_prefix); |
break; |
default: |
kgem_debug_print(data, offset, i++, "%s: unknown\n", instr_prefix); |
kgem_debug_print(data, offset, i++, "%s\n", instr_prefix); |
kgem_debug_print(data, offset, i++, "%s\n", instr_prefix); |
break; |
} |
} |
static const char * |
gen3_decode_compare_func(uint32_t op) |
{ |
switch (op&0x7) { |
case 0: return "always"; |
case 1: return "never"; |
case 2: return "less"; |
case 3: return "equal"; |
case 4: return "lequal"; |
case 5: return "greater"; |
case 6: return "notequal"; |
case 7: return "gequal"; |
} |
return ""; |
} |
static const char * |
gen3_decode_stencil_op(uint32_t op) |
{ |
switch (op&0x7) { |
case 0: return "keep"; |
case 1: return "zero"; |
case 2: return "replace"; |
case 3: return "incr_sat"; |
case 4: return "decr_sat"; |
case 5: return "greater"; |
case 6: return "incr"; |
case 7: return "decr"; |
} |
return ""; |
} |
#if 0 |
/* part of MODES_4 */ |
static const char * |
gen3_decode_logic_op(uint32_t op) |
{ |
switch (op&0xf) { |
case 0: return "clear"; |
case 1: return "nor"; |
case 2: return "and_inv"; |
case 3: return "copy_inv"; |
case 4: return "and_rvrse"; |
case 5: return "inv"; |
case 6: return "xor"; |
case 7: return "nand"; |
case 8: return "and"; |
case 9: return "equiv"; |
case 10: return "noop"; |
case 11: return "or_inv"; |
case 12: return "copy"; |
case 13: return "or_rvrse"; |
case 14: return "or"; |
case 15: return "set"; |
} |
return ""; |
} |
#endif |
static const char * |
gen3_decode_blend_fact(uint32_t op) |
{ |
switch (op&0xf) { |
case 1: return "zero"; |
case 2: return "one"; |
case 3: return "src_colr"; |
case 4: return "inv_src_colr"; |
case 5: return "src_alpha"; |
case 6: return "inv_src_alpha"; |
case 7: return "dst_alpha"; |
case 8: return "inv_dst_alpha"; |
case 9: return "dst_colr"; |
case 10: return "inv_dst_colr"; |
case 11: return "src_alpha_sat"; |
case 12: return "cnst_colr"; |
case 13: return "inv_cnst_colr"; |
case 14: return "cnst_alpha"; |
case 15: return "inv_const_alpha"; |
} |
return ""; |
} |
static const char * |
decode_tex_coord_mode(uint32_t mode) |
{ |
switch (mode&0x7) { |
case 0: return "wrap"; |
case 1: return "mirror"; |
case 2: return "clamp_edge"; |
case 3: return "cube"; |
case 4: return "clamp_border"; |
case 5: return "mirror_once"; |
} |
return ""; |
} |
static const char * |
gen3_decode_sample_filter(uint32_t mode) |
{ |
switch (mode&0x7) { |
case 0: return "nearest"; |
case 1: return "linear"; |
case 2: return "anisotropic"; |
case 3: return "4x4_1"; |
case 4: return "4x4_2"; |
case 5: return "4x4_flat"; |
case 6: return "6x5_mono"; |
} |
return ""; |
} |
static int |
gen3_decode_load_state_immediate_1(struct kgem *kgem, uint32_t offset) |
{ |
const uint32_t *data = kgem->batch + offset; |
int len, i, word; |
kgem_debug_print(data, offset, 0, "3DSTATE_LOAD_STATE_IMMEDIATE_1\n"); |
len = (data[0] & 0x0000000f) + 2; |
i = 1; |
for (word = 0; word <= 8; word++) { |
if (data[0] & (1 << (4 + word))) { |
switch (word) { |
case 0: |
kgem_debug_print(data, offset, i, "S0: vbo offset: 0x%08x%s\n", |
data[i]&(~1),data[i]&1?", auto cache invalidate disabled":""); |
gen3_update_vertex_buffer_addr(kgem, offset + i); |
break; |
case 1: |
kgem_debug_print(data, offset, i, "S1: vertex width: %i, vertex pitch: %i\n", |
(data[i]>>24)&0x3f,(data[i]>>16)&0x3f); |
gen3_update_vertex_buffer_pitch(kgem, offset + i); |
break; |
case 2: |
{ |
char buf[200]; |
int len = 0; |
int tex_num; |
for (tex_num = 0; tex_num < 8; tex_num++) { |
switch((data[i]>>tex_num*4)&0xf) { |
case 0: len += sprintf(buf + len, "%i=2D ", tex_num); break; |
case 1: len += sprintf(buf + len, "%i=3D ", tex_num); break; |
case 2: len += sprintf(buf + len, "%i=4D ", tex_num); break; |
case 3: len += sprintf(buf + len, "%i=1D ", tex_num); break; |
case 4: len += sprintf(buf + len, "%i=2D_16 ", tex_num); break; |
case 5: len += sprintf(buf + len, "%i=4D_16 ", tex_num); break; |
case 0xf: len += sprintf(buf + len, "%i=NP ", tex_num); break; |
} |
} |
kgem_debug_print(data, offset, i, "S2: texcoord formats: %s\n", buf); |
gen3_update_vertex_texcoords(kgem, data[i]); |
} |
break; |
case 3: |
kgem_debug_print(data, offset, i, "S3: not documented\n"); |
break; |
case 4: |
{ |
const char *cullmode = ""; |
const char *vfmt_xyzw = ""; |
switch((data[i]>>13)&0x3) { |
case 0: cullmode = "both"; break; |
case 1: cullmode = "none"; break; |
case 2: cullmode = "cw"; break; |
case 3: cullmode = "ccw"; break; |
} |
switch(data[i] & (7<<6 | 1<<2)) { |
case 1<<6: vfmt_xyzw = "XYZ,"; break; |
case 2<<6: vfmt_xyzw = "XYZW,"; break; |
case 3<<6: vfmt_xyzw = "XY,"; break; |
case 4<<6: vfmt_xyzw = "XYW,"; break; |
case 1<<6 | 1<<2: vfmt_xyzw = "XYZF,"; break; |
case 2<<6 | 1<<2: vfmt_xyzw = "XYZWF,"; break; |
case 3<<6 | 1<<2: vfmt_xyzw = "XYF,"; break; |
case 4<<6 | 1<<2: vfmt_xyzw = "XYWF,"; break; |
} |
kgem_debug_print(data, offset, i, "S4: point_width=%i, line_width=%.1f," |
"%s%s%s%s%s cullmode=%s, vfmt=%s%s%s%s%s%s%s%s " |
"%s%s%s\n", |
(data[i]>>23)&0x1ff, |
((data[i]>>19)&0xf) / 2.0, |
data[i]&(0xf<<15)?" flatshade=":"", |
data[i]&(1<<18)?"Alpha,":"", |
data[i]&(1<<17)?"Fog,":"", |
data[i]&(1<<16)?"Specular,":"", |
data[i]&(1<<15)?"Color,":"", |
cullmode, |
data[i]&(1<<12)?"PointWidth,":"", |
data[i]&(1<<11)?"SpecFog,":"", |
data[i]&(1<<10)?"Color,":"", |
data[i]&(1<<9)?"DepthOfs,":"", |
vfmt_xyzw, |
data[i]&(1<<9)?"FogParam,":"", |
data[i]&(1<<5)?"force default diffuse, ":"", |
data[i]&(1<<4)?"force default specular, ":"", |
data[i]&(1<<3)?"local depth ofs enable, ":"", |
data[i]&(1<<1)?"point sprite enable, ":"", |
data[i]&(1<<0)?"line AA enable, ":""); |
gen3_update_vertex_elements(kgem, data[i]); |
break; |
} |
case 5: |
{ |
kgem_debug_print(data, offset, i, "S5:%s%s%s%s%s" |
"%s%s%s%s stencil_ref=0x%x, stencil_test=%s, " |
"stencil_fail=%s, stencil_pass_z_fail=%s, " |
"stencil_pass_z_pass=%s, %s%s%s%s\n", |
data[i]&(0xf<<28)?" write_disable=":"", |
data[i]&(1<<31)?"Alpha,":"", |
data[i]&(1<<30)?"Red,":"", |
data[i]&(1<<29)?"Green,":"", |
data[i]&(1<<28)?"Blue,":"", |
data[i]&(1<<27)?" force default point size,":"", |
data[i]&(1<<26)?" last pixel enable,":"", |
data[i]&(1<<25)?" global depth ofs enable,":"", |
data[i]&(1<<24)?" fog enable,":"", |
(data[i]>>16)&0xff, |
gen3_decode_compare_func(data[i]>>13), |
gen3_decode_stencil_op(data[i]>>10), |
gen3_decode_stencil_op(data[i]>>7), |
gen3_decode_stencil_op(data[i]>>4), |
data[i]&(1<<3)?"stencil write enable, ":"", |
data[i]&(1<<2)?"stencil test enable, ":"", |
data[i]&(1<<1)?"color dither enable, ":"", |
data[i]&(1<<0)?"logicop enable, ":""); |
} |
break; |
case 6: |
kgem_debug_print(data, offset, i, "S6: %salpha_test=%s, alpha_ref=0x%x, " |
"depth_test=%s, %ssrc_blnd_fct=%s, dst_blnd_fct=%s, " |
"%s%stristrip_provoking_vertex=%i\n", |
data[i]&(1<<31)?"alpha test enable, ":"", |
gen3_decode_compare_func(data[i]>>28), |
data[i]&(0xff<<20), |
gen3_decode_compare_func(data[i]>>16), |
data[i]&(1<<15)?"cbuf blend enable, ":"", |
gen3_decode_blend_fact(data[i]>>8), |
gen3_decode_blend_fact(data[i]>>4), |
data[i]&(1<<3)?"depth write enable, ":"", |
data[i]&(1<<2)?"cbuf write enable, ":"", |
data[i]&(0x3)); |
break; |
case 7: |
kgem_debug_print(data, offset, i, "S7: depth offset constant: 0x%08x\n", data[i]); |
break; |
} |
i++; |
} |
} |
assert(len == i); |
return len; |
} |
static int |
gen3_decode_3d_1d(struct kgem *kgem, uint32_t offset) |
{ |
uint32_t *data = kgem->batch + offset; |
unsigned int len, i, c, idx, word, map, sampler, instr; |
const char *format, *zformat, *type; |
uint32_t opcode; |
static const struct { |
uint32_t opcode; |
int min_len; |
int max_len; |
const char *name; |
} opcodes_3d_1d[] = { |
{ 0x86, 4, 4, "3DSTATE_CHROMA_KEY" }, |
{ 0x88, 2, 2, "3DSTATE_CONSTANT_BLEND_COLOR" }, |
{ 0x99, 2, 2, "3DSTATE_DEFAULT_DIFFUSE" }, |
{ 0x9a, 2, 2, "3DSTATE_DEFAULT_SPECULAR" }, |
{ 0x98, 2, 2, "3DSTATE_DEFAULT_Z" }, |
{ 0x97, 2, 2, "3DSTATE_DEPTH_OFFSET_SCALE" }, |
{ 0x9d, 65, 65, "3DSTATE_FILTER_COEFFICIENTS_4X4" }, |
{ 0x9e, 4, 4, "3DSTATE_MONO_FILTER" }, |
{ 0x89, 4, 4, "3DSTATE_FOG_MODE" }, |
{ 0x8f, 2, 16, "3DSTATE_MAP_PALLETE_LOAD_32" }, |
{ 0x83, 2, 2, "3DSTATE_SPAN_STIPPLE" }, |
}, *opcode_3d_1d; |
opcode = (data[0] & 0x00ff0000) >> 16; |
switch (opcode) { |
case 0x07: |
/* This instruction is unusual. A 0 length means just 1 DWORD instead of |
* 2. The 0 length is specified in one place to be unsupported, but |
* stated to be required in another, and 0 length LOAD_INDIRECTs appear |
* to cause no harm at least. |
*/ |
kgem_debug_print(data, offset, 0, "3DSTATE_LOAD_INDIRECT\n"); |
len = (data[0] & 0x000000ff) + 1; |
i = 1; |
if (data[0] & (0x01 << 8)) { |
kgem_debug_print(data, offset, i++, "SIS.0\n"); |
kgem_debug_print(data, offset, i++, "SIS.1\n"); |
} |
if (data[0] & (0x02 << 8)) { |
kgem_debug_print(data, offset, i++, "DIS.0\n"); |
} |
if (data[0] & (0x04 << 8)) { |
kgem_debug_print(data, offset, i++, "SSB.0\n"); |
kgem_debug_print(data, offset, i++, "SSB.1\n"); |
} |
if (data[0] & (0x08 << 8)) { |
kgem_debug_print(data, offset, i++, "MSB.0\n"); |
kgem_debug_print(data, offset, i++, "MSB.1\n"); |
} |
if (data[0] & (0x10 << 8)) { |
kgem_debug_print(data, offset, i++, "PSP.0\n"); |
kgem_debug_print(data, offset, i++, "PSP.1\n"); |
} |
if (data[0] & (0x20 << 8)) { |
kgem_debug_print(data, offset, i++, "PSC.0\n"); |
kgem_debug_print(data, offset, i++, "PSC.1\n"); |
} |
assert(len == i); |
return len; |
case 0x04: |
return gen3_decode_load_state_immediate_1(kgem, offset); |
case 0x03: |
kgem_debug_print(data, offset, 0, "3DSTATE_LOAD_STATE_IMMEDIATE_2\n"); |
len = (data[0] & 0x0000000f) + 2; |
i = 1; |
for (word = 6; word <= 14; word++) { |
if (data[0] & (1 << word)) { |
if (word == 6) |
kgem_debug_print(data, offset, i++, "TBCF\n"); |
else if (word >= 7 && word <= 10) { |
kgem_debug_print(data, offset, i++, "TB%dC\n", word - 7); |
kgem_debug_print(data, offset, i++, "TB%dA\n", word - 7); |
} else if (word >= 11 && word <= 14) { |
kgem_debug_print(data, offset, i, "TM%dS0: offset=0x%08x, %s\n", |
word - 11, |
data[i]&0xfffffffe, |
data[i]&1?"use fence":""); |
i++; |
kgem_debug_print(data, offset, i, "TM%dS1: height=%i, width=%i, %s\n", |
word - 11, |
data[i]>>21, (data[i]>>10)&0x3ff, |
data[i]&2?(data[i]&1?"y-tiled":"x-tiled"):""); |
i++; |
kgem_debug_print(data, offset, i, "TM%dS2: pitch=%i, \n", |
word - 11, |
((data[i]>>21) + 1)*4); |
i++; |
kgem_debug_print(data, offset, i++, "TM%dS3\n", word - 11); |
kgem_debug_print(data, offset, i++, "TM%dS4: dflt color\n", word - 11); |
} |
} |
} |
assert(len == i); |
return len; |
case 0x00: |
kgem_debug_print(data, offset, 0, "3DSTATE_MAP_STATE\n"); |
len = (data[0] & 0x0000003f) + 2; |
kgem_debug_print(data, offset, 1, "mask\n"); |
i = 2; |
for (map = 0; map <= 15; map++) { |
if (data[1] & (1 << map)) { |
int width, height, pitch, dword; |
struct drm_i915_gem_relocation_entry *reloc; |
const char *tiling; |
reloc = kgem_debug_get_reloc_entry(kgem, &data[i] - kgem->batch); |
assert(reloc->target_handle); |
dword = data[i]; |
kgem_debug_print(data, offset, i++, "map %d MS2 %s%s%s, handle=%d\n", map, |
dword&(1<<31)?"untrusted surface, ":"", |
dword&(1<<1)?"vertical line stride enable, ":"", |
dword&(1<<0)?"vertical ofs enable, ":"", |
reloc->target_handle); |
dword = data[i]; |
width = ((dword >> 10) & ((1 << 11) - 1))+1; |
height = ((dword >> 21) & ((1 << 11) - 1))+1; |
tiling = "none"; |
if (dword & (1 << 2)) |
tiling = "fenced"; |
else if (dword & (1 << 1)) |
tiling = dword & (1 << 0) ? "Y" : "X"; |
type = " BAD"; |
format = " (invalid)"; |
switch ((dword>>7) & 0x7) { |
case 1: |
type = "8"; |
switch ((dword>>3) & 0xf) { |
case 0: format = "I"; break; |
case 1: format = "L"; break; |
case 4: format = "A"; break; |
case 5: format = " mono"; break; |
} |
break; |
case 2: |
type = "16"; |
switch ((dword>>3) & 0xf) { |
case 0: format = " rgb565"; break; |
case 1: format = " argb1555"; break; |
case 2: format = " argb4444"; break; |
case 3: format = " ay88"; break; |
case 5: format = " 88dvdu"; break; |
case 6: format = " bump655"; break; |
case 7: format = "I"; break; |
case 8: format = "L"; break; |
case 9: format = "A"; break; |
} |
break; |
case 3: |
type = "32"; |
switch ((dword>>3) & 0xf) { |
case 0: format = " argb8888"; break; |
case 1: format = " abgr8888"; break; |
case 2: format = " xrgb8888"; break; |
case 3: format = " xbgr8888"; break; |
case 4: format = " qwvu8888"; break; |
case 5: format = " axvu8888"; break; |
case 6: format = " lxvu8888"; break; |
case 7: format = " xlvu8888"; break; |
case 8: format = " argb2101010"; break; |
case 9: format = " abgr2101010"; break; |
case 10: format = " awvu2101010"; break; |
case 11: format = " gr1616"; break; |
case 12: format = " vu1616"; break; |
case 13: format = " xI824"; break; |
case 14: format = " xA824"; break; |
case 15: format = " xL824"; break; |
} |
break; |
case 5: |
type = "422"; |
switch ((dword>>3) & 0xf) { |
case 0: format = " yuv_swapy"; break; |
case 1: format = " yuv"; break; |
case 2: format = " yuv_swapuv"; break; |
case 3: format = " yuv_swapuvy"; break; |
} |
break; |
case 6: |
type = "compressed"; |
switch ((dword>>3) & 0x7) { |
case 0: format = " dxt1"; break; |
case 1: format = " dxt2_3"; break; |
case 2: format = " dxt4_5"; break; |
case 3: format = " fxt1"; break; |
case 4: format = " dxt1_rb"; break; |
} |
break; |
case 7: |
type = "4b indexed"; |
switch ((dword>>3) & 0xf) { |
case 7: format = " argb8888"; break; |
} |
break; |
default: |
format = "BAD"; |
break; |
} |
dword = data[i]; |
kgem_debug_print(data, offset, i++, "map %d MS3 [width=%d, height=%d, format=%s%s, tiling=%s%s]\n", |
map, width, height, type, format, tiling, |
dword&(1<<9)?" palette select":""); |
dword = data[i]; |
pitch = 4*(((dword >> 21) & ((1 << 11) - 1))+1); |
kgem_debug_print(data, offset, i++, "map %d MS4 [pitch=%d, max_lod=%i, vol_depth=%i, cube_face_ena=%x, %s]\n", |
map, pitch, |
(dword>>9)&0x3f, dword&0xff, (dword>>15)&0x3f, |
dword&(1<<8)?"miplayout legacy":"miplayout right"); |
} |
} |
assert(len == i); |
return len; |
case 0x06: |
kgem_debug_print(data, offset, 0, "3DSTATE_PIXEL_SHADER_CONSTANTS\n"); |
len = (data[0] & 0x000000ff) + 2; |
i = 2; |
for (c = 0; c <= 31; c++) { |
if (data[1] & (1 << c)) { |
kgem_debug_print(data, offset, i, "C%d.X = %f\n", |
c, int_as_float(data[i])); |
i++; |
kgem_debug_print(data, offset, i, "C%d.Y = %f\n", |
c, int_as_float(data[i])); |
i++; |
kgem_debug_print(data, offset, i, "C%d.Z = %f\n", |
c, int_as_float(data[i])); |
i++; |
kgem_debug_print(data, offset, i, "C%d.W = %f\n", |
c, int_as_float(data[i])); |
i++; |
} |
} |
assert(len == i); |
return len; |
case 0x05: |
kgem_debug_print(data, offset, 0, "3DSTATE_PIXEL_SHADER_PROGRAM\n"); |
len = (data[0] & 0x000000ff) + 2; |
assert(((len-1) % 3) == 0); |
assert(len <= 370); |
i = 1; |
for (instr = 0; instr < (len - 1) / 3; instr++) { |
char instr_prefix[10]; |
sprintf(instr_prefix, "PS%03d", instr); |
gen3_decode_instruction(data, offset, i, instr_prefix); |
i += 3; |
} |
return len; |
case 0x01: |
kgem_debug_print(data, offset, 0, "3DSTATE_SAMPLER_STATE\n"); |
kgem_debug_print(data, offset, 1, "mask\n"); |
len = (data[0] & 0x0000003f) + 2; |
i = 2; |
for (sampler = 0; sampler <= 15; sampler++) { |
if (data[1] & (1 << sampler)) { |
uint32_t dword; |
const char *mip_filter = ""; |
dword = data[i]; |
switch ((dword>>20)&0x3) { |
case 0: mip_filter = "none"; break; |
case 1: mip_filter = "nearest"; break; |
case 3: mip_filter = "linear"; break; |
} |
kgem_debug_print(data, offset, i++, "sampler %d SS2:%s%s%s " |
"base_mip_level=%i, mip_filter=%s, mag_filter=%s, min_filter=%s " |
"lod_bias=%.2f,%s max_aniso=%i, shadow_func=%s\n", sampler, |
dword&(1<<31)?" reverse gamma,":"", |
dword&(1<<30)?" packed2planar,":"", |
dword&(1<<29)?" colorspace conversion,":"", |
(dword>>22)&0x1f, |
mip_filter, |
gen3_decode_sample_filter(dword>>17), |
gen3_decode_sample_filter(dword>>14), |
((dword>>5)&0x1ff)/(0x10*1.0), |
dword&(1<<4)?" shadow,":"", |
dword&(1<<3)?4:2, |
gen3_decode_compare_func(dword)); |
dword = data[i]; |
kgem_debug_print(data, offset, i++, "sampler %d SS3: min_lod=%.2f,%s " |
"tcmode_x=%s, tcmode_y=%s, tcmode_z=%s,%s texmap_idx=%i,%s\n", |
sampler, ((dword>>24)&0xff)/(0x10*1.0), |
dword&(1<<17)?" kill pixel enable,":"", |
decode_tex_coord_mode(dword>>12), |
decode_tex_coord_mode(dword>>9), |
decode_tex_coord_mode(dword>>6), |
dword&(1<<5)?" normalized coords,":"", |
(dword>>1)&0xf, |
dword&(1<<0)?" deinterlacer,":""); |
kgem_debug_print(data, offset, i++, "sampler %d SS4: border color\n", |
sampler); |
} |
} |
assert(len == i); |
return len; |
case 0x85: |
len = (data[0] & 0x0000000f) + 2; |
assert(len == 2); |
kgem_debug_print(data, offset, 0, |
"3DSTATE_DEST_BUFFER_VARIABLES\n"); |
switch ((data[1] >> 8) & 0xf) { |
case 0x0: format = "g8"; break; |
case 0x1: format = "x1r5g5b5"; break; |
case 0x2: format = "r5g6b5"; break; |
case 0x3: format = "a8r8g8b8"; break; |
case 0x4: format = "ycrcb_swapy"; break; |
case 0x5: format = "ycrcb_normal"; break; |
case 0x6: format = "ycrcb_swapuv"; break; |
case 0x7: format = "ycrcb_swapuvy"; break; |
case 0x8: format = "a4r4g4b4"; break; |
case 0x9: format = "a1r5g5b5"; break; |
case 0xa: format = "a2r10g10b10"; break; |
default: format = "BAD"; break; |
} |
switch ((data[1] >> 2) & 0x3) { |
case 0x0: zformat = "u16"; break; |
case 0x1: zformat = "f16"; break; |
case 0x2: zformat = "u24x8"; break; |
default: zformat = "BAD"; break; |
} |
kgem_debug_print(data, offset, 1, "%s format, %s depth format, early Z %sabled\n", |
format, zformat, |
(data[1] & (1 << 31)) ? "en" : "dis"); |
return len; |
case 0x8e: |
{ |
const char *name, *tiling; |
len = (data[0] & 0x0000000f) + 2; |
assert(len == 3); |
switch((data[1] >> 24) & 0x7) { |
case 0x3: name = "color"; break; |
case 0x7: name = "depth"; break; |
default: name = "unknown"; break; |
} |
tiling = "none"; |
if (data[1] & (1 << 23)) |
tiling = "fenced"; |
else if (data[1] & (1 << 22)) |
tiling = data[1] & (1 << 21) ? "Y" : "X"; |
kgem_debug_print(data, offset, 0, "3DSTATE_BUFFER_INFO\n"); |
kgem_debug_print(data, offset, 1, "%s, tiling = %s, pitch=%d\n", name, tiling, data[1]&0xffff); |
kgem_debug_print(data, offset, 2, "address\n"); |
return len; |
} |
case 0x81: |
len = (data[0] & 0x0000000f) + 2; |
assert(len == 3); |
kgem_debug_print(data, offset, 0, |
"3DSTATE_SCISSOR_RECTANGLE\n"); |
kgem_debug_print(data, offset, 1, "(%d,%d)\n", |
data[1] & 0xffff, data[1] >> 16); |
kgem_debug_print(data, offset, 2, "(%d,%d)\n", |
data[2] & 0xffff, data[2] >> 16); |
return len; |
case 0x80: |
len = (data[0] & 0x0000000f) + 2; |
assert(len == 5); |
kgem_debug_print(data, offset, 0, |
"3DSTATE_DRAWING_RECTANGLE\n"); |
kgem_debug_print(data, offset, 1, "%s\n", |
data[1]&(1<<30)?"depth ofs disabled ":""); |
kgem_debug_print(data, offset, 2, "(%d,%d)\n", |
data[2] & 0xffff, data[2] >> 16); |
kgem_debug_print(data, offset, 3, "(%d,%d)\n", |
data[3] & 0xffff, data[3] >> 16); |
kgem_debug_print(data, offset, 4, "(%d,%d)\n", |
(int16_t)(data[4] & 0xffff), |
(int16_t)(data[4] >> 16)); |
return len; |
case 0x9c: |
len = (data[0] & 0x0000000f) + 2; |
assert(len == 7); |
kgem_debug_print(data, offset, 0, |
"3DSTATE_CLEAR_PARAMETERS\n"); |
kgem_debug_print(data, offset, 1, "prim_type=%s, clear=%s%s%s\n", |
data[1]&(1<<16)?"CLEAR_RECT":"ZONE_INIT", |
data[1]&(1<<2)?"color,":"", |
data[1]&(1<<1)?"depth,":"", |
data[1]&(1<<0)?"stencil,":""); |
kgem_debug_print(data, offset, 2, "clear color\n"); |
kgem_debug_print(data, offset, 3, "clear depth/stencil\n"); |
kgem_debug_print(data, offset, 4, "color value (rgba8888)\n"); |
kgem_debug_print(data, offset, 5, "depth value %f\n", |
int_as_float(data[5])); |
kgem_debug_print(data, offset, 6, "clear stencil\n"); |
return len; |
} |
for (idx = 0; idx < ARRAY_SIZE(opcodes_3d_1d); idx++) { |
opcode_3d_1d = &opcodes_3d_1d[idx]; |
if (((data[0] & 0x00ff0000) >> 16) == opcode_3d_1d->opcode) { |
len = (data[0] & 0xf) + 2; |
kgem_debug_print(data, offset, 0, "%s\n", opcode_3d_1d->name); |
for (i = 1; i < len; i++) |
kgem_debug_print(data, offset, i, "dword %d\n", i); |
return len; |
} |
} |
kgem_debug_print(data, offset, 0, "3D UNKNOWN: 3d_1d opcode = 0x%x\n", opcode); |
assert(0); |
return 1; |
} |
#define VERTEX_OUT(fmt, ...) do { \ |
kgem_debug_print(data, offset, i, " V%d."fmt"\n", vertex, __VA_ARGS__); \ |
i++; \ |
} while (0) |
static int |
gen3_decode_3d_primitive(struct kgem *kgem, uint32_t offset) |
{ |
uint32_t *data = kgem->batch + offset; |
char immediate = (data[0] & (1 << 23)) == 0; |
unsigned int len, i, ret; |
const char *primtype; |
unsigned int vertex = 0; |
switch ((data[0] >> 18) & 0xf) { |
case 0x0: primtype = "TRILIST"; break; |
case 0x1: primtype = "TRISTRIP"; break; |
case 0x2: primtype = "TRISTRIP_REVERSE"; break; |
case 0x3: primtype = "TRIFAN"; break; |
case 0x4: primtype = "POLYGON"; break; |
case 0x5: primtype = "LINELIST"; break; |
case 0x6: primtype = "LINESTRIP"; break; |
case 0x7: primtype = "RECTLIST"; break; |
case 0x8: primtype = "POINTLIST"; break; |
case 0x9: primtype = "DIB"; break; |
case 0xa: primtype = "CLEAR_RECT"; assert(0); break; |
default: primtype = "unknown"; break; |
} |
gen3_update_vertex_elements_offsets(kgem); |
/* XXX: 3DPRIM_DIB not supported */ |
if (immediate) { |
len = (data[0] & 0x0003ffff) + 2; |
kgem_debug_print(data, offset, 0, "3DPRIMITIVE inline %s\n", primtype); |
for (i = 1; i < len; ) { |
ErrorF(" [%d]: ", vertex); |
i += inline_vertex_out(kgem, data + i) / sizeof(uint32_t); |
ErrorF("\n"); |
vertex++; |
} |
ret = len; |
} else { |
/* indirect vertices */ |
len = data[0] & 0x0000ffff; /* index count */ |
if (data[0] & (1 << 17)) { |
/* random vertex access */ |
kgem_debug_print(data, offset, 0, |
"3DPRIMITIVE random indirect %s (%d)\n", primtype, len); |
assert(0); |
if (len == 0) { |
/* vertex indices continue until 0xffff is found */ |
} else { |
/* fixed size vertex index buffer */ |
} |
ret = (len + 1) / 2 + 1; |
goto out; |
} else { |
/* sequential vertex access */ |
vertex = data[1] & 0xffff; |
kgem_debug_print(data, offset, 0, |
"3DPRIMITIVE sequential indirect %s, %d starting from " |
"%d\n", primtype, len, vertex); |
kgem_debug_print(data, offset, 1, " start\n"); |
for (i = 0; i < len; i++) { |
ErrorF(" [%d]: ", vertex); |
indirect_vertex_out(kgem, vertex++); |
ErrorF("\n"); |
} |
ret = 2; |
goto out; |
} |
} |
out: |
return ret; |
} |
int kgem_gen3_decode_3d(struct kgem *kgem, uint32_t offset) |
{ |
static const struct { |
uint32_t opcode; |
int min_len; |
int max_len; |
const char *name; |
} opcodes[] = { |
{ 0x06, 1, 1, "3DSTATE_ANTI_ALIASING" }, |
{ 0x08, 1, 1, "3DSTATE_BACKFACE_STENCIL_OPS" }, |
{ 0x09, 1, 1, "3DSTATE_BACKFACE_STENCIL_MASKS" }, |
{ 0x16, 1, 1, "3DSTATE_COORD_SET_BINDINGS" }, |
{ 0x15, 1, 1, "3DSTATE_FOG_COLOR" }, |
{ 0x0b, 1, 1, "3DSTATE_INDEPENDENT_ALPHA_BLEND" }, |
{ 0x0d, 1, 1, "3DSTATE_MODES_4" }, |
{ 0x0c, 1, 1, "3DSTATE_MODES_5" }, |
{ 0x07, 1, 1, "3DSTATE_RASTERIZATION_RULES" }, |
}; |
uint32_t *data = kgem->batch + offset; |
uint32_t opcode; |
unsigned int idx; |
opcode = (data[0] & 0x1f000000) >> 24; |
switch (opcode) { |
case 0x1f: |
return gen3_decode_3d_primitive(kgem, offset); |
case 0x1d: |
return gen3_decode_3d_1d(kgem, offset); |
case 0x1c: |
return gen3_decode_3d_1c(kgem, offset); |
} |
for (idx = 0; idx < ARRAY_SIZE(opcodes); idx++) { |
if (opcode == opcodes[idx].opcode) { |
unsigned int len = 1, i; |
kgem_debug_print(data, offset, 0, "%s\n", opcodes[idx].name); |
if (opcodes[idx].max_len > 1) { |
len = (data[0] & 0xff) + 2; |
assert(len >= opcodes[idx].min_len || |
len <= opcodes[idx].max_len); |
} |
for (i = 1; i < len; i++) |
kgem_debug_print(data, offset, i, "dword %d\n", i); |
return len; |
} |
} |
kgem_debug_print(data, offset, 0, "3D UNKNOWN: 3d opcode = 0x%x\n", opcode); |
return 1; |
} |
void kgem_gen3_finish_state(struct kgem *kgem) |
{ |
memset(&state, 0, sizeof(state)); |
} |
/drivers/video/Intel-2D/sna/kgem_debug_gen4.c |
---|
0,0 → 1,687 |
/* |
* Copyright © 2007-2011 Intel Corporation |
* |
* Permission is hereby granted, free of charge, to any person obtaining a |
* copy of this software and associated documentation files (the "Software"), |
* to deal in the Software without restriction, including without limitation |
* the rights to use, copy, modify, merge, publish, distribute, sublicense, |
* and/or sell copies of the Software, and to permit persons to whom the |
* Software is furnished to do so, subject to the following conditions: |
* |
* The above copyright notice and this permission notice (including the next |
* paragraph) shall be included in all copies or substantial portions of the |
* Software. |
* |
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL |
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, |
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE |
* SOFTWARE. |
* |
* Authors: |
* Eric Anholt <eric@anholt.net> |
* Chris Wilson <chris@chris-wilson.co.uk> |
* |
*/ |
#ifdef HAVE_CONFIG_H |
#include "config.h" |
#endif |
#include <assert.h> |
#include "sna.h" |
#include "sna_reg.h" |
#include "gen4_render.h" |
#include "kgem_debug.h" |
static struct state { |
struct vertex_buffer { |
int handle; |
void *base; |
const char *ptr; |
int pitch; |
struct kgem_bo *current; |
} vb[33]; |
struct vertex_elements { |
int buffer; |
int offset; |
bool valid; |
uint32_t type; |
uint8_t swizzle[4]; |
} ve[33]; |
int num_ve; |
struct dynamic_state { |
struct kgem_bo *current; |
void *base, *ptr; |
} dynamic_state; |
} state; |
static void gen4_update_vertex_buffer(struct kgem *kgem, const uint32_t *data) |
{ |
uint32_t reloc = sizeof(uint32_t) * (&data[1] - kgem->batch); |
struct kgem_bo *bo = NULL; |
void *base, *ptr; |
int i; |
for (i = 0; i < kgem->nreloc; i++) |
if (kgem->reloc[i].offset == reloc) |
break; |
assert(i < kgem->nreloc); |
reloc = kgem->reloc[i].target_handle; |
if (reloc == 0) { |
base = kgem->batch; |
} else { |
list_for_each_entry(bo, &kgem->next_request->buffers, request) |
if (bo->handle == reloc) |
break; |
assert(&bo->request != &kgem->next_request->buffers); |
base = kgem_bo_map__debug(kgem, bo); |
} |
ptr = (char *)base + kgem->reloc[i].delta; |
i = data[0] >> 27; |
state.vb[i].current = bo; |
state.vb[i].base = base; |
state.vb[i].ptr = ptr; |
state.vb[i].pitch = data[0] & 0x7ff; |
} |
static uint32_t |
get_ve_component(uint32_t data, int component) |
{ |
return (data >> (16 + (3 - component) * 4)) & 0x7; |
} |
static void gen4_update_vertex_elements(struct kgem *kgem, int id, const uint32_t *data) |
{ |
state.ve[id].buffer = data[0] >> 27; |
state.ve[id].valid = !!(data[0] & (1 << 26)); |
state.ve[id].type = (data[0] >> 16) & 0x1ff; |
state.ve[id].offset = data[0] & 0x7ff; |
state.ve[id].swizzle[0] = get_ve_component(data[1], 0); |
state.ve[id].swizzle[1] = get_ve_component(data[1], 1); |
state.ve[id].swizzle[2] = get_ve_component(data[1], 2); |
state.ve[id].swizzle[3] = get_ve_component(data[1], 3); |
} |
static void vertices_sint16_out(const struct vertex_elements *ve, const int16_t *v, int max) |
{ |
int c; |
ErrorF("("); |
for (c = 0; c < max; c++) { |
switch (ve->swizzle[c]) { |
case 0: ErrorF("#"); break; |
case 1: ErrorF("%d", v[c]); break; |
case 2: ErrorF("0.0"); break; |
case 3: ErrorF("1.0"); break; |
case 4: ErrorF("0x1"); break; |
case 5: break; |
default: ErrorF("?"); |
} |
if (c < 3) |
ErrorF(", "); |
} |
for (; c < 4; c++) { |
switch (ve->swizzle[c]) { |
case 0: ErrorF("#"); break; |
case 1: ErrorF("1.0"); break; |
case 2: ErrorF("0.0"); break; |
case 3: ErrorF("1.0"); break; |
case 4: ErrorF("0x1"); break; |
case 5: break; |
default: ErrorF("?"); |
} |
if (c < 3) |
ErrorF(", "); |
} |
ErrorF(")"); |
} |
static void vertices_float_out(const struct vertex_elements *ve, const float *f, int max) |
{ |
int c, o; |
ErrorF("("); |
for (c = o = 0; c < 4 && o < max; c++) { |
switch (ve->swizzle[c]) { |
case 0: ErrorF("#"); break; |
case 1: ErrorF("%f", f[o++]); break; |
case 2: ErrorF("0.0"); break; |
case 3: ErrorF("1.0"); break; |
case 4: ErrorF("0x1"); break; |
case 5: break; |
default: ErrorF("?"); |
} |
if (c < 3) |
ErrorF(", "); |
} |
for (; c < 4; c++) { |
switch (ve->swizzle[c]) { |
case 0: ErrorF("#"); break; |
case 1: ErrorF("1.0"); break; |
case 2: ErrorF("0.0"); break; |
case 3: ErrorF("1.0"); break; |
case 4: ErrorF("0x1"); break; |
case 5: break; |
default: ErrorF("?"); |
} |
if (c < 3) |
ErrorF(", "); |
} |
ErrorF(")"); |
} |
static void ve_out(const struct vertex_elements *ve, const void *ptr) |
{ |
switch (ve->type) { |
case GEN4_SURFACEFORMAT_R32_FLOAT: |
vertices_float_out(ve, ptr, 1); |
break; |
case GEN4_SURFACEFORMAT_R32G32_FLOAT: |
vertices_float_out(ve, ptr, 2); |
break; |
case GEN4_SURFACEFORMAT_R32G32B32_FLOAT: |
vertices_float_out(ve, ptr, 3); |
break; |
case GEN4_SURFACEFORMAT_R32G32B32A32_FLOAT: |
vertices_float_out(ve, ptr, 4); |
break; |
case GEN4_SURFACEFORMAT_R16_SINT: |
vertices_sint16_out(ve, ptr, 1); |
break; |
case GEN4_SURFACEFORMAT_R16G16_SINT: |
vertices_sint16_out(ve, ptr, 2); |
break; |
case GEN4_SURFACEFORMAT_R16G16B16A16_SINT: |
vertices_sint16_out(ve, ptr, 4); |
break; |
case GEN4_SURFACEFORMAT_R16_SSCALED: |
vertices_sint16_out(ve, ptr, 1); |
break; |
case GEN4_SURFACEFORMAT_R16G16_SSCALED: |
vertices_sint16_out(ve, ptr, 2); |
break; |
case GEN4_SURFACEFORMAT_R16G16B16A16_SSCALED: |
vertices_sint16_out(ve, ptr, 4); |
break; |
} |
} |
static void indirect_vertex_out(struct kgem *kgem, uint32_t v) |
{ |
int i = 0; |
do { |
const struct vertex_elements *ve = &state.ve[i]; |
const struct vertex_buffer *vb = &state.vb[ve->buffer]; |
const void *ptr = vb->ptr + v * vb->pitch + ve->offset; |
if (!ve->valid) |
continue; |
ve_out(ve, ptr); |
while (++i <= state.num_ve && !state.ve[i].valid) |
; |
if (i <= state.num_ve) |
ErrorF(", "); |
} while (i <= state.num_ve); |
} |
static void primitive_out(struct kgem *kgem, uint32_t *data) |
{ |
int n; |
assert((data[0] & (1<<15)) == 0); /* XXX index buffers */ |
for (n = 0; n < data[1]; n++) { |
int v = data[2] + n; |
ErrorF(" [%d:%d] = ", n, v); |
indirect_vertex_out(kgem, v); |
ErrorF("\n"); |
} |
} |
static void |
state_base_out(uint32_t *data, uint32_t offset, unsigned int index, |
const char *name) |
{ |
if (data[index] & 1) |
kgem_debug_print(data, offset, index, |
"%s state base address 0x%08x\n", |
name, data[index] & ~1); |
else |
kgem_debug_print(data, offset, index, |
"%s state base not updated\n", |
name); |
} |
static void |
state_max_out(uint32_t *data, uint32_t offset, unsigned int index, |
const char *name) |
{ |
if (data[index] == 1) |
kgem_debug_print(data, offset, index, |
"%s state upper bound disabled\n", name); |
else if (data[index] & 1) |
kgem_debug_print(data, offset, index, |
"%s state upper bound 0x%08x\n", |
name, data[index] & ~1); |
else |
kgem_debug_print(data, offset, index, |
"%s state upper bound not updated\n", |
name); |
} |
static const char * |
get_965_surfacetype(unsigned int surfacetype) |
{ |
switch (surfacetype) { |
case 0: return "1D"; |
case 1: return "2D"; |
case 2: return "3D"; |
case 3: return "CUBE"; |
case 4: return "BUFFER"; |
case 7: return "NULL"; |
default: return "unknown"; |
} |
} |
static const char * |
get_965_depthformat(unsigned int depthformat) |
{ |
switch (depthformat) { |
case 0: return "s8_z24float"; |
case 1: return "z32float"; |
case 2: return "z24s8"; |
case 5: return "z16"; |
default: return "unknown"; |
} |
} |
static const char * |
get_965_element_component(uint32_t data, int component) |
{ |
uint32_t component_control = (data >> (16 + (3 - component) * 4)) & 0x7; |
switch (component_control) { |
case 0: |
return "nostore"; |
case 1: |
switch (component) { |
case 0: return "X"; |
case 1: return "Y"; |
case 2: return "Z"; |
case 3: return "W"; |
default: return "fail"; |
} |
case 2: |
return "0.0"; |
case 3: |
return "1.0"; |
case 4: |
return "0x1"; |
case 5: |
return "VID"; |
default: |
return "fail"; |
} |
} |
static const char * |
get_965_prim_type(uint32_t data) |
{ |
uint32_t primtype = (data >> 10) & 0x1f; |
switch (primtype) { |
case 0x01: return "point list"; |
case 0x02: return "line list"; |
case 0x03: return "line strip"; |
case 0x04: return "tri list"; |
case 0x05: return "tri strip"; |
case 0x06: return "tri fan"; |
case 0x07: return "quad list"; |
case 0x08: return "quad strip"; |
case 0x09: return "line list adj"; |
case 0x0a: return "line strip adj"; |
case 0x0b: return "tri list adj"; |
case 0x0c: return "tri strip adj"; |
case 0x0d: return "tri strip reverse"; |
case 0x0e: return "polygon"; |
case 0x0f: return "rect list"; |
case 0x10: return "line loop"; |
case 0x11: return "point list bf"; |
case 0x12: return "line strip cont"; |
case 0x13: return "line strip bf"; |
case 0x14: return "line strip cont bf"; |
case 0x15: return "tri fan no stipple"; |
default: return "fail"; |
} |
} |
#if 0 |
struct reloc { |
struct kgem_bo *bo; |
void *base; |
}; |
static void * |
get_reloc(struct kgem *kgem, |
void *base, const uint32_t *reloc, |
struct reloc *r) |
{ |
uint32_t delta = *reloc; |
memset(r, 0, sizeof(*r)); |
if (base == 0) { |
uint32_t handle = sizeof(uint32_t) * (reloc - kgem->batch); |
struct kgem_bo *bo = NULL; |
int i; |
for (i = 0; i < kgem->nreloc; i++) |
if (kgem->reloc[i].offset == handle) |
break; |
assert(i < kgem->nreloc); |
handle = kgem->reloc[i].target_handle; |
delta = kgem->reloc[i].delta; |
if (handle == 0) { |
base = kgem->batch; |
} else { |
list_for_each_entry(bo, &kgem->next_request->buffers, request) |
if (bo->handle == handle) |
break; |
assert(&bo->request != &kgem->next_request->buffers); |
base = kgem_bo_map__debug(kgem, bo); |
r->bo = bo; |
r->base = base; |
} |
} |
return (char *)base + delta; |
} |
#endif |
int kgem_gen4_decode_3d(struct kgem *kgem, uint32_t offset) |
{ |
static const struct { |
uint32_t opcode; |
int min_len; |
int max_len; |
const char *name; |
} opcodes[] = { |
{ 0x6000, 3, 3, "URB_FENCE" }, |
{ 0x6001, 2, 2, "CS_URB_FENCE" }, |
{ 0x6002, 2, 2, "CONSTANT_BUFFER" }, |
{ 0x6101, 6, 6, "STATE_BASE_ADDRESS" }, |
{ 0x6102, 2, 2 , "STATE_SIP" }, |
{ 0x6104, 1, 1, "3DSTATE_PIPELINE_SELECT" }, |
{ 0x680b, 1, 1, "3DSTATE_VF_STATISTICS" }, |
{ 0x6904, 1, 1, "3DSTATE_PIPELINE_SELECT" }, |
{ 0x7800, 7, 7, "3DSTATE_PIPELINED_POINTERS" }, |
{ 0x7801, 6, 6, "3DSTATE_BINDING_TABLE_POINTERS" }, |
{ 0x7808, 5, 257, "3DSTATE_VERTEX_BUFFERS" }, |
{ 0x7809, 3, 256, "3DSTATE_VERTEX_ELEMENTS" }, |
{ 0x780a, 3, 3, "3DSTATE_INDEX_BUFFER" }, |
{ 0x780b, 1, 1, "3DSTATE_VF_STATISTICS" }, |
{ 0x7900, 4, 4, "3DSTATE_DRAWING_RECTANGLE" }, |
{ 0x7901, 5, 5, "3DSTATE_CONSTANT_COLOR" }, |
{ 0x7905, 5, 7, "3DSTATE_DEPTH_BUFFER" }, |
{ 0x7906, 2, 2, "3DSTATE_POLY_STIPPLE_OFFSET" }, |
{ 0x7907, 33, 33, "3DSTATE_POLY_STIPPLE_PATTERN" }, |
{ 0x7908, 3, 3, "3DSTATE_LINE_STIPPLE" }, |
{ 0x7909, 2, 2, "3DSTATE_GLOBAL_DEPTH_OFFSET_CLAMP" }, |
{ 0x7909, 2, 2, "3DSTATE_CLEAR_PARAMS" }, |
{ 0x790a, 3, 3, "3DSTATE_AA_LINE_PARAMETERS" }, |
{ 0x790b, 4, 4, "3DSTATE_GS_SVB_INDEX" }, |
{ 0x790d, 3, 3, "3DSTATE_MULTISAMPLE" }, |
{ 0x7910, 2, 2, "3DSTATE_CLEAR_PARAMS" }, |
{ 0x7b00, 6, 6, "3DPRIMITIVE" }, |
{ 0x7805, 3, 3, "3DSTATE_URB" }, |
{ 0x7815, 5, 5, "3DSTATE_CONSTANT_VS_STATE" }, |
{ 0x7816, 5, 5, "3DSTATE_CONSTANT_GS_STATE" }, |
{ 0x7817, 5, 5, "3DSTATE_CONSTANT_PS_STATE" }, |
{ 0x7818, 2, 2, "3DSTATE_SAMPLE_MASK" }, |
}; |
uint32_t *data = kgem->batch + offset; |
uint32_t op; |
unsigned int len; |
int i; |
const char *desc1 = NULL; |
len = (data[0] & 0xff) + 2; |
op = (data[0] & 0xffff0000) >> 16; |
switch (op) { |
case 0x6000: |
assert(len == 3); |
kgem_debug_print(data, offset, 0, "URB_FENCE: %s%s%s%s%s%s\n", |
(data[0] >> 13) & 1 ? "cs " : "", |
(data[0] >> 12) & 1 ? "vfe " : "", |
(data[0] >> 11) & 1 ? "sf " : "", |
(data[0] >> 10) & 1 ? "clip " : "", |
(data[0] >> 9) & 1 ? "gs " : "", |
(data[0] >> 8) & 1 ? "vs " : ""); |
kgem_debug_print(data, offset, 1, |
"vs fence: %d, gs_fence: %d, clip_fence: %d\n", |
data[1] & 0x3ff, |
(data[1] >> 10) & 0x3ff, |
(data[1] >> 20) & 0x3ff); |
kgem_debug_print(data, offset, 2, |
"sf fence: %d, vfe_fence: %d, cs_fence: %d\n", |
data[2] & 0x3ff, |
(data[2] >> 10) & 0x3ff, |
(data[2] >> 20) & 0x7ff); |
return len; |
case 0x6001: |
kgem_debug_print(data, offset, 0, "CS_URB_STATE\n"); |
kgem_debug_print(data, offset, 1, "entry_size: %d [%d bytes], n_entries: %d\n", |
(data[1] >> 4) & 0x1f, |
(((data[1] >> 4) & 0x1f) + 1) * 64, |
data[1] & 0x7); |
return len; |
case 0x6002: |
kgem_debug_print(data, offset, 0, "CONSTANT_BUFFER: %s\n", |
(data[0] >> 8) & 1 ? "valid" : "invalid"); |
kgem_debug_print(data, offset, 1, "offset: 0x%08x, length: %d bytes\n", |
data[1] & ~0x3f, ((data[1] & 0x3f) + 1) * 64); |
return len; |
case 0x6101: |
i = 0; |
kgem_debug_print(data, offset, i++, "STATE_BASE_ADDRESS\n"); |
assert(len == 6); |
state_base_out(data, offset, i++, "general"); |
state_base_out(data, offset, i++, "surface"); |
state_base_out(data, offset, i++, "media"); |
state_max_out(data, offset, i++, "general"); |
state_max_out(data, offset, i++, "media"); |
return len; |
case 0x7801: |
assert(len == 6); |
kgem_debug_print(data, offset, 0, |
"3DSTATE_BINDING_TABLE_POINTERS\n"); |
kgem_debug_print(data, offset, 1, "VS binding table\n"); |
kgem_debug_print(data, offset, 2, "GS binding table\n"); |
kgem_debug_print(data, offset, 3, "CLIP binding table\n"); |
kgem_debug_print(data, offset, 4, "SF binding table\n"); |
kgem_debug_print(data, offset, 5, "WM binding table\n"); |
return len; |
case 0x7808: |
assert((len - 1) % 4 == 0); |
kgem_debug_print(data, offset, 0, "3DSTATE_VERTEX_BUFFERS\n"); |
for (i = 1; i < len;) { |
gen4_update_vertex_buffer(kgem, data + i); |
kgem_debug_print(data, offset, i, "buffer %d: %s, pitch %db\n", |
data[i] >> 27, |
data[i] & (1 << 20) ? "random" : "sequential", |
data[i] & 0x07ff); |
i++; |
kgem_debug_print(data, offset, i++, "buffer address\n"); |
kgem_debug_print(data, offset, i++, "max index\n"); |
kgem_debug_print(data, offset, i++, "mbz\n"); |
} |
return len; |
case 0x7809: |
assert((len + 1) % 2 == 0); |
kgem_debug_print(data, offset, 0, "3DSTATE_VERTEX_ELEMENTS\n"); |
memset(state.ve, 0, sizeof(state.ve)); /* XXX? */ |
for (i = 1; i < len;) { |
gen4_update_vertex_elements(kgem, (i - 1)/2, data + i); |
kgem_debug_print(data, offset, i, "buffer %d: %svalid, type 0x%04x, " |
"src offset 0x%04x bytes\n", |
data[i] >> 27, |
data[i] & (1 << 26) ? "" : "in", |
(data[i] >> 16) & 0x1ff, |
data[i] & 0x07ff); |
i++; |
kgem_debug_print(data, offset, i, "(%s, %s, %s, %s), " |
"dst offset 0x%02x bytes\n", |
get_965_element_component(data[i], 0), |
get_965_element_component(data[i], 1), |
get_965_element_component(data[i], 2), |
get_965_element_component(data[i], 3), |
(data[i] & 0xff) * 4); |
i++; |
} |
state.num_ve = (len - 1) / 2; /* XXX? */ |
return len; |
case 0x780a: |
assert(len == 3); |
kgem_debug_print(data, offset, 0, "3DSTATE_INDEX_BUFFER\n"); |
kgem_debug_print(data, offset, 1, "beginning buffer address\n"); |
kgem_debug_print(data, offset, 2, "ending buffer address\n"); |
return len; |
case 0x7900: |
assert(len == 4); |
kgem_debug_print(data, offset, 0, |
"3DSTATE_DRAWING_RECTANGLE\n"); |
kgem_debug_print(data, offset, 1, "top left: %d,%d\n", |
data[1] & 0xffff, |
(data[1] >> 16) & 0xffff); |
kgem_debug_print(data, offset, 2, "bottom right: %d,%d\n", |
data[2] & 0xffff, |
(data[2] >> 16) & 0xffff); |
kgem_debug_print(data, offset, 3, "origin: %d,%d\n", |
(int)data[3] & 0xffff, |
((int)data[3] >> 16) & 0xffff); |
return len; |
case 0x7905: |
assert(len == 7); |
kgem_debug_print(data, offset, 0, |
"3DSTATE_DEPTH_BUFFER\n"); |
kgem_debug_print(data, offset, 1, "%s, %s, pitch = %d bytes, %stiled, HiZ %d, Seperate Stencil %d\n", |
get_965_surfacetype(data[1] >> 29), |
get_965_depthformat((data[1] >> 18) & 0x7), |
(data[1] & 0x0001ffff) + 1, |
data[1] & (1 << 27) ? "" : "not ", |
(data[1] & (1 << 22)) != 0, |
(data[1] & (1 << 21)) != 0); |
kgem_debug_print(data, offset, 2, "depth offset\n"); |
kgem_debug_print(data, offset, 3, "%dx%d\n", |
((data[3] & 0x0007ffc0) >> 6) + 1, |
((data[3] & 0xfff80000) >> 19) + 1); |
kgem_debug_print(data, offset, 4, "volume depth\n"); |
kgem_debug_print(data, offset, 5, "\n"); |
kgem_debug_print(data, offset, 6, "\n"); |
return len; |
case 0x7a00: |
assert(len == 4 || len == 5); |
switch ((data[1] >> 14) & 0x3) { |
case 0: desc1 = "no write"; break; |
case 1: desc1 = "qword write"; break; |
case 2: desc1 = "PS_DEPTH_COUNT write"; break; |
case 3: desc1 = "TIMESTAMP write"; break; |
} |
kgem_debug_print(data, offset, 0, "PIPE_CONTROL\n"); |
kgem_debug_print(data, offset, 1, |
"%s, %scs stall, %stlb invalidate, " |
"%ssync gfdt, %sdepth stall, %sRC write flush, " |
"%sinst flush, %sTC flush\n", |
desc1, |
data[1] & (1 << 20) ? "" : "no ", |
data[1] & (1 << 18) ? "" : "no ", |
data[1] & (1 << 17) ? "" : "no ", |
data[1] & (1 << 13) ? "" : "no ", |
data[1] & (1 << 12) ? "" : "no ", |
data[1] & (1 << 11) ? "" : "no ", |
data[1] & (1 << 10) ? "" : "no "); |
if (len == 5) { |
kgem_debug_print(data, offset, 2, "destination address\n"); |
kgem_debug_print(data, offset, 3, "immediate dword low\n"); |
kgem_debug_print(data, offset, 4, "immediate dword high\n"); |
} else { |
for (i = 2; i < len; i++) { |
kgem_debug_print(data, offset, i, "\n"); |
} |
} |
return len; |
case 0x7b00: |
assert(len == 6); |
kgem_debug_print(data, offset, 0, |
"3DPRIMITIVE: %s %s\n", |
get_965_prim_type(data[0]), |
(data[0] & (1 << 15)) ? "random" : "sequential"); |
kgem_debug_print(data, offset, 1, "vertex count\n"); |
kgem_debug_print(data, offset, 2, "start vertex\n"); |
kgem_debug_print(data, offset, 3, "instance count\n"); |
kgem_debug_print(data, offset, 4, "start instance\n"); |
kgem_debug_print(data, offset, 5, "index bias\n"); |
primitive_out(kgem, data); |
return len; |
} |
/* For the rest, just dump the bytes */ |
for (i = 0; i < ARRAY_SIZE(opcodes); i++) |
if (op == opcodes[i].opcode) |
break; |
assert(i < ARRAY_SIZE(opcodes)); |
len = 1; |
kgem_debug_print(data, offset, 0, "%s\n", opcodes[i].name); |
if (opcodes[i].max_len > 1) { |
len = (data[0] & 0xff) + 2; |
assert(len >= opcodes[i].min_len && |
len <= opcodes[i].max_len); |
} |
for (i = 1; i < len; i++) |
kgem_debug_print(data, offset, i, "dword %d\n", i); |
return len; |
} |
void kgem_gen4_finish_state(struct kgem *kgem) |
{ |
memset(&state, 0, sizeof(state)); |
} |
/drivers/video/Intel-2D/sna/kgem_debug_gen5.c |
---|
0,0 → 1,662 |
/* |
* Copyright © 2007-2011 Intel Corporation |
* |
* Permission is hereby granted, free of charge, to any person obtaining a |
* copy of this software and associated documentation files (the "Software"), |
* to deal in the Software without restriction, including without limitation |
* the rights to use, copy, modify, merge, publish, distribute, sublicense, |
* and/or sell copies of the Software, and to permit persons to whom the |
* Software is furnished to do so, subject to the following conditions: |
* |
* The above copyright notice and this permission notice (including the next |
* paragraph) shall be included in all copies or substantial portions of the |
* Software. |
* |
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL |
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, |
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE |
* SOFTWARE. |
* |
* Authors: |
* Eric Anholt <eric@anholt.net> |
* Chris Wilson <chris@chris-wilson.co.uk> |
* |
*/ |
#ifdef HAVE_CONFIG_H |
#include "config.h" |
#endif |
#include <assert.h> |
#include "sna.h" |
#include "sna_reg.h" |
#include "gen5_render.h" |
#include "kgem_debug.h" |
static struct state { |
struct vertex_buffer { |
int handle; |
void *base; |
int size; |
const char *ptr; |
int pitch; |
struct kgem_bo *current; |
} vb[17]; |
struct vertex_elements { |
int buffer; |
int offset; |
bool valid; |
uint32_t type; |
uint8_t swizzle[4]; |
} ve[17]; |
int num_ve; |
struct dynamic_state { |
struct kgem_bo *current; |
void *base, *ptr; |
} dynamic_state; |
} state; |
static void gen5_update_vertex_buffer(struct kgem *kgem, const uint32_t *data) |
{ |
struct drm_i915_gem_relocation_entry *reloc; |
struct kgem_bo *bo = NULL; |
void *base, *ptr; |
int i, size; |
reloc = kgem_debug_get_reloc_entry(kgem, &data[1] - kgem->batch); |
if (reloc->target_handle == -1) { |
base = kgem->batch; |
size = kgem->nbatch * sizeof(uint32_t); |
} else { |
bo = kgem_debug_get_bo_for_reloc_entry(kgem, reloc); |
base = kgem_bo_map__debug(kgem, bo); |
size = kgem_bo_size(bo); |
} |
ptr = (char *)base + reloc->delta; |
i = data[0] >> 27; |
state.vb[i].handle = reloc->target_handle; |
state.vb[i].current = bo; |
state.vb[i].base = base; |
state.vb[i].ptr = ptr; |
state.vb[i].pitch = data[0] & 0x7ff; |
state.vb[i].size = size; |
} |
static uint32_t |
get_ve_component(uint32_t data, int component) |
{ |
return (data >> (16 + (3 - component) * 4)) & 0x7; |
} |
static void gen5_update_vertex_elements(struct kgem *kgem, int id, const uint32_t *data) |
{ |
state.ve[id].buffer = data[0] >> 27; |
state.ve[id].valid = !!(data[0] & (1 << 26)); |
state.ve[id].type = (data[0] >> 16) & 0x1ff; |
state.ve[id].offset = data[0] & 0x7ff; |
state.ve[id].swizzle[0] = get_ve_component(data[1], 0); |
state.ve[id].swizzle[1] = get_ve_component(data[1], 1); |
state.ve[id].swizzle[2] = get_ve_component(data[1], 2); |
state.ve[id].swizzle[3] = get_ve_component(data[1], 3); |
} |
static void vertices_sint16_out(const struct vertex_elements *ve, const int16_t *v, int max) |
{ |
int c, o; |
ErrorF("("); |
for (c = o = 0; c < 4 && o < max; c++) { |
switch (ve->swizzle[c]) { |
case 0: ErrorF("#"); break; |
case 1: ErrorF("%d", v[o++]); break; |
case 2: ErrorF("0.0"); break; |
case 3: ErrorF("1.0"); break; |
case 4: ErrorF("0x1"); break; |
case 5: break; |
default: ErrorF("?"); |
} |
if (o < max) |
ErrorF(", "); |
} |
ErrorF(")"); |
} |
static void vertices_float_out(const struct vertex_elements *ve, const float *f, int max) |
{ |
int c, o; |
ErrorF("("); |
for (c = o = 0; c < 4 && o < max; c++) { |
switch (ve->swizzle[c]) { |
case 0: ErrorF("#"); break; |
case 1: ErrorF("%f", f[o++]); break; |
case 2: ErrorF("0.0"); break; |
case 3: ErrorF("1.0"); break; |
case 4: ErrorF("0x1"); break; |
case 5: break; |
default: ErrorF("?"); |
} |
if (o < max) |
ErrorF(", "); |
} |
ErrorF(")"); |
} |
static void ve_out(const struct vertex_elements *ve, const void *ptr) |
{ |
switch (ve->type) { |
case GEN5_SURFACEFORMAT_R32_FLOAT: |
vertices_float_out(ve, ptr, 1); |
break; |
case GEN5_SURFACEFORMAT_R32G32_FLOAT: |
vertices_float_out(ve, ptr, 2); |
break; |
case GEN5_SURFACEFORMAT_R32G32B32_FLOAT: |
vertices_float_out(ve, ptr, 3); |
break; |
case GEN5_SURFACEFORMAT_R32G32B32A32_FLOAT: |
vertices_float_out(ve, ptr, 4); |
break; |
case GEN5_SURFACEFORMAT_R16_SINT: |
vertices_sint16_out(ve, ptr, 1); |
break; |
case GEN5_SURFACEFORMAT_R16G16_SINT: |
vertices_sint16_out(ve, ptr, 2); |
break; |
case GEN5_SURFACEFORMAT_R16G16B16A16_SINT: |
vertices_sint16_out(ve, ptr, 4); |
break; |
case GEN5_SURFACEFORMAT_R16_SSCALED: |
vertices_sint16_out(ve, ptr, 1); |
break; |
case GEN5_SURFACEFORMAT_R16G16_SSCALED: |
vertices_sint16_out(ve, ptr, 2); |
break; |
case GEN5_SURFACEFORMAT_R16G16B16A16_SSCALED: |
vertices_sint16_out(ve, ptr, 4); |
break; |
} |
} |
static void indirect_vertex_out(struct kgem *kgem, uint32_t v) |
{ |
int i = 1; |
do { |
const struct vertex_elements *ve = &state.ve[i]; |
const struct vertex_buffer *vb = &state.vb[ve->buffer]; |
const void *ptr = vb->ptr + v * vb->pitch + ve->offset; |
if (!ve->valid) |
continue; |
assert(vb->pitch); |
assert(ve->offset + v*vb->pitch < vb->size); |
ve_out(ve, ptr); |
while (++i <= state.num_ve && !state.ve[i].valid) |
; |
if (i <= state.num_ve) |
ErrorF(", "); |
} while (i <= state.num_ve); |
} |
static void primitive_out(struct kgem *kgem, uint32_t *data) |
{ |
int n; |
assert((data[0] & (1<<15)) == 0); /* XXX index buffers */ |
for (n = 0; n < data[1]; n++) { |
int v = data[2] + n; |
ErrorF(" [%d:%d] = ", n, v); |
indirect_vertex_out(kgem, v); |
ErrorF("\n"); |
} |
} |
static void |
state_base_out(uint32_t *data, uint32_t offset, unsigned int index, |
const char *name) |
{ |
if (data[index] & 1) |
kgem_debug_print(data, offset, index, |
"%s state base address 0x%08x\n", |
name, data[index] & ~1); |
else |
kgem_debug_print(data, offset, index, |
"%s state base not updated\n", |
name); |
} |
static void |
state_max_out(uint32_t *data, uint32_t offset, unsigned int index, |
const char *name) |
{ |
if (data[index] == 1) |
kgem_debug_print(data, offset, index, |
"%s state upper bound disabled\n", name); |
else if (data[index] & 1) |
kgem_debug_print(data, offset, index, |
"%s state upper bound 0x%08x\n", |
name, data[index] & ~1); |
else |
kgem_debug_print(data, offset, index, |
"%s state upper bound not updated\n", |
name); |
} |
static const char * |
get_965_surfacetype(unsigned int surfacetype) |
{ |
switch (surfacetype) { |
case 0: return "1D"; |
case 1: return "2D"; |
case 2: return "3D"; |
case 3: return "CUBE"; |
case 4: return "BUFFER"; |
case 7: return "NULL"; |
default: return "unknown"; |
} |
} |
static const char * |
get_965_depthformat(unsigned int depthformat) |
{ |
switch (depthformat) { |
case 0: return "s8_z24float"; |
case 1: return "z32float"; |
case 2: return "z24s8"; |
case 5: return "z16"; |
default: return "unknown"; |
} |
} |
static const char * |
get_965_element_component(uint32_t data, int component) |
{ |
uint32_t component_control = (data >> (16 + (3 - component) * 4)) & 0x7; |
switch (component_control) { |
case 0: |
return "nostore"; |
case 1: |
switch (component) { |
case 0: return "X"; |
case 1: return "Y"; |
case 2: return "Z"; |
case 3: return "W"; |
default: return "fail"; |
} |
case 2: |
return "0.0"; |
case 3: |
return "1.0"; |
case 4: |
return "0x1"; |
case 5: |
return "VID"; |
default: |
return "fail"; |
} |
} |
static const char * |
get_965_prim_type(uint32_t data) |
{ |
uint32_t primtype = (data >> 10) & 0x1f; |
switch (primtype) { |
case 0x01: return "point list"; |
case 0x02: return "line list"; |
case 0x03: return "line strip"; |
case 0x04: return "tri list"; |
case 0x05: return "tri strip"; |
case 0x06: return "tri fan"; |
case 0x07: return "quad list"; |
case 0x08: return "quad strip"; |
case 0x09: return "line list adj"; |
case 0x0a: return "line strip adj"; |
case 0x0b: return "tri list adj"; |
case 0x0c: return "tri strip adj"; |
case 0x0d: return "tri strip reverse"; |
case 0x0e: return "polygon"; |
case 0x0f: return "rect list"; |
case 0x10: return "line loop"; |
case 0x11: return "point list bf"; |
case 0x12: return "line strip cont"; |
case 0x13: return "line strip bf"; |
case 0x14: return "line strip cont bf"; |
case 0x15: return "tri fan no stipple"; |
default: return "fail"; |
} |
} |
#if 0 |
struct reloc { |
struct kgem_bo *bo; |
void *base; |
}; |
static void * |
get_reloc(struct kgem *kgem, |
void *base, const uint32_t *reloc, |
struct reloc *r) |
{ |
uint32_t delta = *reloc; |
memset(r, 0, sizeof(*r)); |
if (base == 0) { |
uint32_t handle = sizeof(uint32_t) * (reloc - kgem->batch); |
struct kgem_bo *bo = NULL; |
int i; |
for (i = 0; i < kgem->nreloc; i++) |
if (kgem->reloc[i].offset == handle) |
break; |
assert(i < kgem->nreloc); |
handle = kgem->reloc[i].target_handle; |
delta = kgem->reloc[i].delta; |
if (handle == 0) { |
base = kgem->batch; |
} else { |
list_for_each_entry(bo, &kgem->next_request->buffers, request) |
if (bo->handle == handle) |
break; |
assert(&bo->request != &kgem->next_request->buffers); |
base = kgem_bo_map(kgem, bo, PROT_READ); |
r->bo = bo; |
r->base = base; |
} |
} |
return (char *)base + delta; |
} |
#endif |
int kgem_gen5_decode_3d(struct kgem *kgem, uint32_t offset) |
{ |
static const struct { |
uint32_t opcode; |
int min_len; |
int max_len; |
const char *name; |
} opcodes[] = { |
{ 0x6000, 3, 3, "URB_FENCE" }, |
{ 0x6001, 2, 2, "CS_URB_FENCE" }, |
{ 0x6002, 2, 2, "CONSTANT_BUFFER" }, |
{ 0x6101, 6, 6, "STATE_BASE_ADDRESS" }, |
{ 0x6102, 2, 2 , "STATE_SIP" }, |
{ 0x6104, 1, 1, "3DSTATE_PIPELINE_SELECT" }, |
{ 0x680b, 1, 1, "3DSTATE_VF_STATISTICS" }, |
{ 0x6904, 1, 1, "3DSTATE_PIPELINE_SELECT" }, |
{ 0x7800, 7, 7, "3DSTATE_PIPELINED_POINTERS" }, |
{ 0x7801, 6, 6, "3DSTATE_BINDING_TABLE_POINTERS" }, |
{ 0x7808, 5, 257, "3DSTATE_VERTEX_BUFFERS" }, |
{ 0x7809, 3, 256, "3DSTATE_VERTEX_ELEMENTS" }, |
{ 0x780a, 3, 3, "3DSTATE_INDEX_BUFFER" }, |
{ 0x780b, 1, 1, "3DSTATE_VF_STATISTICS" }, |
{ 0x7900, 4, 4, "3DSTATE_DRAWING_RECTANGLE" }, |
{ 0x7901, 5, 5, "3DSTATE_CONSTANT_COLOR" }, |
{ 0x7905, 5, 7, "3DSTATE_DEPTH_BUFFER" }, |
{ 0x7906, 2, 2, "3DSTATE_POLY_STIPPLE_OFFSET" }, |
{ 0x7907, 33, 33, "3DSTATE_POLY_STIPPLE_PATTERN" }, |
{ 0x7908, 3, 3, "3DSTATE_LINE_STIPPLE" }, |
{ 0x7909, 2, 2, "3DSTATE_GLOBAL_DEPTH_OFFSET_CLAMP" }, |
{ 0x7909, 2, 2, "3DSTATE_CLEAR_PARAMS" }, |
{ 0x790a, 3, 3, "3DSTATE_AA_LINE_PARAMETERS" }, |
{ 0x790b, 4, 4, "3DSTATE_GS_SVB_INDEX" }, |
{ 0x790d, 3, 3, "3DSTATE_MULTISAMPLE" }, |
{ 0x7910, 2, 2, "3DSTATE_CLEAR_PARAMS" }, |
{ 0x7b00, 6, 6, "3DPRIMITIVE" }, |
{ 0x7805, 3, 3, "3DSTATE_URB" }, |
{ 0x7815, 5, 5, "3DSTATE_CONSTANT_VS_STATE" }, |
{ 0x7816, 5, 5, "3DSTATE_CONSTANT_GS_STATE" }, |
{ 0x7817, 5, 5, "3DSTATE_CONSTANT_PS_STATE" }, |
{ 0x7818, 2, 2, "3DSTATE_SAMPLE_MASK" }, |
}; |
uint32_t *data = kgem->batch + offset; |
uint32_t op; |
unsigned int len; |
int i; |
const char *desc1 = NULL; |
len = (data[0] & 0xff) + 2; |
op = (data[0] & 0xffff0000) >> 16; |
switch (op) { |
case 0x6000: |
assert(len == 3); |
kgem_debug_print(data, offset, 0, "URB_FENCE: %s%s%s%s%s%s\n", |
(data[0] >> 13) & 1 ? "cs " : "", |
(data[0] >> 12) & 1 ? "vfe " : "", |
(data[0] >> 11) & 1 ? "sf " : "", |
(data[0] >> 10) & 1 ? "clip " : "", |
(data[0] >> 9) & 1 ? "gs " : "", |
(data[0] >> 8) & 1 ? "vs " : ""); |
kgem_debug_print(data, offset, 1, |
"vs fence: %d, gs_fence: %d, clip_fence: %d\n", |
data[1] & 0x3ff, |
(data[1] >> 10) & 0x3ff, |
(data[1] >> 20) & 0x3ff); |
kgem_debug_print(data, offset, 2, |
"sf fence: %d, vfe_fence: %d, cs_fence: %d\n", |
data[2] & 0x3ff, |
(data[2] >> 10) & 0x3ff, |
(data[2] >> 20) & 0x7ff); |
return len; |
case 0x6001: |
kgem_debug_print(data, offset, 0, "CS_URB_STATE\n"); |
kgem_debug_print(data, offset, 1, "entry_size: %d [%d bytes], n_entries: %d\n", |
(data[1] >> 4) & 0x1f, |
(((data[1] >> 4) & 0x1f) + 1) * 64, |
data[1] & 0x7); |
return len; |
case 0x6002: |
kgem_debug_print(data, offset, 0, "CONSTANT_BUFFER: %s\n", |
(data[0] >> 8) & 1 ? "valid" : "invalid"); |
kgem_debug_print(data, offset, 1, "offset: 0x%08x, length: %d bytes\n", |
data[1] & ~0x3f, ((data[1] & 0x3f) + 1) * 64); |
return len; |
case 0x6101: |
i = 0; |
kgem_debug_print(data, offset, i++, "STATE_BASE_ADDRESS\n"); |
assert(len == 8); |
state_base_out(data, offset, i++, "general"); |
state_base_out(data, offset, i++, "surface"); |
state_base_out(data, offset, i++, "media"); |
state_base_out(data, offset, i++, "instruction"); |
state_max_out(data, offset, i++, "general"); |
state_max_out(data, offset, i++, "media"); |
state_max_out(data, offset, i++, "instruction"); |
return len; |
case 0x7801: |
assert(len == 6); |
kgem_debug_print(data, offset, 0, |
"3DSTATE_BINDING_TABLE_POINTERS\n"); |
kgem_debug_print(data, offset, 1, "VS binding table\n"); |
kgem_debug_print(data, offset, 2, "GS binding table\n"); |
kgem_debug_print(data, offset, 3, "CLIP binding table\n"); |
kgem_debug_print(data, offset, 4, "SF binding table\n"); |
kgem_debug_print(data, offset, 5, "WM binding table\n"); |
return len; |
case 0x7808: |
assert((len - 1) % 4 == 0); |
kgem_debug_print(data, offset, 0, "3DSTATE_VERTEX_BUFFERS\n"); |
for (i = 1; i < len;) { |
gen5_update_vertex_buffer(kgem, data + i); |
kgem_debug_print(data, offset, i, "buffer %d: %s, pitch %db\n", |
data[i] >> 27, |
data[i] & (1 << 20) ? "random" : "sequential", |
data[i] & 0x07ff); |
i++; |
kgem_debug_print(data, offset, i++, "buffer address\n"); |
kgem_debug_print(data, offset, i++, "max index\n"); |
kgem_debug_print(data, offset, i++, "mbz\n"); |
} |
return len; |
case 0x7809: |
assert((len + 1) % 2 == 0); |
kgem_debug_print(data, offset, 0, "3DSTATE_VERTEX_ELEMENTS\n"); |
memset(state.ve, 0, sizeof(state.ve)); /* XXX? */ |
for (i = 1; i < len;) { |
gen5_update_vertex_elements(kgem, (i - 1)/2, data + i); |
kgem_debug_print(data, offset, i, |
"buffer %d: %svalid, type 0x%04x, " |
"src offset 0x%04x bytes\n", |
data[i] >> 27, |
data[i] & (1 << 26) ? "" : "in", |
(data[i] >> 16) & 0x1ff, |
data[i] & 0x07ff); |
i++; |
kgem_debug_print(data, offset, i, "(%s, %s, %s, %s)\n", |
get_965_element_component(data[i], 0), |
get_965_element_component(data[i], 1), |
get_965_element_component(data[i], 2), |
get_965_element_component(data[i], 3)); |
i++; |
} |
state.num_ve = (len - 1) / 2; /* XXX? */ |
return len; |
case 0x780a: |
assert(len == 3); |
kgem_debug_print(data, offset, 0, "3DSTATE_INDEX_BUFFER\n"); |
kgem_debug_print(data, offset, 1, "beginning buffer address\n"); |
kgem_debug_print(data, offset, 2, "ending buffer address\n"); |
return len; |
case 0x7900: |
assert(len == 4); |
kgem_debug_print(data, offset, 0, |
"3DSTATE_DRAWING_RECTANGLE\n"); |
kgem_debug_print(data, offset, 1, "top left: %d,%d\n", |
data[1] & 0xffff, |
(data[1] >> 16) & 0xffff); |
kgem_debug_print(data, offset, 2, "bottom right: %d,%d\n", |
data[2] & 0xffff, |
(data[2] >> 16) & 0xffff); |
kgem_debug_print(data, offset, 3, "origin: %d,%d\n", |
(int)data[3] & 0xffff, |
((int)data[3] >> 16) & 0xffff); |
return len; |
case 0x7905: |
assert(len == 7); |
kgem_debug_print(data, offset, 0, |
"3DSTATE_DEPTH_BUFFER\n"); |
kgem_debug_print(data, offset, 1, "%s, %s, pitch = %d bytes, %stiled, HiZ %d, Seperate Stencil %d\n", |
get_965_surfacetype(data[1] >> 29), |
get_965_depthformat((data[1] >> 18) & 0x7), |
(data[1] & 0x0001ffff) + 1, |
data[1] & (1 << 27) ? "" : "not ", |
(data[1] & (1 << 22)) != 0, |
(data[1] & (1 << 21)) != 0); |
kgem_debug_print(data, offset, 2, "depth offset\n"); |
kgem_debug_print(data, offset, 3, "%dx%d\n", |
((data[3] & 0x0007ffc0) >> 6) + 1, |
((data[3] & 0xfff80000) >> 19) + 1); |
kgem_debug_print(data, offset, 4, "volume depth\n"); |
kgem_debug_print(data, offset, 5, "\n"); |
kgem_debug_print(data, offset, 6, "\n"); |
return len; |
case 0x7a00: |
assert(len == 4 || len == 5); |
switch ((data[1] >> 14) & 0x3) { |
case 0: desc1 = "no write"; break; |
case 1: desc1 = "qword write"; break; |
case 2: desc1 = "PS_DEPTH_COUNT write"; break; |
case 3: desc1 = "TIMESTAMP write"; break; |
} |
kgem_debug_print(data, offset, 0, "PIPE_CONTROL\n"); |
kgem_debug_print(data, offset, 1, |
"%s, %scs stall, %stlb invalidate, " |
"%ssync gfdt, %sdepth stall, %sRC write flush, " |
"%sinst flush, %sTC flush\n", |
desc1, |
data[1] & (1 << 20) ? "" : "no ", |
data[1] & (1 << 18) ? "" : "no ", |
data[1] & (1 << 17) ? "" : "no ", |
data[1] & (1 << 13) ? "" : "no ", |
data[1] & (1 << 12) ? "" : "no ", |
data[1] & (1 << 11) ? "" : "no ", |
data[1] & (1 << 10) ? "" : "no "); |
if (len == 5) { |
kgem_debug_print(data, offset, 2, "destination address\n"); |
kgem_debug_print(data, offset, 3, "immediate dword low\n"); |
kgem_debug_print(data, offset, 4, "immediate dword high\n"); |
} else { |
for (i = 2; i < len; i++) { |
kgem_debug_print(data, offset, i, "\n"); |
} |
} |
return len; |
case 0x7b00: |
assert(len == 6); |
kgem_debug_print(data, offset, 0, |
"3DPRIMITIVE: %s %s\n", |
get_965_prim_type(data[0]), |
(data[0] & (1 << 15)) ? "random" : "sequential"); |
kgem_debug_print(data, offset, 1, "vertex count\n"); |
kgem_debug_print(data, offset, 2, "start vertex\n"); |
kgem_debug_print(data, offset, 3, "instance count\n"); |
kgem_debug_print(data, offset, 4, "start instance\n"); |
kgem_debug_print(data, offset, 5, "index bias\n"); |
primitive_out(kgem, data); |
return len; |
} |
/* For the rest, just dump the bytes */ |
for (i = 0; i < ARRAY_SIZE(opcodes); i++) |
if (op == opcodes[i].opcode) |
break; |
assert(i < ARRAY_SIZE(opcodes)); |
len = 1; |
kgem_debug_print(data, offset, 0, "%s\n", opcodes[i].name); |
if (opcodes[i].max_len > 1) { |
len = (data[0] & 0xff) + 2; |
assert(len >= opcodes[i].min_len && |
len <= opcodes[i].max_len); |
} |
for (i = 1; i < len; i++) |
kgem_debug_print(data, offset, i, "dword %d\n", i); |
return len; |
} |
void kgem_gen5_finish_state(struct kgem *kgem) |
{ |
memset(&state, 0, sizeof(state)); |
} |
/drivers/video/Intel-2D/sna/kgem_debug_gen6.c |
---|
0,0 → 1,1075 |
/* |
* Copyright © 2007-2011 Intel Corporation |
* |
* Permission is hereby granted, free of charge, to any person obtaining a |
* copy of this software and associated documentation files (the "Software"), |
* to deal in the Software without restriction, including without limitation |
* the rights to use, copy, modify, merge, publish, distribute, sublicense, |
* and/or sell copies of the Software, and to permit persons to whom the |
* Software is furnished to do so, subject to the following conditions: |
* |
* The above copyright notice and this permission notice (including the next |
* paragraph) shall be included in all copies or substantial portions of the |
* Software. |
* |
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL |
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, |
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE |
* SOFTWARE. |
* |
* Authors: |
* Eric Anholt <eric@anholt.net> |
* Chris Wilson <chris"chris-wilson.co.uk> |
* |
*/ |
#ifdef HAVE_CONFIG_H |
#include "config.h" |
#endif |
//#include <sys/mman.h> |
#include <assert.h> |
#include "sna.h" |
#include "sna_reg.h" |
#include "gen6_render.h" |
#include "kgem_debug.h" |
static struct state { |
struct vertex_buffer { |
int handle; |
const char *ptr; |
int pitch; |
struct kgem_bo *current; |
} vb[33]; |
struct vertex_elements { |
int buffer; |
int offset; |
bool valid; |
uint32_t type; |
uint8_t swizzle[4]; |
} ve[33]; |
int num_ve; |
struct dynamic_state { |
struct kgem_bo *current; |
void *base, *ptr; |
} dynamic_state; |
} state; |
static void gen6_update_vertex_buffer(struct kgem *kgem, const uint32_t *data) |
{ |
uint32_t reloc = sizeof(uint32_t) * (&data[1] - kgem->batch); |
struct kgem_bo *bo = NULL; |
void *base; |
int i; |
for (i = 0; i < kgem->nreloc; i++) |
if (kgem->reloc[i].offset == reloc) |
break; |
assert(i < kgem->nreloc); |
reloc = kgem->reloc[i].target_handle; |
if (reloc == -1) { |
base = kgem->batch; |
} else { |
list_for_each_entry(bo, &kgem->next_request->buffers, request) |
if (bo->target_handle == reloc) |
break; |
assert(&bo->request != &kgem->next_request->buffers); |
base = kgem_bo_map__debug(kgem, bo); |
} |
base = (char *)base + kgem->reloc[i].delta; |
i = data[0] >> 26; |
state.vb[i].current = bo; |
state.vb[i].ptr = base; |
state.vb[i].pitch = data[0] & 0x7ff; |
} |
static void gen6_update_dynamic_buffer(struct kgem *kgem, const uint32_t offset) |
{ |
uint32_t reloc = sizeof(uint32_t) * offset; |
struct kgem_bo *bo = NULL; |
void *base, *ptr; |
int i; |
if ((kgem->batch[offset] & 1) == 0) |
return; |
for (i = 0; i < kgem->nreloc; i++) |
if (kgem->reloc[i].offset == reloc) |
break; |
if(i < kgem->nreloc) { |
reloc = kgem->reloc[i].target_handle; |
if (reloc == 0) { |
base = kgem->batch; |
} else { |
list_for_each_entry(bo, &kgem->next_request->buffers, request) |
if (bo->handle == reloc) |
break; |
assert(&bo->request != &kgem->next_request->buffers); |
base = kgem_bo_map__debug(kgem, bo); |
} |
ptr = (char *)base + (kgem->reloc[i].delta & ~1); |
} else { |
bo = NULL; |
base = NULL; |
ptr = NULL; |
} |
state.dynamic_state.current = bo; |
state.dynamic_state.base = base; |
state.dynamic_state.ptr = ptr; |
} |
static uint32_t |
get_ve_component(uint32_t data, int component) |
{ |
return (data >> (16 + (3 - component) * 4)) & 0x7; |
} |
static void gen6_update_vertex_elements(struct kgem *kgem, int id, const uint32_t *data) |
{ |
state.ve[id].buffer = data[0] >> 26; |
state.ve[id].valid = !!(data[0] & (1 << 25)); |
state.ve[id].type = (data[0] >> 16) & 0x1ff; |
state.ve[id].offset = data[0] & 0x7ff; |
state.ve[id].swizzle[0] = get_ve_component(data[1], 0); |
state.ve[id].swizzle[1] = get_ve_component(data[1], 1); |
state.ve[id].swizzle[2] = get_ve_component(data[1], 2); |
state.ve[id].swizzle[3] = get_ve_component(data[1], 3); |
} |
static void gen6_update_sf_state(struct kgem *kgem, uint32_t *data) |
{ |
state.num_ve = 1 + ((data[1] >> 22) & 0x3f); |
} |
static void vertices_sint16_out(const struct vertex_elements *ve, const int16_t *v, int max) |
{ |
int c; |
ErrorF("("); |
for (c = 0; c < max; c++) { |
switch (ve->swizzle[c]) { |
case 0: ErrorF("#"); break; |
case 1: ErrorF("%d", v[c]); break; |
case 2: ErrorF("0.0"); break; |
case 3: ErrorF("1.0"); break; |
case 4: ErrorF("0x1"); break; |
case 5: break; |
default: ErrorF("?"); |
} |
if (c < 3) |
ErrorF(", "); |
} |
for (; c < 4; c++) { |
switch (ve->swizzle[c]) { |
case 0: ErrorF("#"); break; |
case 1: ErrorF("1.0"); break; |
case 2: ErrorF("0.0"); break; |
case 3: ErrorF("1.0"); break; |
case 4: ErrorF("0x1"); break; |
case 5: break; |
default: ErrorF("?"); |
} |
if (c < 3) |
ErrorF(", "); |
} |
ErrorF(")"); |
} |
static void vertices_float_out(const struct vertex_elements *ve, const float *f, int max) |
{ |
int c, o; |
ErrorF("("); |
for (c = o = 0; c < 4 && o < max; c++) { |
switch (ve->swizzle[c]) { |
case 0: ErrorF("#"); break; |
case 1: ErrorF("%f", f[o++]); break; |
case 2: ErrorF("0.0"); break; |
case 3: ErrorF("1.0"); break; |
case 4: ErrorF("0x1"); break; |
case 5: break; |
default: ErrorF("?"); |
} |
if (c < 3) |
ErrorF(", "); |
} |
for (; c < 4; c++) { |
switch (ve->swizzle[c]) { |
case 0: ErrorF("#"); break; |
case 1: ErrorF("1.0"); break; |
case 2: ErrorF("0.0"); break; |
case 3: ErrorF("1.0"); break; |
case 4: ErrorF("0x1"); break; |
case 5: break; |
default: ErrorF("?"); |
} |
if (c < 3) |
ErrorF(", "); |
} |
ErrorF(")"); |
} |
static void ve_out(const struct vertex_elements *ve, const void *ptr) |
{ |
switch (ve->type) { |
case GEN6_SURFACEFORMAT_R32_FLOAT: |
vertices_float_out(ve, ptr, 1); |
break; |
case GEN6_SURFACEFORMAT_R32G32_FLOAT: |
vertices_float_out(ve, ptr, 2); |
break; |
case GEN6_SURFACEFORMAT_R32G32B32_FLOAT: |
vertices_float_out(ve, ptr, 3); |
break; |
case GEN6_SURFACEFORMAT_R32G32B32A32_FLOAT: |
vertices_float_out(ve, ptr, 4); |
break; |
case GEN6_SURFACEFORMAT_R16_SINT: |
vertices_sint16_out(ve, ptr, 1); |
break; |
case GEN6_SURFACEFORMAT_R16G16_SINT: |
vertices_sint16_out(ve, ptr, 2); |
break; |
case GEN6_SURFACEFORMAT_R16G16B16A16_SINT: |
vertices_sint16_out(ve, ptr, 4); |
break; |
case GEN6_SURFACEFORMAT_R16_SSCALED: |
vertices_sint16_out(ve, ptr, 1); |
break; |
case GEN6_SURFACEFORMAT_R16G16_SSCALED: |
vertices_sint16_out(ve, ptr, 2); |
break; |
case GEN6_SURFACEFORMAT_R16G16B16A16_SSCALED: |
vertices_sint16_out(ve, ptr, 4); |
break; |
} |
} |
static void indirect_vertex_out(struct kgem *kgem, uint32_t v) |
{ |
int i = 1; |
do { |
const struct vertex_elements *ve = &state.ve[i]; |
const struct vertex_buffer *vb = &state.vb[ve->buffer]; |
const void *ptr = vb->ptr + v * vb->pitch + ve->offset; |
if (ve->valid) |
ve_out(ve, ptr); |
while (++i <= state.num_ve && !state.ve[i].valid) |
; |
if (i <= state.num_ve) |
ErrorF(", "); |
} while (i <= state.num_ve); |
} |
static void primitive_out(struct kgem *kgem, uint32_t *data) |
{ |
int n; |
assert((data[0] & (1<<15)) == 0); /* XXX index buffers */ |
for (n = 0; n < data[1]; n++) { |
int v = data[2] + n; |
ErrorF(" [%d:%d] = ", n, v); |
indirect_vertex_out(kgem, v); |
ErrorF("\n"); |
} |
} |
static void finish_state(struct kgem *kgem) |
{ |
memset(&state, 0, sizeof(state)); |
} |
static void |
state_base_out(uint32_t *data, uint32_t offset, unsigned int index, |
const char *name) |
{ |
if (data[index] & 1) |
kgem_debug_print(data, offset, index, |
"%s state base address 0x%08x\n", |
name, data[index] & ~1); |
else |
kgem_debug_print(data, offset, index, |
"%s state base not updated\n", |
name); |
} |
static void |
state_max_out(uint32_t *data, uint32_t offset, unsigned int index, |
const char *name) |
{ |
if (data[index] == 1) |
kgem_debug_print(data, offset, index, |
"%s state upper bound disabled\n", name); |
else if (data[index] & 1) |
kgem_debug_print(data, offset, index, |
"%s state upper bound 0x%08x\n", |
name, data[index] & ~1); |
else |
kgem_debug_print(data, offset, index, |
"%s state upper bound not updated\n", |
name); |
} |
static const char * |
get_965_surfacetype(unsigned int surfacetype) |
{ |
switch (surfacetype) { |
case 0: return "1D"; |
case 1: return "2D"; |
case 2: return "3D"; |
case 3: return "CUBE"; |
case 4: return "BUFFER"; |
case 7: return "NULL"; |
default: return "unknown"; |
} |
} |
static const char * |
get_965_depthformat(unsigned int depthformat) |
{ |
switch (depthformat) { |
case 0: return "s8_z24float"; |
case 1: return "z32float"; |
case 2: return "z24s8"; |
case 5: return "z16"; |
default: return "unknown"; |
} |
} |
static const char * |
get_965_element_component(uint32_t data, int component) |
{ |
uint32_t component_control = (data >> (16 + (3 - component) * 4)) & 0x7; |
switch (component_control) { |
case 0: |
return "nostore"; |
case 1: |
switch (component) { |
case 0: return "X"; |
case 1: return "Y"; |
case 2: return "Z"; |
case 3: return "W"; |
default: return "fail"; |
} |
case 2: |
return "0.0"; |
case 3: |
return "1.0"; |
case 4: |
return "0x1"; |
case 5: |
return "VID"; |
default: |
return "fail"; |
} |
} |
static const char * |
get_965_prim_type(uint32_t data) |
{ |
uint32_t primtype = (data >> 10) & 0x1f; |
switch (primtype) { |
case 0x01: return "point list"; |
case 0x02: return "line list"; |
case 0x03: return "line strip"; |
case 0x04: return "tri list"; |
case 0x05: return "tri strip"; |
case 0x06: return "tri fan"; |
case 0x07: return "quad list"; |
case 0x08: return "quad strip"; |
case 0x09: return "line list adj"; |
case 0x0a: return "line strip adj"; |
case 0x0b: return "tri list adj"; |
case 0x0c: return "tri strip adj"; |
case 0x0d: return "tri strip reverse"; |
case 0x0e: return "polygon"; |
case 0x0f: return "rect list"; |
case 0x10: return "line loop"; |
case 0x11: return "point list bf"; |
case 0x12: return "line strip cont"; |
case 0x13: return "line strip bf"; |
case 0x14: return "line strip cont bf"; |
case 0x15: return "tri fan no stipple"; |
default: return "fail"; |
} |
} |
struct reloc { |
struct kgem_bo *bo; |
void *base; |
}; |
static void * |
get_reloc(struct kgem *kgem, |
void *base, const uint32_t *reloc, |
struct reloc *r) |
{ |
uint32_t delta = *reloc; |
memset(r, 0, sizeof(*r)); |
if (base == 0) { |
uint32_t handle = sizeof(uint32_t) * (reloc - kgem->batch); |
struct kgem_bo *bo = NULL; |
int i; |
for (i = 0; i < kgem->nreloc; i++) |
if (kgem->reloc[i].offset == handle) |
break; |
assert(i < kgem->nreloc); |
handle = kgem->reloc[i].target_handle; |
delta = kgem->reloc[i].delta; |
if (handle == 0) { |
base = kgem->batch; |
} else { |
list_for_each_entry(bo, &kgem->next_request->buffers, request) |
if (bo->handle == handle) |
break; |
assert(&bo->request != &kgem->next_request->buffers); |
base = kgem_bo_map__debug(kgem, bo); |
r->bo = bo; |
r->base = base; |
} |
} |
return (char *)base + (delta & ~3); |
} |
static const char * |
gen6_filter_to_string(uint32_t filter) |
{ |
switch (filter) { |
default: |
case GEN6_MAPFILTER_NEAREST: return "nearest"; |
case GEN6_MAPFILTER_LINEAR: return "linear"; |
} |
} |
static const char * |
gen6_repeat_to_string(uint32_t repeat) |
{ |
switch (repeat) { |
default: |
case GEN6_TEXCOORDMODE_CLAMP_BORDER: return "border"; |
case GEN6_TEXCOORDMODE_WRAP: return "wrap"; |
case GEN6_TEXCOORDMODE_CLAMP: return "clamp"; |
case GEN6_TEXCOORDMODE_MIRROR: return "mirror"; |
} |
} |
static void |
gen6_decode_sampler_state(struct kgem *kgem, const uint32_t *reloc) |
{ |
const struct gen6_sampler_state *ss; |
struct reloc r; |
const char *min, *mag; |
const char *s_wrap, *t_wrap, *r_wrap; |
ss = get_reloc(kgem, state.dynamic_state.ptr, reloc, &r); |
min = gen6_filter_to_string(ss->ss0.min_filter); |
mag = gen6_filter_to_string(ss->ss0.mag_filter); |
s_wrap = gen6_repeat_to_string(ss->ss1.s_wrap_mode); |
t_wrap = gen6_repeat_to_string(ss->ss1.t_wrap_mode); |
r_wrap = gen6_repeat_to_string(ss->ss1.r_wrap_mode); |
ErrorF(" Sampler 0:\n"); |
ErrorF(" filter: min=%s, mag=%s\n", min, mag); |
ErrorF(" wrap: s=%s, t=%s, r=%s\n", s_wrap, t_wrap, r_wrap); |
ss++; |
min = gen6_filter_to_string(ss->ss0.min_filter); |
mag = gen6_filter_to_string(ss->ss0.mag_filter); |
s_wrap = gen6_repeat_to_string(ss->ss1.s_wrap_mode); |
t_wrap = gen6_repeat_to_string(ss->ss1.t_wrap_mode); |
r_wrap = gen6_repeat_to_string(ss->ss1.r_wrap_mode); |
ErrorF(" Sampler 1:\n"); |
ErrorF(" filter: min=%s, mag=%s\n", min, mag); |
ErrorF(" wrap: s=%s, t=%s, r=%s\n", s_wrap, t_wrap, r_wrap); |
} |
static const char * |
gen6_blend_factor_to_string(uint32_t v) |
{ |
switch (v) { |
#define C(x) case GEN6_BLENDFACTOR_##x: return #x; |
C(ONE); |
C(SRC_COLOR); |
C(SRC_ALPHA); |
C(DST_ALPHA); |
C(DST_COLOR); |
C(SRC_ALPHA_SATURATE); |
C(CONST_COLOR); |
C(CONST_ALPHA); |
C(SRC1_COLOR); |
C(SRC1_ALPHA); |
C(ZERO); |
C(INV_SRC_COLOR); |
C(INV_SRC_ALPHA); |
C(INV_DST_ALPHA); |
C(INV_DST_COLOR); |
C(INV_CONST_COLOR); |
C(INV_CONST_ALPHA); |
C(INV_SRC1_COLOR); |
C(INV_SRC1_ALPHA); |
#undef C |
default: return "???"; |
} |
} |
static const char * |
gen6_blend_function_to_string(uint32_t v) |
{ |
switch (v) { |
#define C(x) case GEN6_BLENDFUNCTION_##x: return #x; |
C(ADD); |
C(SUBTRACT); |
C(REVERSE_SUBTRACT); |
C(MIN); |
C(MAX); |
#undef C |
default: return "???"; |
} |
} |
static float unpack_float(uint32_t dw) |
{ |
union { |
float f; |
uint32_t dw; |
} u; |
u.dw = dw; |
return u.f; |
} |
static void |
gen6_decode_blend(struct kgem *kgem, const uint32_t *reloc) |
{ |
const struct gen6_blend_state *blend; |
struct reloc r; |
const char *dst, *src; |
const char *func; |
blend = get_reloc(kgem, state.dynamic_state.ptr, reloc, &r); |
dst = gen6_blend_factor_to_string(blend->blend0.dest_blend_factor); |
src = gen6_blend_factor_to_string(blend->blend0.source_blend_factor); |
func = gen6_blend_function_to_string(blend->blend0.blend_func); |
ErrorF(" Blend (%s): function %s, src=%s, dst=%s\n", |
blend->blend0.blend_enable ? "enabled" : "disabled", |
func, src, dst); |
} |
int kgem_gen6_decode_3d(struct kgem *kgem, uint32_t offset) |
{ |
static const struct { |
uint32_t opcode; |
int min_len; |
int max_len; |
const char *name; |
} opcodes[] = { |
{ 0x6101, 6, 6, "STATE_BASE_ADDRESS" }, |
{ 0x6102, 2, 2 , "STATE_SIP" }, |
{ 0x6104, 1, 1, "3DSTATE_PIPELINE_SELECT" }, |
{ 0x680b, 1, 1, "3DSTATE_VF_STATISTICS" }, |
{ 0x6904, 1, 1, "3DSTATE_PIPELINE_SELECT" }, |
{ 0x7800, 7, 7, "3DSTATE_PIPELINED_POINTERS" }, |
{ 0x7801, 6, 6, "3DSTATE_BINDING_TABLE_POINTERS" }, |
{ 0x7808, 5, 257, "3DSTATE_VERTEX_BUFFERS" }, |
{ 0x7809, 3, 256, "3DSTATE_VERTEX_ELEMENTS" }, |
{ 0x780a, 3, 3, "3DSTATE_INDEX_BUFFER" }, |
{ 0x780b, 1, 1, "3DSTATE_VF_STATISTICS" }, |
{ 0x7900, 4, 4, "3DSTATE_DRAWING_RECTANGLE" }, |
{ 0x7901, 5, 5, "3DSTATE_CONSTANT_COLOR" }, |
{ 0x7905, 5, 7, "3DSTATE_DEPTH_BUFFER" }, |
{ 0x7906, 2, 2, "3DSTATE_POLY_STIPPLE_OFFSET" }, |
{ 0x7907, 33, 33, "3DSTATE_POLY_STIPPLE_PATTERN" }, |
{ 0x7908, 3, 3, "3DSTATE_LINE_STIPPLE" }, |
{ 0x7909, 2, 2, "3DSTATE_GLOBAL_DEPTH_OFFSET_CLAMP" }, |
{ 0x7909, 2, 2, "3DSTATE_CLEAR_PARAMS" }, |
{ 0x790a, 3, 3, "3DSTATE_AA_LINE_PARAMETERS" }, |
{ 0x790b, 4, 4, "3DSTATE_GS_SVB_INDEX" }, |
{ 0x790d, 3, 3, "3DSTATE_MULTISAMPLE" }, |
{ 0x7910, 2, 2, "3DSTATE_CLEAR_PARAMS" }, |
{ 0x7b00, 6, 6, "3DPRIMITIVE" }, |
{ 0x7802, 4, 4, "3DSTATE_SAMPLER_STATE_POINTERS" }, |
{ 0x7805, 3, 3, "3DSTATE_URB" }, |
{ 0x780d, 4, 4, "3DSTATE_VIEWPORT_STATE_POINTERS" }, |
{ 0x780e, 4, 4, "3DSTATE_CC_STATE_POINTERS" }, |
{ 0x780f, 2, 2, "3DSTATE_SCISSOR_STATE_POINTERS" }, |
{ 0x7810, 6, 6, "3DSTATE_VS_STATE" }, |
{ 0x7811, 7, 7, "3DSTATE_GS_STATE" }, |
{ 0x7812, 4, 4, "3DSTATE_CLIP_STATE" }, |
{ 0x7813, 20, 20, "3DSTATE_SF_STATE" }, |
{ 0x7814, 9, 9, "3DSTATE_WM_STATE" }, |
{ 0x7815, 5, 5, "3DSTATE_CONSTANT_VS_STATE" }, |
{ 0x7816, 5, 5, "3DSTATE_CONSTANT_GS_STATE" }, |
{ 0x7817, 5, 5, "3DSTATE_CONSTANT_WM_STATE" }, |
{ 0x7818, 2, 2, "3DSTATE_SAMPLE_MASK" }, |
}; |
uint32_t *data = kgem->batch + offset; |
uint32_t op; |
unsigned int len; |
int i, j; |
const char *desc1 = NULL; |
len = (data[0] & 0xff) + 2; |
op = (data[0] & 0xffff0000) >> 16; |
switch (op) { |
case 0x6101: |
i = 0; |
kgem_debug_print(data, offset, i++, "STATE_BASE_ADDRESS\n"); |
if (kgem->gen >= 060) { |
assert(len == 10); |
state_base_out(data, offset, i++, "general"); |
state_base_out(data, offset, i++, "surface"); |
state_base_out(data, offset, i++, "dynamic"); |
state_base_out(data, offset, i++, "indirect"); |
state_base_out(data, offset, i++, "instruction"); |
state_max_out(data, offset, i++, "general"); |
state_max_out(data, offset, i++, "dynamic"); |
state_max_out(data, offset, i++, "indirect"); |
state_max_out(data, offset, i++, "instruction"); |
gen6_update_dynamic_buffer(kgem, offset + 3); |
} else if (kgem->gen >= 050) { |
assert(len == 8); |
state_base_out(data, offset, i++, "general"); |
state_base_out(data, offset, i++, "surface"); |
state_base_out(data, offset, i++, "media"); |
state_base_out(data, offset, i++, "instruction"); |
state_max_out(data, offset, i++, "general"); |
state_max_out(data, offset, i++, "media"); |
state_max_out(data, offset, i++, "instruction"); |
} |
return len; |
case 0x7801: |
if (kgem->gen >= 060) { |
assert(len == 4); |
kgem_debug_print(data, offset, 0, |
"3DSTATE_BINDING_TABLE_POINTERS: VS mod %d, " |
"GS mod %d, WM mod %d\n", |
(data[0] & (1 << 8)) != 0, |
(data[0] & (1 << 9)) != 0, |
(data[0] & (1 << 12)) != 0); |
kgem_debug_print(data, offset, 1, "VS binding table\n"); |
kgem_debug_print(data, offset, 2, "GS binding table\n"); |
kgem_debug_print(data, offset, 3, "WM binding table\n"); |
} else if (kgem->gen >= 040) { |
assert(len == 6); |
kgem_debug_print(data, offset, 0, |
"3DSTATE_BINDING_TABLE_POINTERS\n"); |
kgem_debug_print(data, offset, 1, "VS binding table\n"); |
kgem_debug_print(data, offset, 2, "GS binding table\n"); |
kgem_debug_print(data, offset, 3, "CLIP binding table\n"); |
kgem_debug_print(data, offset, 4, "SF binding table\n"); |
kgem_debug_print(data, offset, 5, "WM binding table\n"); |
} |
return len; |
case 0x7802: |
assert(len == 4); |
kgem_debug_print(data, offset, 0, "3DSTATE_SAMPLER_STATE_POINTERS: VS mod %d, " |
"GS mod %d, WM mod %d\n", |
(data[0] & (1 << 8)) != 0, |
(data[0] & (1 << 9)) != 0, |
(data[0] & (1 << 12)) != 0); |
kgem_debug_print(data, offset, 1, "VS sampler state\n"); |
kgem_debug_print(data, offset, 2, "GS sampler state\n"); |
kgem_debug_print(data, offset, 3, "WM sampler state\n"); |
gen6_decode_sampler_state(kgem, &data[3]); |
return len; |
case 0x7808: |
assert((len - 1) % 4 == 0); |
kgem_debug_print(data, offset, 0, "3DSTATE_VERTEX_BUFFERS\n"); |
for (i = 1; i < len;) { |
gen6_update_vertex_buffer(kgem, data + i); |
kgem_debug_print(data, offset, i, "buffer %d: %s, pitch %db\n", |
data[i] >> 26, |
data[i] & (1 << 20) ? "random" : "sequential", |
data[i] & 0x07ff); |
i++; |
kgem_debug_print(data, offset, i++, "buffer address\n"); |
kgem_debug_print(data, offset, i++, "max index\n"); |
kgem_debug_print(data, offset, i++, "mbz\n"); |
} |
return len; |
case 0x7809: |
assert((len + 1) % 2 == 0); |
kgem_debug_print(data, offset, 0, "3DSTATE_VERTEX_ELEMENTS\n"); |
for (i = 1; i < len;) { |
gen6_update_vertex_elements(kgem, (i - 1)/2, data + i); |
kgem_debug_print(data, offset, i, "buffer %d: %svalid, type 0x%04x, " |
"src offset 0x%04x bytes\n", |
data[i] >> 26, |
data[i] & (1 << 25) ? "" : "in", |
(data[i] >> 16) & 0x1ff, |
data[i] & 0x07ff); |
i++; |
kgem_debug_print(data, offset, i, "(%s, %s, %s, %s), " |
"dst offset 0x%02x bytes\n", |
get_965_element_component(data[i], 0), |
get_965_element_component(data[i], 1), |
get_965_element_component(data[i], 2), |
get_965_element_component(data[i], 3), |
(data[i] & 0xff) * 4); |
i++; |
} |
return len; |
case 0x780d: |
assert(len == 4); |
kgem_debug_print(data, offset, 0, "3DSTATE_VIEWPORT_STATE_POINTERS\n"); |
kgem_debug_print(data, offset, 1, "clip\n"); |
kgem_debug_print(data, offset, 2, "sf\n"); |
kgem_debug_print(data, offset, 3, "cc\n"); |
return len; |
case 0x780a: |
assert(len == 3); |
kgem_debug_print(data, offset, 0, "3DSTATE_INDEX_BUFFER\n"); |
kgem_debug_print(data, offset, 1, "beginning buffer address\n"); |
kgem_debug_print(data, offset, 2, "ending buffer address\n"); |
return len; |
case 0x780e: |
assert(len == 4); |
kgem_debug_print(data, offset, 0, "3DSTATE_CC_STATE_POINTERS\n"); |
kgem_debug_print(data, offset, 1, "blend%s\n", |
data[1] & 1 ? " update" : ""); |
if (data[1] & 1) |
gen6_decode_blend(kgem, data+1); |
kgem_debug_print(data, offset, 2, "depth+stencil%s\n", |
data[2] & 1 ? " update" : ""); |
kgem_debug_print(data, offset, 3, "cc%s\n", |
data[3] & 1 ? " update" : ""); |
return len; |
case 0x780f: |
assert(len == 2); |
kgem_debug_print(data, offset, 0, "3DSTATE_SCISSOR_POINTERS\n"); |
kgem_debug_print(data, offset, 1, "scissor rect offset\n"); |
return len; |
case 0x7810: |
assert(len == 6); |
kgem_debug_print(data, offset, 0, "3DSTATE_VS\n"); |
kgem_debug_print(data, offset, 1, "kernel pointer\n"); |
kgem_debug_print(data, offset, 2, "SPF=%d, VME=%d, Sampler Count %d, " |
"Binding table count %d\n", |
(data[2] >> 31) & 1, |
(data[2] >> 30) & 1, |
(data[2] >> 27) & 7, |
(data[2] >> 18) & 0xff); |
kgem_debug_print(data, offset, 3, "scratch offset\n"); |
kgem_debug_print(data, offset, 4, "Dispatch GRF start %d, VUE read length %d, " |
"VUE read offset %d\n", |
(data[4] >> 20) & 0x1f, |
(data[4] >> 11) & 0x3f, |
(data[4] >> 4) & 0x3f); |
kgem_debug_print(data, offset, 5, "Max Threads %d, Vertex Cache %sable, " |
"VS func %sable\n", |
((data[5] >> 25) & 0x7f) + 1, |
(data[5] & (1 << 1)) != 0 ? "dis" : "en", |
(data[5] & 1) != 0 ? "en" : "dis"); |
return len; |
case 0x7811: |
assert(len == 7); |
kgem_debug_print(data, offset, 0, "3DSTATE_GS\n"); |
kgem_debug_print(data, offset, 1, "kernel pointer\n"); |
kgem_debug_print(data, offset, 2, "SPF=%d, VME=%d, Sampler Count %d, " |
"Binding table count %d\n", |
(data[2] >> 31) & 1, |
(data[2] >> 30) & 1, |
(data[2] >> 27) & 7, |
(data[2] >> 18) & 0xff); |
kgem_debug_print(data, offset, 3, "scratch offset\n"); |
kgem_debug_print(data, offset, 4, "Dispatch GRF start %d, VUE read length %d, " |
"VUE read offset %d\n", |
(data[4] & 0xf), |
(data[4] >> 11) & 0x3f, |
(data[4] >> 4) & 0x3f); |
kgem_debug_print(data, offset, 5, "Max Threads %d, Rendering %sable\n", |
((data[5] >> 25) & 0x7f) + 1, |
(data[5] & (1 << 8)) != 0 ? "en" : "dis"); |
kgem_debug_print(data, offset, 6, "Reorder %sable, Discard Adjaceny %sable, " |
"GS %sable\n", |
(data[6] & (1 << 30)) != 0 ? "en" : "dis", |
(data[6] & (1 << 29)) != 0 ? "en" : "dis", |
(data[6] & (1 << 15)) != 0 ? "en" : "dis"); |
return len; |
case 0x7812: |
assert(len == 4); |
kgem_debug_print(data, offset, 0, "3DSTATE_CLIP\n"); |
kgem_debug_print(data, offset, 1, "UserClip distance cull test mask 0x%x\n", |
data[1] & 0xff); |
kgem_debug_print(data, offset, 2, "Clip %sable, API mode %s, Viewport XY test %sable, " |
"Viewport Z test %sable, Guardband test %sable, Clip mode %d, " |
"Perspective Divide %sable, Non-Perspective Barycentric %sable, " |
"Tri Provoking %d, Line Provoking %d, Trifan Provoking %d\n", |
(data[2] & (1 << 31)) != 0 ? "en" : "dis", |
(data[2] & (1 << 30)) != 0 ? "D3D" : "OGL", |
(data[2] & (1 << 28)) != 0 ? "en" : "dis", |
(data[2] & (1 << 27)) != 0 ? "en" : "dis", |
(data[2] & (1 << 26)) != 0 ? "en" : "dis", |
(data[2] >> 13) & 7, |
(data[2] & (1 << 9)) != 0 ? "dis" : "en", |
(data[2] & (1 << 8)) != 0 ? "en" : "dis", |
(data[2] >> 4) & 3, |
(data[2] >> 2) & 3, |
(data[2] & 3)); |
kgem_debug_print(data, offset, 3, "Min PointWidth %d, Max PointWidth %d, " |
"Force Zero RTAIndex %sable, Max VPIndex %d\n", |
(data[3] >> 17) & 0x7ff, |
(data[3] >> 6) & 0x7ff, |
(data[3] & (1 << 5)) != 0 ? "en" : "dis", |
(data[3] & 0xf)); |
return len; |
case 0x7813: |
gen6_update_sf_state(kgem, data); |
assert(len == 20); |
kgem_debug_print(data, offset, 0, "3DSTATE_SF\n"); |
kgem_debug_print(data, offset, 1, "Attrib Out %d, Attrib Swizzle %sable, VUE read length %d, " |
"VUE read offset %d\n", |
(data[1] >> 22) & 0x3f, |
(data[1] & (1 << 21)) != 0 ? "en" : "dis", |
(data[1] >> 11) & 0x1f, |
(data[1] >> 4) & 0x3f); |
kgem_debug_print(data, offset, 2, "Legacy Global DepthBias %sable, FrontFace fill %d, BF fill %d, " |
"VP transform %sable, FrontWinding_%s\n", |
(data[2] & (1 << 11)) != 0 ? "en" : "dis", |
(data[2] >> 5) & 3, |
(data[2] >> 3) & 3, |
(data[2] & (1 << 1)) != 0 ? "en" : "dis", |
(data[2] & 1) != 0 ? "CCW" : "CW"); |
kgem_debug_print(data, offset, 3, "AA %sable, CullMode %d, Scissor %sable, Multisample m ode %d\n", |
(data[3] & (1 << 31)) != 0 ? "en" : "dis", |
(data[3] >> 29) & 3, |
(data[3] & (1 << 11)) != 0 ? "en" : "dis", |
(data[3] >> 8) & 3); |
kgem_debug_print(data, offset, 4, "Last Pixel %sable, SubPixel Precision %d, Use PixelWidth %d\n", |
(data[4] & (1 << 31)) != 0 ? "en" : "dis", |
(data[4] & (1 << 12)) != 0 ? 4 : 8, |
(data[4] & (1 << 11)) != 0); |
kgem_debug_print(data, offset, 5, "Global Depth Offset Constant %f\n", unpack_float(data[5])); |
kgem_debug_print(data, offset, 6, "Global Depth Offset Scale %f\n", unpack_float(data[6])); |
kgem_debug_print(data, offset, 7, "Global Depth Offset Clamp %f\n", unpack_float(data[7])); |
for (i = 0, j = 0; i < 8; i++, j+=2) |
kgem_debug_print(data, offset, i+8, "Attrib %d (Override %s%s%s%s, Const Source %d, Swizzle Select %d, " |
"Source %d); Attrib %d (Override %s%s%s%s, Const Source %d, Swizzle Select %d, Source %d)\n", |
j+1, |
(data[8+i] & (1 << 31)) != 0 ? "W":"", |
(data[8+i] & (1 << 30)) != 0 ? "Z":"", |
(data[8+i] & (1 << 29)) != 0 ? "Y":"", |
(data[8+i] & (1 << 28)) != 0 ? "X":"", |
(data[8+i] >> 25) & 3, (data[8+i] >> 22) & 3, |
(data[8+i] >> 16) & 0x1f, |
j, |
(data[8+i] & (1 << 15)) != 0 ? "W":"", |
(data[8+i] & (1 << 14)) != 0 ? "Z":"", |
(data[8+i] & (1 << 13)) != 0 ? "Y":"", |
(data[8+i] & (1 << 12)) != 0 ? "X":"", |
(data[8+i] >> 9) & 3, (data[8+i] >> 6) & 3, |
(data[8+i] & 0x1f)); |
kgem_debug_print(data, offset, 16, "Point Sprite TexCoord Enable\n"); |
kgem_debug_print(data, offset, 17, "Const Interp Enable\n"); |
kgem_debug_print(data, offset, 18, "Attrib 7-0 WrapShortest Enable\n"); |
kgem_debug_print(data, offset, 19, "Attrib 15-8 WrapShortest Enable\n"); |
return len; |
case 0x7814: |
assert(len == 9); |
kgem_debug_print(data, offset, 0, "3DSTATE_WM\n"); |
kgem_debug_print(data, offset, 1, "kernel start pointer 0\n"); |
kgem_debug_print(data, offset, 2, "SPF=%d, VME=%d, Sampler Count %d, " |
"Binding table count %d\n", |
(data[2] >> 31) & 1, |
(data[2] >> 30) & 1, |
(data[2] >> 27) & 7, |
(data[2] >> 18) & 0xff); |
kgem_debug_print(data, offset, 3, "scratch offset\n"); |
kgem_debug_print(data, offset, 4, "Depth Clear %d, Depth Resolve %d, HiZ Resolve %d, " |
"Dispatch GRF start[0] %d, start[1] %d, start[2] %d\n", |
(data[4] & (1 << 30)) != 0, |
(data[4] & (1 << 28)) != 0, |
(data[4] & (1 << 27)) != 0, |
(data[4] >> 16) & 0x7f, |
(data[4] >> 8) & 0x7f, |
(data[4] & 0x7f)); |
kgem_debug_print(data, offset, 5, "MaxThreads %d, PS KillPixel %d, PS computed Z %d, " |
"PS use sourceZ %d, Thread Dispatch %d, PS use sourceW %d, Dispatch32 %d, " |
"Dispatch16 %d, Dispatch8 %d\n", |
((data[5] >> 25) & 0x7f) + 1, |
(data[5] & (1 << 22)) != 0, |
(data[5] & (1 << 21)) != 0, |
(data[5] & (1 << 20)) != 0, |
(data[5] & (1 << 19)) != 0, |
(data[5] & (1 << 8)) != 0, |
(data[5] & (1 << 2)) != 0, |
(data[5] & (1 << 1)) != 0, |
(data[5] & (1 << 0)) != 0); |
kgem_debug_print(data, offset, 6, "Num SF output %d, Pos XY offset %d, ZW interp mode %d , " |
"Barycentric interp mode 0x%x, Point raster rule %d, Multisample mode %d, " |
"Multisample Dispatch mode %d\n", |
(data[6] >> 20) & 0x3f, |
(data[6] >> 18) & 3, |
(data[6] >> 16) & 3, |
(data[6] >> 10) & 0x3f, |
(data[6] & (1 << 9)) != 0, |
(data[6] >> 1) & 3, |
(data[6] & 1)); |
kgem_debug_print(data, offset, 7, "kernel start pointer 1\n"); |
kgem_debug_print(data, offset, 8, "kernel start pointer 2\n"); |
return len; |
case 0x7900: |
assert(len == 4); |
kgem_debug_print(data, offset, 0, |
"3DSTATE_DRAWING_RECTANGLE\n"); |
kgem_debug_print(data, offset, 1, "top left: %d, %d\n", |
(uint16_t)(data[1] & 0xffff), |
(uint16_t)(data[1] >> 16)); |
kgem_debug_print(data, offset, 2, "bottom right: %d, %d\n", |
(uint16_t)(data[2] & 0xffff), |
(uint16_t)(data[2] >> 16)); |
kgem_debug_print(data, offset, 3, "origin: %d, %d\n", |
(int16_t)(data[3] & 0xffff), |
(int16_t)(data[3] >> 16)); |
return len; |
case 0x7905: |
assert(len == 7); |
kgem_debug_print(data, offset, 0, |
"3DSTATE_DEPTH_BUFFER\n"); |
kgem_debug_print(data, offset, 1, "%s, %s, pitch = %d bytes, %stiled, HiZ %d, Seperate Stencil %d\n", |
get_965_surfacetype(data[1] >> 29), |
get_965_depthformat((data[1] >> 18) & 0x7), |
(data[1] & 0x0001ffff) + 1, |
data[1] & (1 << 27) ? "" : "not ", |
(data[1] & (1 << 22)) != 0, |
(data[1] & (1 << 21)) != 0); |
kgem_debug_print(data, offset, 2, "depth offset\n"); |
kgem_debug_print(data, offset, 3, "%dx%d\n", |
((data[3] & 0x0007ffc0) >> 6) + 1, |
((data[3] & 0xfff80000) >> 19) + 1); |
kgem_debug_print(data, offset, 4, "volume depth\n"); |
kgem_debug_print(data, offset, 5, "\n"); |
kgem_debug_print(data, offset, 6, "\n"); |
return len; |
case 0x7a00: |
assert(len == 4 || len == 5); |
switch ((data[1] >> 14) & 0x3) { |
case 0: desc1 = "no write"; break; |
case 1: desc1 = "qword write"; break; |
case 2: desc1 = "PS_DEPTH_COUNT write"; break; |
case 3: desc1 = "TIMESTAMP write"; break; |
} |
kgem_debug_print(data, offset, 0, "PIPE_CONTROL\n"); |
kgem_debug_print(data, offset, 1, |
"%s, %scs stall, %stlb invalidate, " |
"%ssync gfdt, %sdepth stall, %sRC write flush, " |
"%sinst flush, %sTC flush\n", |
desc1, |
data[1] & (1 << 20) ? "" : "no ", |
data[1] & (1 << 18) ? "" : "no ", |
data[1] & (1 << 17) ? "" : "no ", |
data[1] & (1 << 13) ? "" : "no ", |
data[1] & (1 << 12) ? "" : "no ", |
data[1] & (1 << 11) ? "" : "no ", |
data[1] & (1 << 10) ? "" : "no "); |
if (len == 5) { |
kgem_debug_print(data, offset, 2, "destination address\n"); |
kgem_debug_print(data, offset, 3, "immediate dword low\n"); |
kgem_debug_print(data, offset, 4, "immediate dword high\n"); |
} else { |
for (i = 2; i < len; i++) { |
kgem_debug_print(data, offset, i, "\n"); |
} |
} |
return len; |
case 0x7b00: |
assert(len == 6); |
kgem_debug_print(data, offset, 0, |
"3DPRIMITIVE: %s %s\n", |
get_965_prim_type(data[0]), |
(data[0] & (1 << 15)) ? "random" : "sequential"); |
kgem_debug_print(data, offset, 1, "vertex count\n"); |
kgem_debug_print(data, offset, 2, "start vertex\n"); |
kgem_debug_print(data, offset, 3, "instance count\n"); |
kgem_debug_print(data, offset, 4, "start instance\n"); |
kgem_debug_print(data, offset, 5, "index bias\n"); |
primitive_out(kgem, data); |
return len; |
} |
/* For the rest, just dump the bytes */ |
for (i = 0; i < ARRAY_SIZE(opcodes); i++) |
if (op == opcodes[i].opcode) |
break; |
assert(i < ARRAY_SIZE(opcodes)); |
len = 1; |
kgem_debug_print(data, offset, 0, "%s\n", opcodes[i].name); |
if (opcodes[i].max_len > 1) { |
len = (data[0] & 0xff) + 2; |
assert(len >= opcodes[i].min_len && |
len <= opcodes[i].max_len); |
} |
for (i = 1; i < len; i++) |
kgem_debug_print(data, offset, i, "dword %d\n", i); |
return len; |
} |
void kgem_gen6_finish_state(struct kgem *kgem) |
{ |
finish_state(kgem); |
} |
/drivers/video/Intel-2D/sna/kgem_debug_gen7.c |
---|
0,0 → 1,715 |
/* |
* Copyright © 2007-2011 Intel Corporation |
* |
* Permission is hereby granted, free of charge, to any person obtaining a |
* copy of this software and associated documentation files (the "Software"), |
* to deal in the Software without restriction, including without limitation |
* the rights to use, copy, modify, merge, publish, distribute, sublicense, |
* and/or sell copies of the Software, and to permit persons to whom the |
* Software is furnished to do so, subject to the following conditions: |
* |
* The above copyright notice and this permission notice (including the next |
* paragraph) shall be included in all copies or substantial portions of the |
* Software. |
* |
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL |
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, |
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE |
* SOFTWARE. |
* |
* Authors: |
* Eric Anholt <eric@anholt.net> |
* Chris Wilson <chris"chris-wilson.co.uk> |
* |
*/ |
#ifdef HAVE_CONFIG_H |
#include "config.h" |
#endif |
#include <assert.h> |
#include "sna.h" |
#include "sna_reg.h" |
#include "gen7_render.h" |
#include "kgem_debug.h" |
static struct state { |
struct vertex_buffer { |
int handle; |
void *base; |
const char *ptr; |
int pitch; |
struct kgem_bo *current; |
} vb[33]; |
struct vertex_elements { |
int buffer; |
int offset; |
bool valid; |
uint32_t type; |
uint8_t swizzle[4]; |
} ve[33]; |
int num_ve; |
struct dynamic_state { |
struct kgem_bo *current; |
void *base, *ptr; |
} dynamic_state; |
} state; |
static void gen7_update_vertex_buffer(struct kgem *kgem, const uint32_t *data) |
{ |
uint32_t reloc = sizeof(uint32_t) * (&data[1] - kgem->batch); |
struct kgem_bo *bo = NULL; |
void *base, *ptr; |
int i; |
for (i = 0; i < kgem->nreloc; i++) |
if (kgem->reloc[i].offset == reloc) |
break; |
assert(i < kgem->nreloc); |
reloc = kgem->reloc[i].target_handle; |
if (reloc == 0) { |
base = kgem->batch; |
} else { |
list_for_each_entry(bo, &kgem->next_request->buffers, request) |
if (bo->handle == reloc) |
break; |
assert(&bo->request != &kgem->next_request->buffers); |
base = kgem_bo_map__debug(kgem, bo); |
} |
ptr = (char *)base + kgem->reloc[i].delta; |
i = data[0] >> 26; |
state.vb[i].current = bo; |
state.vb[i].base = base; |
state.vb[i].ptr = ptr; |
state.vb[i].pitch = data[0] & 0x7ff; |
} |
static void gen7_update_dynamic_buffer(struct kgem *kgem, const uint32_t offset) |
{ |
uint32_t reloc = sizeof(uint32_t) * offset; |
struct kgem_bo *bo = NULL; |
void *base, *ptr; |
int i; |
if ((kgem->batch[offset] & 1) == 0) |
return; |
for (i = 0; i < kgem->nreloc; i++) |
if (kgem->reloc[i].offset == reloc) |
break; |
if(i < kgem->nreloc) { |
reloc = kgem->reloc[i].target_handle; |
if (reloc == 0) { |
base = kgem->batch; |
} else { |
list_for_each_entry(bo, &kgem->next_request->buffers, request) |
if (bo->handle == reloc) |
break; |
assert(&bo->request != &kgem->next_request->buffers); |
base = kgem_bo_map__debug(kgem, bo); |
} |
ptr = (char *)base + (kgem->reloc[i].delta & ~1); |
} else { |
bo = NULL; |
base = NULL; |
ptr = NULL; |
} |
state.dynamic_state.current = bo; |
state.dynamic_state.base = base; |
state.dynamic_state.ptr = ptr; |
} |
static uint32_t |
get_ve_component(uint32_t data, int component) |
{ |
return (data >> (16 + (3 - component) * 4)) & 0x7; |
} |
static void gen7_update_vertex_elements(struct kgem *kgem, int id, const uint32_t *data) |
{ |
state.ve[id].buffer = data[0] >> 26; |
state.ve[id].valid = !!(data[0] & (1 << 25)); |
state.ve[id].type = (data[0] >> 16) & 0x1ff; |
state.ve[id].offset = data[0] & 0x7ff; |
state.ve[id].swizzle[0] = get_ve_component(data[1], 0); |
state.ve[id].swizzle[1] = get_ve_component(data[1], 1); |
state.ve[id].swizzle[2] = get_ve_component(data[1], 2); |
state.ve[id].swizzle[3] = get_ve_component(data[1], 3); |
} |
static void gen7_update_sf_state(struct kgem *kgem, uint32_t *data) |
{ |
state.num_ve = 1 + ((data[1] >> 22) & 0x3f); |
} |
static void vertices_sint16_out(const struct vertex_elements *ve, const int16_t *v, int max) |
{ |
int c; |
ErrorF("("); |
for (c = 0; c < max; c++) { |
switch (ve->swizzle[c]) { |
case 0: ErrorF("#"); break; |
case 1: ErrorF("%d", v[c]); break; |
case 2: ErrorF("0.0"); break; |
case 3: ErrorF("1.0"); break; |
case 4: ErrorF("0x1"); break; |
case 5: break; |
default: ErrorF("?"); |
} |
if (c < 3) |
ErrorF(", "); |
} |
for (; c < 4; c++) { |
switch (ve->swizzle[c]) { |
case 0: ErrorF("#"); break; |
case 1: ErrorF("1.0"); break; |
case 2: ErrorF("0.0"); break; |
case 3: ErrorF("1.0"); break; |
case 4: ErrorF("0x1"); break; |
case 5: break; |
default: ErrorF("?"); |
} |
if (c < 3) |
ErrorF(", "); |
} |
ErrorF(")"); |
} |
static void vertices_float_out(const struct vertex_elements *ve, const float *f, int max) |
{ |
int c, o; |
ErrorF("("); |
for (c = o = 0; c < 4 && o < max; c++) { |
switch (ve->swizzle[c]) { |
case 0: ErrorF("#"); break; |
case 1: ErrorF("%f", f[o++]); break; |
case 2: ErrorF("0.0"); break; |
case 3: ErrorF("1.0"); break; |
case 4: ErrorF("0x1"); break; |
case 5: break; |
default: ErrorF("?"); |
} |
if (c < 3) |
ErrorF(", "); |
} |
for (; c < 4; c++) { |
switch (ve->swizzle[c]) { |
case 0: ErrorF("#"); break; |
case 1: ErrorF("1.0"); break; |
case 2: ErrorF("0.0"); break; |
case 3: ErrorF("1.0"); break; |
case 4: ErrorF("0x1"); break; |
case 5: break; |
default: ErrorF("?"); |
} |
if (c < 3) |
ErrorF(", "); |
} |
ErrorF(")"); |
} |
static void ve_out(const struct vertex_elements *ve, const void *ptr) |
{ |
switch (ve->type) { |
case GEN7_SURFACEFORMAT_R32_FLOAT: |
vertices_float_out(ve, ptr, 1); |
break; |
case GEN7_SURFACEFORMAT_R32G32_FLOAT: |
vertices_float_out(ve, ptr, 2); |
break; |
case GEN7_SURFACEFORMAT_R32G32B32_FLOAT: |
vertices_float_out(ve, ptr, 3); |
break; |
case GEN7_SURFACEFORMAT_R32G32B32A32_FLOAT: |
vertices_float_out(ve, ptr, 4); |
break; |
case GEN7_SURFACEFORMAT_R16_SINT: |
vertices_sint16_out(ve, ptr, 1); |
break; |
case GEN7_SURFACEFORMAT_R16G16_SINT: |
vertices_sint16_out(ve, ptr, 2); |
break; |
case GEN7_SURFACEFORMAT_R16G16B16A16_SINT: |
vertices_sint16_out(ve, ptr, 4); |
break; |
case GEN7_SURFACEFORMAT_R16_SSCALED: |
vertices_sint16_out(ve, ptr, 1); |
break; |
case GEN7_SURFACEFORMAT_R16G16_SSCALED: |
vertices_sint16_out(ve, ptr, 2); |
break; |
case GEN7_SURFACEFORMAT_R16G16B16A16_SSCALED: |
vertices_sint16_out(ve, ptr, 4); |
break; |
} |
} |
static void indirect_vertex_out(struct kgem *kgem, uint32_t v) |
{ |
int i = 1; |
do { |
const struct vertex_elements *ve = &state.ve[i]; |
const struct vertex_buffer *vb = &state.vb[ve->buffer]; |
const void *ptr = vb->ptr + v * vb->pitch + ve->offset; |
if (!ve->valid) |
continue; |
ve_out(ve, ptr); |
while (++i <= state.num_ve && !state.ve[i].valid) |
; |
if (i <= state.num_ve) |
ErrorF(", "); |
} while (i <= state.num_ve); |
} |
static void primitive_out(struct kgem *kgem, uint32_t *data) |
{ |
int n; |
assert((data[0] & (1<<15)) == 0); /* XXX index buffers */ |
for (n = 0; n < data[2]; n++) { |
int v = data[3] + n; |
ErrorF(" [%d:%d] = ", n, v); |
indirect_vertex_out(kgem, v); |
ErrorF("\n"); |
} |
} |
static void finish_state(struct kgem *kgem) |
{ |
memset(&state, 0, sizeof(state)); |
} |
static void |
state_base_out(uint32_t *data, uint32_t offset, unsigned int index, |
const char *name) |
{ |
if (data[index] & 1) |
kgem_debug_print(data, offset, index, |
"%s state base address 0x%08x\n", |
name, data[index] & ~1); |
else |
kgem_debug_print(data, offset, index, |
"%s state base not updated\n", |
name); |
} |
static void |
state_max_out(uint32_t *data, uint32_t offset, unsigned int index, |
const char *name) |
{ |
if (data[index] == 1) |
kgem_debug_print(data, offset, index, |
"%s state upper bound disabled\n", name); |
else if (data[index] & 1) |
kgem_debug_print(data, offset, index, |
"%s state upper bound 0x%08x\n", |
name, data[index] & ~1); |
else |
kgem_debug_print(data, offset, index, |
"%s state upper bound not updated\n", |
name); |
} |
static const char * |
get_965_surfacetype(unsigned int surfacetype) |
{ |
switch (surfacetype) { |
case 0: return "1D"; |
case 1: return "2D"; |
case 2: return "3D"; |
case 3: return "CUBE"; |
case 4: return "BUFFER"; |
case 7: return "NULL"; |
default: return "unknown"; |
} |
} |
static const char * |
get_965_depthformat(unsigned int depthformat) |
{ |
switch (depthformat) { |
case 0: return "s8_z24float"; |
case 1: return "z32float"; |
case 2: return "z24s8"; |
case 5: return "z16"; |
default: return "unknown"; |
} |
} |
static const char * |
get_element_component(uint32_t data, int component) |
{ |
uint32_t component_control = (data >> (16 + (3 - component) * 4)) & 0x7; |
switch (component_control) { |
case 0: |
return "nostore"; |
case 1: |
switch (component) { |
case 0: return "X"; |
case 1: return "Y"; |
case 2: return "Z"; |
case 3: return "W"; |
default: return "fail"; |
} |
case 2: |
return "0.0"; |
case 3: |
return "1.0"; |
case 4: |
return "0x1"; |
case 5: |
return "VID"; |
default: |
return "fail"; |
} |
} |
static const char * |
get_prim_type(uint32_t data) |
{ |
uint32_t primtype = data & 0x1f; |
switch (primtype) { |
case 0x01: return "point list"; |
case 0x02: return "line list"; |
case 0x03: return "line strip"; |
case 0x04: return "tri list"; |
case 0x05: return "tri strip"; |
case 0x06: return "tri fan"; |
case 0x07: return "quad list"; |
case 0x08: return "quad strip"; |
case 0x09: return "line list adj"; |
case 0x0a: return "line strip adj"; |
case 0x0b: return "tri list adj"; |
case 0x0c: return "tri strip adj"; |
case 0x0d: return "tri strip reverse"; |
case 0x0e: return "polygon"; |
case 0x0f: return "rect list"; |
case 0x10: return "line loop"; |
case 0x11: return "point list bf"; |
case 0x12: return "line strip cont"; |
case 0x13: return "line strip bf"; |
case 0x14: return "line strip cont bf"; |
case 0x15: return "tri fan no stipple"; |
default: return "fail"; |
} |
} |
struct reloc { |
struct kgem_bo *bo; |
void *base; |
}; |
static void * |
get_reloc(struct kgem *kgem, |
void *base, const uint32_t *reloc, |
struct reloc *r) |
{ |
uint32_t delta = *reloc; |
memset(r, 0, sizeof(*r)); |
if (base == 0) { |
uint32_t handle = sizeof(uint32_t) * (reloc - kgem->batch); |
struct kgem_bo *bo = NULL; |
int i; |
for (i = 0; i < kgem->nreloc; i++) |
if (kgem->reloc[i].offset == handle) |
break; |
assert(i < kgem->nreloc); |
handle = kgem->reloc[i].target_handle; |
delta = kgem->reloc[i].delta; |
if (handle == 0) { |
base = kgem->batch; |
} else { |
list_for_each_entry(bo, &kgem->next_request->buffers, request) |
if (bo->handle == handle) |
break; |
assert(&bo->request != &kgem->next_request->buffers); |
base = kgem_bo_map__debug(kgem, bo); |
r->bo = bo; |
r->base = base; |
} |
} |
return (char *)base + (delta & ~3); |
} |
static const char * |
gen7_filter_to_string(uint32_t filter) |
{ |
switch (filter) { |
default: |
case GEN7_MAPFILTER_NEAREST: return "nearest"; |
case GEN7_MAPFILTER_LINEAR: return "linear"; |
} |
} |
static const char * |
gen7_repeat_to_string(uint32_t repeat) |
{ |
switch (repeat) { |
default: |
case GEN7_TEXCOORDMODE_CLAMP_BORDER: return "border"; |
case GEN7_TEXCOORDMODE_WRAP: return "wrap"; |
case GEN7_TEXCOORDMODE_CLAMP: return "clamp"; |
case GEN7_TEXCOORDMODE_MIRROR: return "mirror"; |
} |
} |
static void |
gen7_decode_sampler_state(struct kgem *kgem, const uint32_t *reloc) |
{ |
const struct gen7_sampler_state *ss; |
struct reloc r; |
const char *min, *mag; |
const char *s_wrap, *t_wrap, *r_wrap; |
ss = get_reloc(kgem, state.dynamic_state.ptr, reloc, &r); |
min = gen7_filter_to_string(ss->ss0.min_filter); |
mag = gen7_filter_to_string(ss->ss0.mag_filter); |
s_wrap = gen7_repeat_to_string(ss->ss3.s_wrap_mode); |
t_wrap = gen7_repeat_to_string(ss->ss3.t_wrap_mode); |
r_wrap = gen7_repeat_to_string(ss->ss3.r_wrap_mode); |
ErrorF(" Sampler 0:\n"); |
ErrorF(" filter: min=%s, mag=%s\n", min, mag); |
ErrorF(" wrap: s=%s, t=%s, r=%s\n", s_wrap, t_wrap, r_wrap); |
ss++; |
min = gen7_filter_to_string(ss->ss0.min_filter); |
mag = gen7_filter_to_string(ss->ss0.mag_filter); |
s_wrap = gen7_repeat_to_string(ss->ss3.s_wrap_mode); |
t_wrap = gen7_repeat_to_string(ss->ss3.t_wrap_mode); |
r_wrap = gen7_repeat_to_string(ss->ss3.r_wrap_mode); |
ErrorF(" Sampler 1:\n"); |
ErrorF(" filter: min=%s, mag=%s\n", min, mag); |
ErrorF(" wrap: s=%s, t=%s, r=%s\n", s_wrap, t_wrap, r_wrap); |
} |
static const char * |
gen7_blend_factor_to_string(uint32_t v) |
{ |
switch (v) { |
#define C(x) case GEN7_BLENDFACTOR_##x: return #x; |
C(ONE); |
C(SRC_COLOR); |
C(SRC_ALPHA); |
C(DST_ALPHA); |
C(DST_COLOR); |
C(SRC_ALPHA_SATURATE); |
C(CONST_COLOR); |
C(CONST_ALPHA); |
C(SRC1_COLOR); |
C(SRC1_ALPHA); |
C(ZERO); |
C(INV_SRC_COLOR); |
C(INV_SRC_ALPHA); |
C(INV_DST_ALPHA); |
C(INV_DST_COLOR); |
C(INV_CONST_COLOR); |
C(INV_CONST_ALPHA); |
C(INV_SRC1_COLOR); |
C(INV_SRC1_ALPHA); |
#undef C |
default: return "???"; |
} |
} |
static const char * |
gen7_blend_function_to_string(uint32_t v) |
{ |
switch (v) { |
#define C(x) case GEN7_BLENDFUNCTION_##x: return #x; |
C(ADD); |
C(SUBTRACT); |
C(REVERSE_SUBTRACT); |
C(MIN); |
C(MAX); |
#undef C |
default: return "???"; |
} |
} |
static void |
gen7_decode_blend(struct kgem *kgem, const uint32_t *reloc) |
{ |
const struct gen7_blend_state *blend; |
struct reloc r; |
const char *dst, *src; |
const char *func; |
blend = get_reloc(kgem, state.dynamic_state.ptr, reloc, &r); |
dst = gen7_blend_factor_to_string(blend->blend0.dest_blend_factor); |
src = gen7_blend_factor_to_string(blend->blend0.source_blend_factor); |
func = gen7_blend_function_to_string(blend->blend0.blend_func); |
ErrorF(" Blend (%s): function %s, src=%s, dst=%s\n", |
blend->blend0.blend_enable ? "enabled" : "disabled", |
func, src, dst); |
} |
int kgem_gen7_decode_3d(struct kgem *kgem, uint32_t offset) |
{ |
static const struct { |
uint32_t opcode; |
int min_len; |
int max_len; |
const char *name; |
} opcodes[] = { |
{ 0x6101, 6, 6, "STATE_BASE_ADDRESS" }, |
{ 0x6102, 2, 2 , "STATE_SIP" }, |
{ 0x6104, 1, 1, "3DSTATE_PIPELINE_SELECT" }, |
{ 0x780a, 3, 3, "3DSTATE_INDEX_BUFFER" }, |
{ 0x7900, 4, 4, "3DSTATE_DRAWING_RECTANGLE" }, |
}; |
uint32_t *data = kgem->batch + offset; |
uint32_t op; |
unsigned int len; |
int i; |
const char *name; |
len = (data[0] & 0xff) + 2; |
op = (data[0] & 0xffff0000) >> 16; |
switch (op) { |
case 0x6101: |
i = 0; |
kgem_debug_print(data, offset, i++, "STATE_BASE_ADDRESS\n"); |
assert(len == 10); |
state_base_out(data, offset, i++, "general"); |
state_base_out(data, offset, i++, "surface"); |
state_base_out(data, offset, i++, "dynamic"); |
state_base_out(data, offset, i++, "indirect"); |
state_base_out(data, offset, i++, "instruction"); |
state_max_out(data, offset, i++, "general"); |
state_max_out(data, offset, i++, "dynamic"); |
state_max_out(data, offset, i++, "indirect"); |
state_max_out(data, offset, i++, "instruction"); |
gen7_update_dynamic_buffer(kgem, offset + 3); |
return len; |
case 0x7808: |
assert((len - 1) % 4 == 0); |
kgem_debug_print(data, offset, 0, "3DSTATE_VERTEX_BUFFERS\n"); |
for (i = 1; i < len;) { |
gen7_update_vertex_buffer(kgem, data + i); |
kgem_debug_print(data, offset, i, "buffer %d: %s, pitch %db\n", |
data[i] >> 26, |
data[i] & (1 << 20) ? "random" : "sequential", |
data[i] & 0x07ff); |
i++; |
kgem_debug_print(data, offset, i++, "buffer address\n"); |
kgem_debug_print(data, offset, i++, "max index\n"); |
kgem_debug_print(data, offset, i++, "mbz\n"); |
} |
return len; |
case 0x7809: |
assert((len + 1) % 2 == 0); |
kgem_debug_print(data, offset, 0, "3DSTATE_VERTEX_ELEMENTS\n"); |
for (i = 1; i < len;) { |
gen7_update_vertex_elements(kgem, (i - 1)/2, data + i); |
kgem_debug_print(data, offset, i, "buffer %d: %svalid, type 0x%04x, " |
"src offset 0x%04x bytes\n", |
data[i] >> 26, |
data[i] & (1 << 25) ? "" : "in", |
(data[i] >> 16) & 0x1ff, |
data[i] & 0x07ff); |
i++; |
kgem_debug_print(data, offset, i, "(%s, %s, %s, %s), " |
"dst offset 0x%02x bytes\n", |
get_element_component(data[i], 0), |
get_element_component(data[i], 1), |
get_element_component(data[i], 2), |
get_element_component(data[i], 3), |
(data[i] & 0xff) * 4); |
i++; |
} |
return len; |
case 0x780a: |
assert(len == 3); |
kgem_debug_print(data, offset, 0, "3DSTATE_INDEX_BUFFER\n"); |
kgem_debug_print(data, offset, 1, "beginning buffer address\n"); |
kgem_debug_print(data, offset, 2, "ending buffer address\n"); |
return len; |
case 0x7b00: |
assert(len == 7); |
kgem_debug_print(data, offset, 0, "3DPRIMITIVE\n"); |
kgem_debug_print(data, offset, 1, "type %s, %s\n", |
get_prim_type(data[1]), |
(data[1] & (1 << 15)) ? "random" : "sequential"); |
kgem_debug_print(data, offset, 2, "vertex count\n"); |
kgem_debug_print(data, offset, 3, "start vertex\n"); |
kgem_debug_print(data, offset, 4, "instance count\n"); |
kgem_debug_print(data, offset, 5, "start instance\n"); |
kgem_debug_print(data, offset, 6, "index bias\n"); |
primitive_out(kgem, data); |
return len; |
} |
/* For the rest, just dump the bytes */ |
name = NULL; |
for (i = 0; i < ARRAY_SIZE(opcodes); i++) |
if (op == opcodes[i].opcode) { |
name = opcodes[i].name; |
break; |
} |
len = (data[0] & 0xff) + 2; |
if (name == NULL) { |
kgem_debug_print(data, offset, 0, "unknown\n"); |
} else { |
kgem_debug_print(data, offset, 0, "%s\n", opcodes[i].name); |
if (opcodes[i].max_len > 1) { |
assert(len >= opcodes[i].min_len && |
len <= opcodes[i].max_len); |
} |
} |
for (i = 1; i < len; i++) |
kgem_debug_print(data, offset, i, "dword %d\n", i); |
return len; |
} |
void kgem_gen7_finish_state(struct kgem *kgem) |
{ |
finish_state(kgem); |
} |
/drivers/video/Intel-2D/sna/pciaccess.h |
---|
0,0 → 1,537 |
/* |
* (C) Copyright IBM Corporation 2006 |
* Copyright 2009 Red Hat, Inc. |
* All Rights Reserved. |
* |
* Permission is hereby granted, free of charge, to any person obtaining a |
* copy of this software and associated documentation files (the "Software"), |
* to deal in the Software without restriction, including without limitation |
* on the rights to use, copy, modify, merge, publish, distribute, sub |
* license, and/or sell copies of the Software, and to permit persons to whom |
* the Software is furnished to do so, subject to the following conditions: |
* |
* The above copyright notice and this permission notice (including the next |
* paragraph) shall be included in all copies or substantial portions of the |
* Software. |
* |
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
* FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL |
* IBM AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING |
* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER |
* DEALINGS IN THE SOFTWARE. |
*/ |
/* |
* Copyright (c) 2007 Paulo R. Zanoni, Tiago Vignatti |
* |
* Permission is hereby granted, free of charge, to any person |
* obtaining a copy of this software and associated documentation |
* files (the "Software"), to deal in the Software without |
* restriction, including without limitation the rights to use, |
* copy, modify, merge, publish, distribute, sublicense, and/or sell |
* copies of the Software, and to permit persons to whom the |
* Software is furnished to do so, subject to the following |
* conditions: |
* |
* The above copyright notice and this permission notice shall be |
* included in all copies or substantial portions of the Software. |
* |
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, |
* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES |
* OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND |
* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT |
* HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, |
* WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING |
* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR |
* OTHER DEALINGS IN THE SOFTWARE. |
* |
*/ |
/** |
* \file pciaccess.h |
* |
* \author Ian Romanick <idr@us.ibm.com> |
*/ |
#ifndef PCIACCESS_H |
#define PCIACCESS_H |
#include <inttypes.h> |
#if __GNUC__ >= 3 |
#define __deprecated __attribute__((deprecated)) |
#else |
#define __deprecated |
#endif |
typedef uint64_t pciaddr_t; |
struct pci_device; |
struct pci_device_iterator; |
struct pci_id_match; |
struct pci_slot_match; |
#ifdef __cplusplus |
extern "C" { |
#endif |
int pci_device_has_kernel_driver(struct pci_device *dev); |
int pci_device_is_boot_vga(struct pci_device *dev); |
int pci_device_read_rom(struct pci_device *dev, void *buffer); |
int __deprecated pci_device_map_region(struct pci_device *dev, |
unsigned region, int write_enable); |
int __deprecated pci_device_unmap_region(struct pci_device *dev, |
unsigned region); |
int pci_device_map_range(struct pci_device *dev, pciaddr_t base, |
pciaddr_t size, unsigned map_flags, void **addr); |
int pci_device_unmap_range(struct pci_device *dev, void *memory, |
pciaddr_t size); |
int __deprecated pci_device_map_memory_range(struct pci_device *dev, |
pciaddr_t base, pciaddr_t size, int write_enable, void **addr); |
int __deprecated pci_device_unmap_memory_range(struct pci_device *dev, |
void *memory, pciaddr_t size); |
int pci_device_probe(struct pci_device *dev); |
const struct pci_agp_info *pci_device_get_agp_info(struct pci_device *dev); |
const struct pci_bridge_info *pci_device_get_bridge_info( |
struct pci_device *dev); |
const struct pci_pcmcia_bridge_info *pci_device_get_pcmcia_bridge_info( |
struct pci_device *dev); |
int pci_device_get_bridge_buses(struct pci_device *dev, int *primary_bus, |
int *secondary_bus, int *subordinate_bus); |
int pci_system_init(void); |
void pci_system_init_dev_mem(int fd); |
void pci_system_cleanup(void); |
struct pci_device_iterator *pci_slot_match_iterator_create( |
const struct pci_slot_match *match); |
struct pci_device_iterator *pci_id_match_iterator_create( |
const struct pci_id_match *match); |
void pci_iterator_destroy(struct pci_device_iterator *iter); |
struct pci_device *pci_device_next(struct pci_device_iterator *iter); |
struct pci_device *pci_device_find_by_slot(uint32_t domain, uint32_t bus, |
uint32_t dev, uint32_t func); |
struct pci_device *pci_device_get_parent_bridge(struct pci_device *dev); |
void pci_get_strings(const struct pci_id_match *m, |
const char **device_name, const char **vendor_name, |
const char **subdevice_name, const char **subvendor_name); |
const char *pci_device_get_device_name(const struct pci_device *dev); |
const char *pci_device_get_subdevice_name(const struct pci_device *dev); |
const char *pci_device_get_vendor_name(const struct pci_device *dev); |
const char *pci_device_get_subvendor_name(const struct pci_device *dev); |
void pci_device_enable(struct pci_device *dev); |
int pci_device_cfg_read (struct pci_device *dev, void *data, |
pciaddr_t offset, pciaddr_t size, pciaddr_t *bytes_read); |
int pci_device_cfg_read_u8 (struct pci_device *dev, uint8_t *data, |
pciaddr_t offset); |
int pci_device_cfg_read_u16(struct pci_device *dev, uint16_t *data, |
pciaddr_t offset); |
int pci_device_cfg_read_u32(struct pci_device *dev, uint32_t *data, |
pciaddr_t offset); |
int pci_device_cfg_write (struct pci_device *dev, const void *data, |
pciaddr_t offset, pciaddr_t size, pciaddr_t *bytes_written); |
int pci_device_cfg_write_u8 (struct pci_device *dev, uint8_t data, |
pciaddr_t offset); |
int pci_device_cfg_write_u16(struct pci_device *dev, uint16_t data, |
pciaddr_t offset); |
int pci_device_cfg_write_u32(struct pci_device *dev, uint32_t data, |
pciaddr_t offset); |
int pci_device_cfg_write_bits(struct pci_device *dev, uint32_t mask, |
uint32_t data, pciaddr_t offset); |
#ifdef __cplusplus |
} |
#endif |
/** |
* \name Mapping flags passed to \c pci_device_map_range |
*/ |
/*@{*/ |
#define PCI_DEV_MAP_FLAG_WRITABLE (1U<<0) |
#define PCI_DEV_MAP_FLAG_WRITE_COMBINE (1U<<1) |
#define PCI_DEV_MAP_FLAG_CACHABLE (1U<<2) |
/*@}*/ |
#define PCI_MATCH_ANY (~0) |
/** |
* Compare two PCI ID values (either vendor or device). This is used |
* internally to compare the fields of \c pci_id_match to the fields of |
* \c pci_device. |
*/ |
#define PCI_ID_COMPARE(a, b) \ |
(((a) == PCI_MATCH_ANY) || ((a) == (b))) |
/** |
*/ |
struct pci_id_match { |
/** |
* \name Device / vendor matching controls |
* |
* Control the search based on the device, vendor, subdevice, or subvendor |
* IDs. Setting any of these fields to \c PCI_MATCH_ANY will cause the |
* field to not be used in the comparison. |
*/ |
/*@{*/ |
uint32_t vendor_id; |
uint32_t device_id; |
uint32_t subvendor_id; |
uint32_t subdevice_id; |
/*@}*/ |
/** |
* \name Device class matching controls |
* |
*/ |
/*@{*/ |
uint32_t device_class; |
uint32_t device_class_mask; |
/*@}*/ |
intptr_t match_data; |
}; |
/** |
*/ |
struct pci_slot_match { |
/** |
* \name Device slot matching controls |
* |
* Control the search based on the domain, bus, slot, and function of |
* the device. Setting any of these fields to \c PCI_MATCH_ANY will cause |
* the field to not be used in the comparison. |
*/ |
/*@{*/ |
uint32_t domain; |
uint32_t bus; |
uint32_t dev; |
uint32_t func; |
/*@}*/ |
intptr_t match_data; |
}; |
/** |
* BAR descriptor for a PCI device. |
*/ |
struct pci_mem_region { |
/** |
* When the region is mapped, this is the pointer to the memory. |
* |
* This field is \b only set when the deprecated \c pci_device_map_region |
* interface is used. Use \c pci_device_map_range instead. |
* |
* \deprecated |
*/ |
void *memory; |
/** |
* Base physical address of the region within its bus / domain. |
* |
* \warning |
* This address is really only useful to other devices in the same |
* domain. It's probably \b not the address applications will ever |
* use. |
* |
* \warning |
* Most (all?) platform back-ends leave this field unset. |
*/ |
pciaddr_t bus_addr; |
/** |
* Base physical address of the region from the CPU's point of view. |
* |
* This address is typically passed to \c pci_device_map_range to create |
* a mapping of the region to the CPU's virtual address space. |
*/ |
pciaddr_t base_addr; |
/** |
* Size, in bytes, of the region. |
*/ |
pciaddr_t size; |
/** |
* Is the region I/O ports or memory? |
*/ |
unsigned is_IO:1; |
/** |
* Is the memory region prefetchable? |
* |
* \note |
* This can only be set if \c is_IO is not set. |
*/ |
unsigned is_prefetchable:1; |
/** |
* Is the memory at a 64-bit address? |
* |
* \note |
* This can only be set if \c is_IO is not set. |
*/ |
unsigned is_64:1; |
}; |
/** |
* PCI device. |
* |
* Contains all of the information about a particular PCI device. |
*/ |
struct pci_device { |
/** |
* \name Device bus identification. |
* |
* Complete bus identification, including domain, of the device. On |
* platforms that do not support PCI domains (e.g., 32-bit x86 hardware), |
* the domain will always be zero. |
*/ |
/*@{*/ |
uint16_t domain; |
uint8_t bus; |
uint8_t dev; |
uint8_t func; |
/*@}*/ |
/** |
* \name Vendor / device ID |
* |
* The vendor ID, device ID, and sub-IDs for the device. |
*/ |
/*@{*/ |
uint16_t vendor_id; |
uint16_t device_id; |
uint16_t subvendor_id; |
uint16_t subdevice_id; |
/*@}*/ |
/** |
* Device's class, subclass, and programming interface packed into a |
* single 32-bit value. The class is at bits [23:16], subclass is at |
* bits [15:8], and programming interface is at [7:0]. |
*/ |
uint32_t device_class; |
/** |
* Device revision number, as read from the configuration header. |
*/ |
uint8_t revision; |
/** |
* BAR descriptors for the device. |
*/ |
struct pci_mem_region regions[6]; |
/** |
* Size, in bytes, of the device's expansion ROM. |
*/ |
pciaddr_t rom_size; |
/** |
* IRQ associated with the device. If there is no IRQ, this value will |
* be -1. |
*/ |
int irq; |
/** |
* Storage for user data. Users of the library can store arbitrary |
* data in this pointer. The library will not use it for any purpose. |
* It is the user's responsability to free this memory before destroying |
* the \c pci_device structure. |
*/ |
intptr_t user_data; |
/** |
* Used by the VGA arbiter. Type of resource decoded by the device and |
* the file descriptor (/dev/vga_arbiter). */ |
int vgaarb_rsrc; |
}; |
/** |
* Description of the AGP capability of the device. |
* |
* \sa pci_device_get_agp_info |
*/ |
struct pci_agp_info { |
/** |
* Offset of the AGP registers in the devices configuration register |
* space. This is generally used so that the offset of the AGP command |
* register can be determined. |
*/ |
unsigned config_offset; |
/** |
* \name AGP major / minor version. |
*/ |
/*@{*/ |
uint8_t major_version; |
uint8_t minor_version; |
/*@}*/ |
/** |
* Logical OR of the supported AGP rates. For example, a value of 0x07 |
* means that the device can support 1x, 2x, and 4x. A value of 0x0c |
* means that the device can support 8x and 4x. |
*/ |
uint8_t rates; |
unsigned int fast_writes:1; /**< Are fast-writes supported? */ |
unsigned int addr64:1; |
unsigned int htrans:1; |
unsigned int gart64:1; |
unsigned int coherent:1; |
unsigned int sideband:1; /**< Is side-band addressing supported? */ |
unsigned int isochronus:1; |
uint8_t async_req_size; |
uint8_t calibration_cycle_timing; |
uint8_t max_requests; |
}; |
/** |
* Description of a PCI-to-PCI bridge device. |
* |
* \sa pci_device_get_bridge_info |
*/ |
struct pci_bridge_info { |
uint8_t primary_bus; |
uint8_t secondary_bus; |
uint8_t subordinate_bus; |
uint8_t secondary_latency_timer; |
uint8_t io_type; |
uint8_t mem_type; |
uint8_t prefetch_mem_type; |
uint16_t secondary_status; |
uint16_t bridge_control; |
uint32_t io_base; |
uint32_t io_limit; |
uint32_t mem_base; |
uint32_t mem_limit; |
uint64_t prefetch_mem_base; |
uint64_t prefetch_mem_limit; |
}; |
/** |
* Description of a PCI-to-PCMCIA bridge device. |
* |
* \sa pci_device_get_pcmcia_bridge_info |
*/ |
struct pci_pcmcia_bridge_info { |
uint8_t primary_bus; |
uint8_t card_bus; |
uint8_t subordinate_bus; |
uint8_t cardbus_latency_timer; |
uint16_t secondary_status; |
uint16_t bridge_control; |
struct { |
uint32_t base; |
uint32_t limit; |
} io[2]; |
struct { |
uint32_t base; |
uint32_t limit; |
} mem[2]; |
}; |
/** |
* VGA Arbiter definitions, functions and related. |
*/ |
/* Legacy VGA regions */ |
#define VGA_ARB_RSRC_NONE 0x00 |
#define VGA_ARB_RSRC_LEGACY_IO 0x01 |
#define VGA_ARB_RSRC_LEGACY_MEM 0x02 |
/* Non-legacy access */ |
#define VGA_ARB_RSRC_NORMAL_IO 0x04 |
#define VGA_ARB_RSRC_NORMAL_MEM 0x08 |
int pci_device_vgaarb_init (void); |
void pci_device_vgaarb_fini (void); |
int pci_device_vgaarb_set_target (struct pci_device *dev); |
/* use the targetted device */ |
int pci_device_vgaarb_decodes (int new_vga_rsrc); |
int pci_device_vgaarb_lock (void); |
int pci_device_vgaarb_trylock (void); |
int pci_device_vgaarb_unlock (void); |
/* return the current device count + resource decodes for the device */ |
int pci_device_vgaarb_get_info (struct pci_device *dev, int *vga_count, int *rsrc_decodes); |
/* |
* I/O space access. |
*/ |
struct pci_io_handle; |
struct pci_io_handle *pci_device_open_io(struct pci_device *dev, pciaddr_t base, |
pciaddr_t size); |
struct pci_io_handle *pci_legacy_open_io(struct pci_device *dev, pciaddr_t base, |
pciaddr_t size); |
void pci_device_close_io(struct pci_device *dev, struct pci_io_handle *handle); |
uint32_t pci_io_read32(struct pci_io_handle *handle, uint32_t reg); |
uint16_t pci_io_read16(struct pci_io_handle *handle, uint32_t reg); |
uint8_t pci_io_read8(struct pci_io_handle *handle, uint32_t reg); |
void pci_io_write32(struct pci_io_handle *handle, uint32_t reg, uint32_t data); |
void pci_io_write16(struct pci_io_handle *handle, uint32_t reg, uint16_t data); |
void pci_io_write8(struct pci_io_handle *handle, uint32_t reg, uint8_t data); |
/* |
* Legacy memory access |
*/ |
int pci_device_map_legacy(struct pci_device *dev, pciaddr_t base, |
pciaddr_t size, unsigned map_flags, void **addr); |
int pci_device_unmap_legacy(struct pci_device *dev, void *addr, pciaddr_t size); |
#endif /* PCIACCESS_H */ |
/drivers/video/Intel-2D/sna/render/exa_sf.g4b |
---|
0,0 → 1,15 |
{ 0x00400031, 0x20c01fbd, 0x0069002c, 0x01110001 }, |
{ 0x00400001, 0x206003be, 0x00690060, 0x00000000 }, |
{ 0x00400040, 0x20e077bd, 0x00690080, 0x006940a0 }, |
{ 0x00400041, 0x202077be, 0x006900e0, 0x000000c0 }, |
{ 0x00400040, 0x20e077bd, 0x006900a0, 0x00694060 }, |
{ 0x00400041, 0x204077be, 0x006900e0, 0x000000c8 }, |
{ 0x00600031, 0x20001fbc, 0x008d0000, 0x8640c800 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
/drivers/video/Intel-2D/sna/render/exa_sf.g5b |
---|
0,0 → 1,7 |
{ 0x00400031, 0x20c01fbd, 0x1069002c, 0x02100001 }, |
{ 0x00400001, 0x206003be, 0x00690060, 0x00000000 }, |
{ 0x00400040, 0x20e077bd, 0x00690080, 0x006940a0 }, |
{ 0x00400041, 0x202077be, 0x006900e0, 0x000000c0 }, |
{ 0x00400040, 0x20e077bd, 0x006900a0, 0x00694060 }, |
{ 0x00400041, 0x204077be, 0x006900e0, 0x000000c8 }, |
{ 0x00600031, 0x20001fbc, 0x648d0000, 0x8808c800 }, |
/drivers/video/Intel-2D/sna/render/exa_sf_mask.g4b |
---|
0,0 → 1,15 |
{ 0x00400031, 0x20c01fbd, 0x0069002c, 0x01110001 }, |
{ 0x00600001, 0x206003be, 0x008d0060, 0x00000000 }, |
{ 0x00600040, 0x20e077bd, 0x008d0080, 0x008d40a0 }, |
{ 0x00600041, 0x202077be, 0x008d00e0, 0x000000c0 }, |
{ 0x00600040, 0x20e077bd, 0x008d00a0, 0x008d4060 }, |
{ 0x00600041, 0x204077be, 0x008d00e0, 0x000000c8 }, |
{ 0x00600031, 0x20001fbc, 0x008d0000, 0x8640c800 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
/drivers/video/Intel-2D/sna/render/exa_sf_mask.g5b |
---|
0,0 → 1,7 |
{ 0x00400031, 0x20c01fbd, 0x1069002c, 0x02100001 }, |
{ 0x00600001, 0x206003be, 0x008d0060, 0x00000000 }, |
{ 0x00600040, 0x20e077bd, 0x008d0080, 0x008d40a0 }, |
{ 0x00600041, 0x202077be, 0x008d00e0, 0x000000c0 }, |
{ 0x00600040, 0x20e077bd, 0x008d00a0, 0x008d4060 }, |
{ 0x00600041, 0x204077be, 0x008d00e0, 0x000000c8 }, |
{ 0x00600031, 0x20001fbc, 0x648d0000, 0x8808c800 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_ca.g4b |
---|
0,0 → 1,4 |
{ 0x00802041, 0x21c077bd, 0x008d01c0, 0x008d02c0 }, |
{ 0x00802041, 0x220077bd, 0x008d0200, 0x008d0300 }, |
{ 0x00802041, 0x224077bd, 0x008d0240, 0x008d0340 }, |
{ 0x00802041, 0x228077bd, 0x008d0280, 0x008d0380 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_ca.g5b |
---|
0,0 → 1,4 |
{ 0x00802041, 0x21c077bd, 0x008d01c0, 0x008d02c0 }, |
{ 0x00802041, 0x220077bd, 0x008d0200, 0x008d0300 }, |
{ 0x00802041, 0x224077bd, 0x008d0240, 0x008d0340 }, |
{ 0x00802041, 0x228077bd, 0x008d0280, 0x008d0380 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_ca.g6b |
---|
0,0 → 1,4 |
{ 0x00800041, 0x21c077bd, 0x008d01c0, 0x008d02c0 }, |
{ 0x00800041, 0x220077bd, 0x008d0200, 0x008d0300 }, |
{ 0x00800041, 0x224077bd, 0x008d0240, 0x008d0340 }, |
{ 0x00800041, 0x228077bd, 0x008d0280, 0x008d0380 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_ca_srcalpha.g4b |
---|
0,0 → 1,4 |
{ 0x00802041, 0x21c077bd, 0x008d02c0, 0x008d0280 }, |
{ 0x00802041, 0x220077bd, 0x008d0300, 0x008d0280 }, |
{ 0x00802041, 0x224077bd, 0x008d0340, 0x008d0280 }, |
{ 0x00802041, 0x228077bd, 0x008d0380, 0x008d0280 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_ca_srcalpha.g5b |
---|
0,0 → 1,4 |
{ 0x00802041, 0x21c077bd, 0x008d02c0, 0x008d0280 }, |
{ 0x00802041, 0x220077bd, 0x008d0300, 0x008d0280 }, |
{ 0x00802041, 0x224077bd, 0x008d0340, 0x008d0280 }, |
{ 0x00802041, 0x228077bd, 0x008d0380, 0x008d0280 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_ca_srcalpha.g6b |
---|
0,0 → 1,4 |
{ 0x00800041, 0x21c077bd, 0x008d02c0, 0x008d0280 }, |
{ 0x00800041, 0x220077bd, 0x008d0300, 0x008d0280 }, |
{ 0x00800041, 0x224077bd, 0x008d0340, 0x008d0280 }, |
{ 0x00800041, 0x228077bd, 0x008d0380, 0x008d0280 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_mask_affine.g4b |
---|
0,0 → 1,8 |
{ 0x00802041, 0x23c077bd, 0x008d0100, 0x000000a0 }, |
{ 0x00802041, 0x238077bd, 0x008d0140, 0x000000a4 }, |
{ 0x00802040, 0x23c077bd, 0x008d03c0, 0x008d0380 }, |
{ 0x00802040, 0x210077be, 0x008d03c0, 0x000000ac }, |
{ 0x00802041, 0x23c077bd, 0x008d0100, 0x000000b0 }, |
{ 0x00802041, 0x238077bd, 0x008d0140, 0x000000b4 }, |
{ 0x00802040, 0x23c077bd, 0x008d03c0, 0x008d0380 }, |
{ 0x00802040, 0x214077be, 0x008d03c0, 0x000000bc }, |
/drivers/video/Intel-2D/sna/render/exa_wm_mask_affine.g5b |
---|
0,0 → 1,4 |
{ 0x00802059, 0x200077bc, 0x000000a0, 0x008d0100 }, |
{ 0x00802048, 0x210077be, 0x000000a4, 0x008d0140 }, |
{ 0x00802059, 0x200077bc, 0x000000b0, 0x008d0100 }, |
{ 0x00802048, 0x214077be, 0x000000b4, 0x008d0140 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_mask_affine.g6b |
---|
0,0 → 1,4 |
{ 0x0060005a, 0x210077be, 0x00000100, 0x008d0040 }, |
{ 0x0060005a, 0x212077be, 0x00000100, 0x008d0080 }, |
{ 0x0060005a, 0x214077be, 0x00000110, 0x008d0040 }, |
{ 0x0060005a, 0x216077be, 0x00000110, 0x008d0080 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_mask_affine.g7b |
---|
0,0 → 1,4 |
{ 0x0060005a, 0x290077bd, 0x00000100, 0x008d0040 }, |
{ 0x0060005a, 0x292077bd, 0x00000100, 0x008d0080 }, |
{ 0x0060005a, 0x294077bd, 0x00000110, 0x008d0040 }, |
{ 0x0060005a, 0x296077bd, 0x00000110, 0x008d0080 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_mask_projective.g4b |
---|
0,0 → 1,16 |
{ 0x00802041, 0x23c077bd, 0x008d0100, 0x000000c0 }, |
{ 0x00802041, 0x238077bd, 0x008d0140, 0x000000c4 }, |
{ 0x00802040, 0x23c077bd, 0x008d03c0, 0x008d0380 }, |
{ 0x00802040, 0x23c077bd, 0x008d03c0, 0x000000cc }, |
{ 0x00600031, 0x21801fbd, 0x008d03c0, 0x01110001 }, |
{ 0x00600031, 0x21a01fbd, 0x008d03e0, 0x01110001 }, |
{ 0x00802041, 0x23c077bd, 0x008d0100, 0x000000a0 }, |
{ 0x00802041, 0x238077bd, 0x008d0140, 0x000000a4 }, |
{ 0x00802040, 0x23c077bd, 0x008d03c0, 0x008d0380 }, |
{ 0x00802040, 0x23c077bd, 0x008d03c0, 0x000000ac }, |
{ 0x00802041, 0x210077be, 0x008d03c0, 0x008d0180 }, |
{ 0x00802041, 0x23c077bd, 0x008d0100, 0x000000b0 }, |
{ 0x00802041, 0x238077bd, 0x008d0140, 0x000000b4 }, |
{ 0x00802040, 0x23c077bd, 0x008d03c0, 0x008d0380 }, |
{ 0x00802040, 0x23c077bd, 0x008d03c0, 0x000000bc }, |
{ 0x00802041, 0x214077be, 0x008d03c0, 0x008d0180 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_mask_projective.g5b |
---|
0,0 → 1,16 |
{ 0x00802041, 0x23c077bd, 0x008d0100, 0x000000c0 }, |
{ 0x00802041, 0x238077bd, 0x008d0140, 0x000000c4 }, |
{ 0x00802040, 0x23c077bd, 0x008d03c0, 0x008d0380 }, |
{ 0x00802040, 0x23c077bd, 0x008d03c0, 0x000000cc }, |
{ 0x00600031, 0x21801fbd, 0x108d03c0, 0x02100001 }, |
{ 0x00600031, 0x21a01fbd, 0x108d03e0, 0x02100001 }, |
{ 0x00802041, 0x23c077bd, 0x008d0100, 0x000000a0 }, |
{ 0x00802041, 0x238077bd, 0x008d0140, 0x000000a4 }, |
{ 0x00802040, 0x23c077bd, 0x008d03c0, 0x008d0380 }, |
{ 0x00802040, 0x23c077bd, 0x008d03c0, 0x000000ac }, |
{ 0x00802041, 0x210077be, 0x008d03c0, 0x008d0180 }, |
{ 0x00802041, 0x23c077bd, 0x008d0100, 0x000000b0 }, |
{ 0x00802041, 0x238077bd, 0x008d0140, 0x000000b4 }, |
{ 0x00802040, 0x23c077bd, 0x008d03c0, 0x008d0380 }, |
{ 0x00802040, 0x23c077bd, 0x008d03c0, 0x000000bc }, |
{ 0x00802041, 0x214077be, 0x008d03c0, 0x008d0180 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_mask_projective.g6b |
---|
0,0 → 1,12 |
{ 0x0060005a, 0x23c077bd, 0x00000120, 0x008d0040 }, |
{ 0x0060005a, 0x23e077bd, 0x00000120, 0x008d0080 }, |
{ 0x01600038, 0x218003bd, 0x008d03c0, 0x00000000 }, |
{ 0x01600038, 0x21a003bd, 0x008d03e0, 0x00000000 }, |
{ 0x0060005a, 0x23c077bd, 0x00000100, 0x008d0040 }, |
{ 0x0060005a, 0x23e077bd, 0x00000100, 0x008d0080 }, |
{ 0x00600041, 0x210077be, 0x008d03c0, 0x008d0180 }, |
{ 0x00600041, 0x212077be, 0x008d03e0, 0x008d01a0 }, |
{ 0x0060005a, 0x23c077bd, 0x00000110, 0x008d0040 }, |
{ 0x0060005a, 0x23e077bd, 0x00000110, 0x008d0080 }, |
{ 0x00600041, 0x214077be, 0x008d03c0, 0x008d0180 }, |
{ 0x00600041, 0x216077be, 0x008d03e0, 0x008d01a0 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_mask_projective.g7b |
---|
0,0 → 1,12 |
{ 0x0060005a, 0x23c077bd, 0x00000120, 0x008d0040 }, |
{ 0x0060005a, 0x23e077bd, 0x00000120, 0x008d0080 }, |
{ 0x01600038, 0x218003bd, 0x008d03c0, 0x00000000 }, |
{ 0x01600038, 0x21a003bd, 0x008d03e0, 0x00000000 }, |
{ 0x0060005a, 0x23c077bd, 0x00000100, 0x008d0040 }, |
{ 0x0060005a, 0x23e077bd, 0x00000100, 0x008d0080 }, |
{ 0x00600041, 0x290077bd, 0x008d03c0, 0x008d0180 }, |
{ 0x00600041, 0x292077bd, 0x008d03e0, 0x008d01a0 }, |
{ 0x0060005a, 0x23c077bd, 0x00000110, 0x008d0040 }, |
{ 0x0060005a, 0x23e077bd, 0x00000110, 0x008d0080 }, |
{ 0x00600041, 0x294077bd, 0x008d03c0, 0x008d0180 }, |
{ 0x00600041, 0x296077bd, 0x008d03e0, 0x008d01a0 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_mask_sample_a.g4b |
---|
0,0 → 1,3 |
{ 0x00000201, 0x20080061, 0x00000000, 0x00007000 }, |
{ 0x00600001, 0x20e00022, 0x008d0000, 0x00000000 }, |
{ 0x07800031, 0x23801c09, 0x00000000, 0x02520102 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_mask_sample_a.g5b |
---|
0,0 → 1,3 |
{ 0x00000201, 0x20080061, 0x00000000, 0x00007000 }, |
{ 0x00600001, 0x20e00022, 0x008d0000, 0x00000000 }, |
{ 0x07800031, 0x23801c09, 0x20000000, 0x0a2a0102 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_mask_sample_a.g6b |
---|
0,0 → 1,3 |
{ 0x00000201, 0x20080061, 0x00000000, 0x00007000 }, |
{ 0x00600001, 0x20e00022, 0x008d0000, 0x00000000 }, |
{ 0x02800031, 0x23801cc9, 0x000000e0, 0x0a2a0102 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_mask_sample_a.g7b |
---|
0,0 → 1,3 |
{ 0x00000201, 0x20080061, 0x00000000, 0x00007000 }, |
{ 0x00600001, 0x28e00021, 0x008d0000, 0x00000000 }, |
{ 0x02800031, 0x23801ca9, 0x000008e0, 0x0a2c0102 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_mask_sample_argb.g4b |
---|
0,0 → 1,3 |
{ 0x00000201, 0x20080061, 0x00000000, 0x00000000 }, |
{ 0x00600001, 0x20e00022, 0x008d0000, 0x00000000 }, |
{ 0x07800031, 0x22c01c09, 0x00000000, 0x02580102 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_mask_sample_argb.g5b |
---|
0,0 → 1,3 |
{ 0x00000201, 0x20080061, 0x00000000, 0x00000000 }, |
{ 0x00600001, 0x20e00022, 0x008d0000, 0x00000000 }, |
{ 0x07800031, 0x22c01c09, 0x20000000, 0x0a8a0102 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_mask_sample_argb.g6b |
---|
0,0 → 1,3 |
{ 0x00000201, 0x20080061, 0x00000000, 0x00000000 }, |
{ 0x00600001, 0x20e00022, 0x008d0000, 0x00000000 }, |
{ 0x02800031, 0x22c01cc9, 0x000000e0, 0x0a8a0102 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_mask_sample_argb.g7b |
---|
0,0 → 1,3 |
{ 0x00000201, 0x20080061, 0x00000000, 0x00000000 }, |
{ 0x00600001, 0x28e00021, 0x008d0000, 0x00000000 }, |
{ 0x02800031, 0x22c01ca9, 0x000008e0, 0x0a8c0102 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_noca.g4b |
---|
0,0 → 1,4 |
{ 0x00802041, 0x21c077bd, 0x008d01c0, 0x008d0380 }, |
{ 0x00802041, 0x220077bd, 0x008d0200, 0x008d0380 }, |
{ 0x00802041, 0x224077bd, 0x008d0240, 0x008d0380 }, |
{ 0x00802041, 0x228077bd, 0x008d0280, 0x008d0380 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_noca.g5b |
---|
0,0 → 1,4 |
{ 0x00802041, 0x21c077bd, 0x008d01c0, 0x008d0380 }, |
{ 0x00802041, 0x220077bd, 0x008d0200, 0x008d0380 }, |
{ 0x00802041, 0x224077bd, 0x008d0240, 0x008d0380 }, |
{ 0x00802041, 0x228077bd, 0x008d0280, 0x008d0380 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_noca.g6b |
---|
0,0 → 1,4 |
{ 0x00800041, 0x21c077bd, 0x008d01c0, 0x008d0380 }, |
{ 0x00800041, 0x220077bd, 0x008d0200, 0x008d0380 }, |
{ 0x00800041, 0x224077bd, 0x008d0240, 0x008d0380 }, |
{ 0x00800041, 0x228077bd, 0x008d0280, 0x008d0380 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_src_affine.g4b |
---|
0,0 → 1,8 |
{ 0x00802041, 0x23c077bd, 0x008d0100, 0x00000060 }, |
{ 0x00802041, 0x238077bd, 0x008d0140, 0x00000064 }, |
{ 0x00802040, 0x23c077bd, 0x008d03c0, 0x008d0380 }, |
{ 0x00802040, 0x204077be, 0x008d03c0, 0x0000006c }, |
{ 0x00802041, 0x23c077bd, 0x008d0100, 0x00000070 }, |
{ 0x00802041, 0x238077bd, 0x008d0140, 0x00000074 }, |
{ 0x00802040, 0x23c077bd, 0x008d03c0, 0x008d0380 }, |
{ 0x00802040, 0x208077be, 0x008d03c0, 0x0000007c }, |
/drivers/video/Intel-2D/sna/render/exa_wm_src_affine.g5b |
---|
0,0 → 1,4 |
{ 0x00802059, 0x200077bc, 0x00000060, 0x008d0100 }, |
{ 0x00802048, 0x204077be, 0x00000064, 0x008d0140 }, |
{ 0x00802059, 0x200077bc, 0x00000070, 0x008d0100 }, |
{ 0x00802048, 0x208077be, 0x00000074, 0x008d0140 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_src_affine.g6b |
---|
0,0 → 1,4 |
{ 0x0060005a, 0x204077be, 0x000000c0, 0x008d0040 }, |
{ 0x0060005a, 0x206077be, 0x000000c0, 0x008d0080 }, |
{ 0x0060005a, 0x208077be, 0x000000d0, 0x008d0040 }, |
{ 0x0060005a, 0x20a077be, 0x000000d0, 0x008d0080 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_src_affine.g7b |
---|
0,0 → 1,4 |
{ 0x0060005a, 0x284077bd, 0x000000c0, 0x008d0040 }, |
{ 0x0060005a, 0x286077bd, 0x000000c0, 0x008d0080 }, |
{ 0x0060005a, 0x288077bd, 0x000000d0, 0x008d0040 }, |
{ 0x0060005a, 0x28a077bd, 0x000000d0, 0x008d0080 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_src_projective.g4b |
---|
0,0 → 1,16 |
{ 0x00802041, 0x23c077bd, 0x008d0100, 0x00000080 }, |
{ 0x00802041, 0x238077bd, 0x008d0140, 0x00000084 }, |
{ 0x00802040, 0x23c077bd, 0x008d03c0, 0x008d0380 }, |
{ 0x00802040, 0x23c077bd, 0x008d03c0, 0x0000008c }, |
{ 0x00600031, 0x21801fbd, 0x008d03c0, 0x01110001 }, |
{ 0x00600031, 0x21a01fbd, 0x008d03e0, 0x01110001 }, |
{ 0x00802041, 0x23c077bd, 0x008d0100, 0x00000060 }, |
{ 0x00802041, 0x238077bd, 0x008d0140, 0x00000064 }, |
{ 0x00802040, 0x23c077bd, 0x008d03c0, 0x008d0380 }, |
{ 0x00802040, 0x23c077bd, 0x008d03c0, 0x0000006c }, |
{ 0x00802041, 0x204077be, 0x008d03c0, 0x008d0180 }, |
{ 0x00802041, 0x23c077bd, 0x008d0100, 0x00000070 }, |
{ 0x00802041, 0x238077bd, 0x008d0140, 0x00000074 }, |
{ 0x00802040, 0x23c077bd, 0x008d03c0, 0x008d0380 }, |
{ 0x00802040, 0x23c077bd, 0x008d03c0, 0x0000007c }, |
{ 0x00802041, 0x208077be, 0x008d03c0, 0x008d0180 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_src_projective.g5b |
---|
0,0 → 1,16 |
{ 0x00802041, 0x23c077bd, 0x008d0100, 0x00000080 }, |
{ 0x00802041, 0x238077bd, 0x008d0140, 0x00000084 }, |
{ 0x00802040, 0x23c077bd, 0x008d03c0, 0x008d0380 }, |
{ 0x00802040, 0x23c077bd, 0x008d03c0, 0x0000008c }, |
{ 0x00600031, 0x21801fbd, 0x108d03c0, 0x02100001 }, |
{ 0x00600031, 0x21a01fbd, 0x108d03e0, 0x02100001 }, |
{ 0x00802041, 0x23c077bd, 0x008d0100, 0x00000060 }, |
{ 0x00802041, 0x238077bd, 0x008d0140, 0x00000064 }, |
{ 0x00802040, 0x23c077bd, 0x008d03c0, 0x008d0380 }, |
{ 0x00802040, 0x23c077bd, 0x008d03c0, 0x0000006c }, |
{ 0x00802041, 0x204077be, 0x008d03c0, 0x008d0180 }, |
{ 0x00802041, 0x23c077bd, 0x008d0100, 0x00000070 }, |
{ 0x00802041, 0x238077bd, 0x008d0140, 0x00000074 }, |
{ 0x00802040, 0x23c077bd, 0x008d03c0, 0x008d0380 }, |
{ 0x00802040, 0x23c077bd, 0x008d03c0, 0x0000007c }, |
{ 0x00802041, 0x208077be, 0x008d03c0, 0x008d0180 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_src_projective.g6b |
---|
0,0 → 1,12 |
{ 0x0060005a, 0x23c077bd, 0x000000e0, 0x008d0040 }, |
{ 0x0060005a, 0x23e077bd, 0x000000e0, 0x008d0080 }, |
{ 0x01600038, 0x218003bd, 0x008d03c0, 0x00000000 }, |
{ 0x01600038, 0x21a003bd, 0x008d03e0, 0x00000000 }, |
{ 0x0060005a, 0x23c077bd, 0x000000c0, 0x008d0040 }, |
{ 0x0060005a, 0x23e077bd, 0x000000c0, 0x008d0080 }, |
{ 0x00600041, 0x204077be, 0x008d03c0, 0x008d0180 }, |
{ 0x00600041, 0x206077be, 0x008d03e0, 0x008d01a0 }, |
{ 0x0060005a, 0x23c077bd, 0x000000d0, 0x008d0040 }, |
{ 0x0060005a, 0x23e077bd, 0x000000d0, 0x008d0080 }, |
{ 0x00600041, 0x208077be, 0x008d03c0, 0x008d0180 }, |
{ 0x00600041, 0x20a077be, 0x008d03e0, 0x008d01a0 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_src_projective.g7b |
---|
0,0 → 1,12 |
{ 0x0060005a, 0x23c077bd, 0x000000e0, 0x008d0040 }, |
{ 0x0060005a, 0x23e077bd, 0x000000e0, 0x008d0080 }, |
{ 0x01600038, 0x218003bd, 0x008d03c0, 0x00000000 }, |
{ 0x01600038, 0x21a003bd, 0x008d03e0, 0x00000000 }, |
{ 0x0060005a, 0x23c077bd, 0x000000c0, 0x008d0040 }, |
{ 0x0060005a, 0x23e077bd, 0x000000c0, 0x008d0080 }, |
{ 0x00600041, 0x284077bd, 0x008d03c0, 0x008d0180 }, |
{ 0x00600041, 0x286077bd, 0x008d03e0, 0x008d01a0 }, |
{ 0x0060005a, 0x23c077bd, 0x000000d0, 0x008d0040 }, |
{ 0x0060005a, 0x23e077bd, 0x000000d0, 0x008d0080 }, |
{ 0x00600041, 0x288077bd, 0x008d03c0, 0x008d0180 }, |
{ 0x00600041, 0x28a077bd, 0x008d03e0, 0x008d01a0 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_src_sample_a.g4b |
---|
0,0 → 1,3 |
{ 0x00000201, 0x20080061, 0x00000000, 0x00007000 }, |
{ 0x00600001, 0x20200022, 0x008d0000, 0x00000000 }, |
{ 0x01800031, 0x22801c09, 0x00000000, 0x02520001 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_src_sample_a.g5b |
---|
0,0 → 1,3 |
{ 0x00000201, 0x20080061, 0x00000000, 0x00007000 }, |
{ 0x00600001, 0x20200022, 0x008d0000, 0x00000000 }, |
{ 0x01800031, 0x22801c09, 0x20000000, 0x0a2a0001 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_src_sample_a.g6b |
---|
0,0 → 1,3 |
{ 0x00000201, 0x20080061, 0x00000000, 0x00007000 }, |
{ 0x00600001, 0x20200022, 0x008d0000, 0x00000000 }, |
{ 0x02800031, 0x22801cc9, 0x00000020, 0x0a2a0001 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_src_sample_a.g7b |
---|
0,0 → 1,3 |
{ 0x00000201, 0x20080061, 0x00000000, 0x00007000 }, |
{ 0x00600001, 0x28200021, 0x008d0000, 0x00000000 }, |
{ 0x02800031, 0x22801ca9, 0x00000820, 0x0a2c0001 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_src_sample_argb.g4b |
---|
0,0 → 1,3 |
{ 0x00000201, 0x20080061, 0x00000000, 0x00000000 }, |
{ 0x00600001, 0x20200022, 0x008d0000, 0x00000000 }, |
{ 0x01800031, 0x21c01c09, 0x00000000, 0x02580001 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_src_sample_argb.g5b |
---|
0,0 → 1,2 |
{ 0x00000201, 0x20080061, 0x00000000, 0x00000000 }, |
{ 0x01800031, 0x21c01d29, 0x208d0000, 0x0a8a0001 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_src_sample_argb.g6b |
---|
0,0 → 1,3 |
{ 0x00000201, 0x20080061, 0x00000000, 0x00000000 }, |
{ 0x00600001, 0x20200022, 0x008d0000, 0x00000000 }, |
{ 0x02800031, 0x21c01cc9, 0x00000020, 0x0a8a0001 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_src_sample_argb.g7b |
---|
0,0 → 1,3 |
{ 0x00000201, 0x20080061, 0x00000000, 0x00000000 }, |
{ 0x00600001, 0x28200021, 0x008d0000, 0x00000000 }, |
{ 0x02800031, 0x21c01ca9, 0x00000820, 0x0a8c0001 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_src_sample_planar.g4b |
---|
0,0 → 1,5 |
{ 0x00000201, 0x20080061, 0x00000000, 0x0000e000 }, |
{ 0x00600001, 0x20200022, 0x008d0000, 0x00000000 }, |
{ 0x01800031, 0x22001c09, 0x00000000, 0x02520001 }, |
{ 0x01800031, 0x21c01c09, 0x00000000, 0x02520003 }, |
{ 0x01800031, 0x22401c09, 0x00000000, 0x02520005 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_src_sample_planar.g5b |
---|
0,0 → 1,5 |
{ 0x00000201, 0x20080061, 0x00000000, 0x0000e000 }, |
{ 0x00600001, 0x20200022, 0x008d0000, 0x00000000 }, |
{ 0x01800031, 0x22001c09, 0x20000000, 0x0a2a0001 }, |
{ 0x01800031, 0x21c01c09, 0x20000000, 0x0a2a0003 }, |
{ 0x01800031, 0x22401c09, 0x20000000, 0x0a2a0005 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_src_sample_planar.g6b |
---|
0,0 → 1,5 |
{ 0x00000201, 0x20080061, 0x00000000, 0x0000e000 }, |
{ 0x00600001, 0x20200022, 0x008d0000, 0x00000000 }, |
{ 0x02800031, 0x22001cc9, 0x00000020, 0x0a2a0001 }, |
{ 0x02800031, 0x21c01cc9, 0x00000020, 0x0a2a0003 }, |
{ 0x02800031, 0x22401cc9, 0x00000020, 0x0a2a0005 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_src_sample_planar.g7b |
---|
0,0 → 1,5 |
{ 0x00000201, 0x20080061, 0x00000000, 0x0000e000 }, |
{ 0x00600001, 0x28200021, 0x008d0000, 0x00000000 }, |
{ 0x02800031, 0x22001ca9, 0x00000820, 0x0a2c0001 }, |
{ 0x02800031, 0x21c01ca9, 0x00000820, 0x0a2c0003 }, |
{ 0x02800031, 0x22401ca9, 0x00000820, 0x0a2c0005 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_write.g4b |
---|
0,0 → 1,18 |
{ 0x00600001, 0x204003be, 0x008d01c0, 0x00000000 }, |
{ 0x00600001, 0x206003be, 0x008d0200, 0x00000000 }, |
{ 0x00600001, 0x208003be, 0x008d0240, 0x00000000 }, |
{ 0x00600001, 0x20a003be, 0x008d0280, 0x00000000 }, |
{ 0x00601001, 0x20c003be, 0x008d01e0, 0x00000000 }, |
{ 0x00601001, 0x20e003be, 0x008d0220, 0x00000000 }, |
{ 0x00601001, 0x210003be, 0x008d0260, 0x00000000 }, |
{ 0x00601001, 0x212003be, 0x008d02a0, 0x00000000 }, |
{ 0x00600201, 0x20200022, 0x008d0020, 0x00000000 }, |
{ 0x00800031, 0x24001d28, 0x008d0000, 0x85a04800 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_write.g5b |
---|
0,0 → 1,6 |
{ 0x00802001, 0x304003be, 0x008d01c0, 0x00000000 }, |
{ 0x00802001, 0x306003be, 0x008d0200, 0x00000000 }, |
{ 0x00802001, 0x308003be, 0x008d0240, 0x00000000 }, |
{ 0x00802001, 0x30a003be, 0x008d0280, 0x00000000 }, |
{ 0x00600201, 0x202003be, 0x008d0020, 0x00000000 }, |
{ 0x00800031, 0x24001d28, 0x548d0000, 0x94084800 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_write.g6b |
---|
0,0 → 1,17 |
{ 0x00600001, 0x204003be, 0x008d01c0, 0x00000000 }, |
{ 0x00600001, 0x206003be, 0x008d01e0, 0x00000000 }, |
{ 0x00600001, 0x208003be, 0x008d0200, 0x00000000 }, |
{ 0x00600001, 0x20a003be, 0x008d0220, 0x00000000 }, |
{ 0x00600001, 0x20c003be, 0x008d0240, 0x00000000 }, |
{ 0x00600001, 0x20e003be, 0x008d0260, 0x00000000 }, |
{ 0x00600001, 0x210003be, 0x008d0280, 0x00000000 }, |
{ 0x00600001, 0x212003be, 0x008d02a0, 0x00000000 }, |
{ 0x05800031, 0x24001cc8, 0x00000040, 0x90019000 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_write.g7b |
---|
0,0 → 1,17 |
{ 0x00600001, 0x284003bd, 0x008d01c0, 0x00000000 }, |
{ 0x00600001, 0x286003bd, 0x008d01e0, 0x00000000 }, |
{ 0x00600001, 0x288003bd, 0x008d0200, 0x00000000 }, |
{ 0x00600001, 0x28a003bd, 0x008d0220, 0x00000000 }, |
{ 0x00600001, 0x28c003bd, 0x008d0240, 0x00000000 }, |
{ 0x00600001, 0x28e003bd, 0x008d0260, 0x00000000 }, |
{ 0x00600001, 0x290003bd, 0x008d0280, 0x00000000 }, |
{ 0x00600001, 0x292003bd, 0x008d02a0, 0x00000000 }, |
{ 0x05800031, 0x24001ca8, 0x00000840, 0x90031000 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_xy.g4b |
---|
0,0 → 1,4 |
{ 0x00800040, 0x23c06d29, 0x00480028, 0x10101010 }, |
{ 0x00800040, 0x23806d29, 0x0048002a, 0x11001100 }, |
{ 0x00802040, 0x2100753d, 0x008d03c0, 0x00004020 }, |
{ 0x00802040, 0x2140753d, 0x008d0380, 0x00004024 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_xy.g5b |
---|
0,0 → 1,4 |
{ 0x00800040, 0x23c06d29, 0x00480028, 0x10101010 }, |
{ 0x00800040, 0x23806d29, 0x0048002a, 0x11001100 }, |
{ 0x00802040, 0x2100753d, 0x008d03c0, 0x00004020 }, |
{ 0x00802040, 0x2140753d, 0x008d0380, 0x00004024 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_yuv_rgb.g4b |
---|
0,0 → 1,12 |
{ 0x00802040, 0x23007fbd, 0x008d0200, 0xbd808081 }, |
{ 0x00802041, 0x23007fbd, 0x008d0300, 0x3f94fdf4 }, |
{ 0x00802040, 0x22c07fbd, 0x008d01c0, 0xbf008084 }, |
{ 0x00802040, 0x23407fbd, 0x008d0240, 0xbf008084 }, |
{ 0x00802001, 0x240003bc, 0x008d0300, 0x00000000 }, |
{ 0x80802048, 0x21c07fbd, 0x008d02c0, 0x3fcc49ba }, |
{ 0x00802001, 0x240003bc, 0x008d0300, 0x00000000 }, |
{ 0x00802048, 0x24007fbc, 0x008d02c0, 0xbf5020c5 }, |
{ 0x80802048, 0x22007fbd, 0x008d0340, 0xbec8b439 }, |
{ 0x00802001, 0x240003bc, 0x008d0300, 0x00000000 }, |
{ 0x80802048, 0x22407fbd, 0x008d0340, 0x40011687 }, |
{ 0x00802001, 0x228003fd, 0x00000000, 0x3f800000 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_yuv_rgb.g5b |
---|
0,0 → 1,12 |
{ 0x00802040, 0x23007fbd, 0x008d0200, 0xbd808081 }, |
{ 0x00802041, 0x23007fbd, 0x008d0300, 0x3f94fdf4 }, |
{ 0x00802040, 0x22c07fbd, 0x008d01c0, 0xbf008084 }, |
{ 0x00802040, 0x23407fbd, 0x008d0240, 0xbf008084 }, |
{ 0x00802001, 0x240003bc, 0x008d0300, 0x00000000 }, |
{ 0x80802048, 0x21c07fbd, 0x008d02c0, 0x3fcc49ba }, |
{ 0x00802001, 0x240003bc, 0x008d0300, 0x00000000 }, |
{ 0x00802048, 0x24007fbc, 0x008d02c0, 0xbf5020c5 }, |
{ 0x80802048, 0x22007fbd, 0x008d0340, 0xbec8b439 }, |
{ 0x00802001, 0x240003bc, 0x008d0300, 0x00000000 }, |
{ 0x80802048, 0x22407fbd, 0x008d0340, 0x40011687 }, |
{ 0x00802001, 0x228003fd, 0x00000000, 0x3f800000 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_yuv_rgb.g6b |
---|
0,0 → 1,12 |
{ 0x00800040, 0x23007fbd, 0x008d0200, 0xbd808081 }, |
{ 0x00800041, 0x23007fbd, 0x008d0300, 0x3f94fdf4 }, |
{ 0x00800040, 0x22c07fbd, 0x008d01c0, 0xbf008084 }, |
{ 0x00800040, 0x23407fbd, 0x008d0240, 0xbf008084 }, |
{ 0x00800001, 0x240003bc, 0x008d0300, 0x00000000 }, |
{ 0x80800048, 0x21c07fbd, 0x008d02c0, 0x3fcc49ba }, |
{ 0x00800001, 0x240003bc, 0x008d0300, 0x00000000 }, |
{ 0x00800048, 0x24007fbc, 0x008d02c0, 0xbf5020c5 }, |
{ 0x80800048, 0x22007fbd, 0x008d0340, 0xbec8b439 }, |
{ 0x00800001, 0x240003bc, 0x008d0300, 0x00000000 }, |
{ 0x80800048, 0x22407fbd, 0x008d0340, 0x40011687 }, |
{ 0x00800001, 0x228003fd, 0x00000000, 0x3f800000 }, |
/drivers/video/Intel-2D/sna/render/exa_wm_yuv_rgb.g7b |
---|
0,0 → 1,12 |
{ 0x00800040, 0x23007fbd, 0x008d0200, 0xbd808081 }, |
{ 0x00800041, 0x23007fbd, 0x008d0300, 0x3f94fdf4 }, |
{ 0x00800040, 0x22c07fbd, 0x008d01c0, 0xbf008084 }, |
{ 0x00800040, 0x23407fbd, 0x008d0240, 0xbf008084 }, |
{ 0x00800001, 0x240003bc, 0x008d0300, 0x00000000 }, |
{ 0x80800048, 0x21c07fbd, 0x008d02c0, 0x3fcc49ba }, |
{ 0x00800001, 0x240003bc, 0x008d0300, 0x00000000 }, |
{ 0x00800048, 0x24007fbc, 0x008d02c0, 0xbf5020c5 }, |
{ 0x80800048, 0x22007fbd, 0x008d0340, 0xbec8b439 }, |
{ 0x00800001, 0x240003bc, 0x008d0300, 0x00000000 }, |
{ 0x80800048, 0x22407fbd, 0x008d0340, 0x40011687 }, |
{ 0x00800001, 0x228003fd, 0x00000000, 0x3f800000 }, |
/drivers/video/Intel-2D/sna/sna.c |
---|
0,0 → 1,991 |
/************************************************************************** |
Copyright 2001 VA Linux Systems Inc., Fremont, California. |
Copyright © 2002 by David Dawes |
All Rights Reserved. |
Permission is hereby granted, free of charge, to any person obtaining a |
copy of this software and associated documentation files (the "Software"), |
to deal in the Software without restriction, including without limitation |
on the rights to use, copy, modify, merge, publish, distribute, sub |
license, and/or sell copies of the Software, and to permit persons to whom |
the Software is furnished to do so, subject to the following conditions: |
The above copyright notice and this permission notice (including the next |
paragraph) shall be included in all copies or substantial portions of the |
Software. |
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL |
THE COPYRIGHT HOLDERS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, |
DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR |
OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE |
USE OR OTHER DEALINGS IN THE SOFTWARE. |
**************************************************************************/ |
/* |
* Authors: Jeff Hartmann <jhartmann@valinux.com> |
* Abraham van der Merwe <abraham@2d3d.co.za> |
* David Dawes <dawes@xfree86.org> |
* Alan Hourihane <alanh@tungstengraphics.com> |
*/ |
#ifdef HAVE_CONFIG_H |
#include "config.h" |
#endif |
#include <memory.h> |
#include <malloc.h> |
#include <kos32sys.h> |
#include <pixlib2.h> |
#include "i915_pciids.h" |
#include "compiler.h" |
#include "sna.h" |
#include "intel_driver.h" |
#define to_surface(x) (surface_t*)((x)->handle) |
static struct sna_fb sna_fb; |
static int tls_mask; |
int tls_alloc(void); |
static inline void *tls_get(int key) |
{ |
void *val; |
__asm__ __volatile__( |
"movl %%fs:(%1), %0" |
:"=r"(val) |
:"r"(key)); |
return val; |
}; |
static inline int |
tls_set(int key, const void *ptr) |
{ |
if(!(key & 3)) |
{ |
__asm__ __volatile__( |
"movl %0, %%fs:(%1)" |
::"r"(ptr),"r"(key)); |
return 0; |
} |
else return -1; |
} |
int kgem_init_fb(struct kgem *kgem, struct sna_fb *fb); |
int kgem_update_fb(struct kgem *kgem, struct sna_fb *fb); |
uint32_t kgem_surface_size(struct kgem *kgem,bool relaxed_fencing, |
unsigned flags, uint32_t width, uint32_t height, |
uint32_t bpp, uint32_t tiling, uint32_t *pitch); |
struct kgem_bo *kgem_bo_from_handle(struct kgem *kgem, int handle, |
int pitch, int height); |
void kgem_close_batches(struct kgem *kgem); |
void sna_bo_destroy(struct kgem *kgem, struct kgem_bo *bo); |
const struct intel_device_info * |
intel_detect_chipset(struct pci_device *pci); |
static bool sna_solid_cache_init(struct sna *sna); |
struct sna *sna_device; |
__LOCK_INIT_RECURSIVE(, __sna_lock); |
static void no_render_reset(struct sna *sna) |
{ |
(void)sna; |
} |
static void no_render_flush(struct sna *sna) |
{ |
(void)sna; |
} |
static void |
no_render_context_switch(struct kgem *kgem, |
int new_mode) |
{ |
if (!kgem->nbatch) |
return; |
if (kgem_ring_is_idle(kgem, kgem->ring)) { |
DBG(("%s: GPU idle, flushing\n", __FUNCTION__)); |
_kgem_submit(kgem); |
} |
(void)new_mode; |
} |
static void |
no_render_retire(struct kgem *kgem) |
{ |
(void)kgem; |
} |
static void |
no_render_expire(struct kgem *kgem) |
{ |
(void)kgem; |
} |
static void |
no_render_fini(struct sna *sna) |
{ |
(void)sna; |
} |
const char *no_render_init(struct sna *sna) |
{ |
struct sna_render *render = &sna->render; |
memset (render,0, sizeof (*render)); |
render->prefer_gpu = PREFER_GPU_BLT; |
render->vertices = render->vertex_data; |
render->vertex_size = ARRAY_SIZE(render->vertex_data); |
render->reset = no_render_reset; |
render->flush = no_render_flush; |
render->fini = no_render_fini; |
sna->kgem.context_switch = no_render_context_switch; |
sna->kgem.retire = no_render_retire; |
sna->kgem.expire = no_render_expire; |
sna->kgem.mode = KGEM_RENDER; |
sna->kgem.ring = KGEM_RENDER; |
sna_vertex_init(sna); |
return "generic"; |
} |
void sna_vertex_init(struct sna *sna) |
{ |
// pthread_mutex_init(&sna->render.lock, NULL); |
// pthread_cond_init(&sna->render.wait, NULL); |
sna->render.active = 0; |
} |
int sna_accel_init(struct sna *sna) |
{ |
const char *backend; |
backend = no_render_init(sna); |
if (sna->info->gen >= 0100) |
(void)backend; |
else if (sna->info->gen >= 070) |
backend = gen7_render_init(sna, backend); |
else if (sna->info->gen >= 060) |
backend = gen6_render_init(sna, backend); |
else if (sna->info->gen >= 050) |
backend = gen5_render_init(sna, backend); |
else if (sna->info->gen >= 040) |
backend = gen4_render_init(sna, backend); |
else if (sna->info->gen >= 030) |
backend = gen3_render_init(sna, backend); |
DBG(("%s(backend=%s, prefer_gpu=%x)\n", |
__FUNCTION__, backend, sna->render.prefer_gpu)); |
kgem_reset(&sna->kgem); |
sna_device = sna; |
return kgem_init_fb(&sna->kgem, &sna_fb); |
} |
int sna_init(uint32_t service) |
{ |
ioctl_t io; |
int caps = 0; |
static struct pci_device device; |
struct sna *sna; |
DBG(("%s\n", __FUNCTION__)); |
__lock_acquire_recursive(__sna_lock); |
if(sna_device) |
goto done; |
io.handle = service; |
io.io_code = SRV_GET_PCI_INFO; |
io.input = &device; |
io.inp_size = sizeof(device); |
io.output = NULL; |
io.out_size = 0; |
if (call_service(&io)!=0) |
goto err1; |
sna = malloc(sizeof(*sna)); |
if (sna == NULL) |
goto err1; |
memset(sna, 0, sizeof(*sna)); |
sna->cpu_features = sna_cpu_detect(); |
sna->PciInfo = &device; |
sna->info = intel_detect_chipset(sna->PciInfo); |
sna->scrn = service; |
kgem_init(&sna->kgem, service, sna->PciInfo, sna->info->gen); |
/* Disable tiling by default */ |
sna->tiling = 0; |
/* Default fail-safe value of 75 Hz */ |
// sna->vblank_interval = 1000 * 1000 * 1000 / 75; |
sna->flags = 0; |
sna_accel_init(sna); |
tls_mask = tls_alloc(); |
// printf("tls mask %x\n", tls_mask); |
done: |
caps = sna_device->render.caps; |
err1: |
__lock_release_recursive(__sna_lock); |
return caps; |
} |
void sna_fini() |
{ |
if( sna_device ) |
{ |
struct kgem_bo *mask; |
__lock_acquire_recursive(__sna_lock); |
mask = tls_get(tls_mask); |
sna_device->render.fini(sna_device); |
if(mask) |
kgem_bo_destroy(&sna_device->kgem, mask); |
kgem_close_batches(&sna_device->kgem); |
kgem_cleanup_cache(&sna_device->kgem); |
sna_device = NULL; |
__lock_release_recursive(__sna_lock); |
}; |
} |
#if 0 |
static bool sna_solid_cache_init(struct sna *sna) |
{ |
struct sna_solid_cache *cache = &sna->render.solid_cache; |
DBG(("%s\n", __FUNCTION__)); |
cache->cache_bo = |
kgem_create_linear(&sna->kgem, sizeof(cache->color)); |
if (!cache->cache_bo) |
return FALSE; |
/* |
* Initialise [0] with white since it is very common and filling the |
* zeroth slot simplifies some of the checks. |
*/ |
cache->color[0] = 0xffffffff; |
cache->bo[0] = kgem_create_proxy(cache->cache_bo, 0, sizeof(uint32_t)); |
cache->bo[0]->pitch = 4; |
cache->dirty = 1; |
cache->size = 1; |
cache->last = 0; |
return TRUE; |
} |
void |
sna_render_flush_solid(struct sna *sna) |
{ |
struct sna_solid_cache *cache = &sna->render.solid_cache; |
DBG(("sna_render_flush_solid(size=%d)\n", cache->size)); |
assert(cache->dirty); |
assert(cache->size); |
kgem_bo_write(&sna->kgem, cache->cache_bo, |
cache->color, cache->size*sizeof(uint32_t)); |
cache->dirty = 0; |
cache->last = 0; |
} |
static void |
sna_render_finish_solid(struct sna *sna, bool force) |
{ |
struct sna_solid_cache *cache = &sna->render.solid_cache; |
int i; |
DBG(("sna_render_finish_solid(force=%d, domain=%d, busy=%d, dirty=%d)\n", |
force, cache->cache_bo->domain, cache->cache_bo->rq != NULL, cache->dirty)); |
if (!force && cache->cache_bo->domain != DOMAIN_GPU) |
return; |
if (cache->dirty) |
sna_render_flush_solid(sna); |
for (i = 0; i < cache->size; i++) { |
if (cache->bo[i] == NULL) |
continue; |
kgem_bo_destroy(&sna->kgem, cache->bo[i]); |
cache->bo[i] = NULL; |
} |
kgem_bo_destroy(&sna->kgem, cache->cache_bo); |
DBG(("sna_render_finish_solid reset\n")); |
cache->cache_bo = kgem_create_linear(&sna->kgem, sizeof(cache->color)); |
cache->bo[0] = kgem_create_proxy(cache->cache_bo, 0, sizeof(uint32_t)); |
cache->bo[0]->pitch = 4; |
if (force) |
cache->size = 1; |
} |
struct kgem_bo * |
sna_render_get_solid(struct sna *sna, uint32_t color) |
{ |
struct sna_solid_cache *cache = &sna->render.solid_cache; |
int i; |
DBG(("%s: %08x\n", __FUNCTION__, color)); |
// if ((color & 0xffffff) == 0) /* alpha only */ |
// return kgem_bo_reference(sna->render.alpha_cache.bo[color>>24]); |
if (color == 0xffffffff) { |
DBG(("%s(white)\n", __FUNCTION__)); |
return kgem_bo_reference(cache->bo[0]); |
} |
if (cache->color[cache->last] == color) { |
DBG(("sna_render_get_solid(%d) = %x (last)\n", |
cache->last, color)); |
return kgem_bo_reference(cache->bo[cache->last]); |
} |
for (i = 1; i < cache->size; i++) { |
if (cache->color[i] == color) { |
if (cache->bo[i] == NULL) { |
DBG(("sna_render_get_solid(%d) = %x (recreate)\n", |
i, color)); |
goto create; |
} else { |
DBG(("sna_render_get_solid(%d) = %x (old)\n", |
i, color)); |
goto done; |
} |
} |
} |
sna_render_finish_solid(sna, i == ARRAY_SIZE(cache->color)); |
i = cache->size++; |
cache->color[i] = color; |
cache->dirty = 1; |
DBG(("sna_render_get_solid(%d) = %x (new)\n", i, color)); |
create: |
cache->bo[i] = kgem_create_proxy(cache->cache_bo, |
i*sizeof(uint32_t), sizeof(uint32_t)); |
cache->bo[i]->pitch = 4; |
done: |
cache->last = i; |
return kgem_bo_reference(cache->bo[i]); |
} |
#endif |
int sna_blit_copy(bitmap_t *src_bitmap, int dst_x, int dst_y, |
int w, int h, int src_x, int src_y) |
{ |
struct sna_copy_op copy; |
struct _Pixmap src, dst; |
struct kgem_bo *src_bo; |
char proc_info[1024]; |
int winx, winy; |
get_proc_info(proc_info); |
winx = *(uint32_t*)(proc_info+34); |
winy = *(uint32_t*)(proc_info+38); |
memset(&src, 0, sizeof(src)); |
memset(&dst, 0, sizeof(dst)); |
src.drawable.bitsPerPixel = 32; |
src.drawable.width = src_bitmap->width; |
src.drawable.height = src_bitmap->height; |
dst.drawable.bitsPerPixel = 32; |
dst.drawable.width = sna_fb.width; |
dst.drawable.height = sna_fb.height; |
memset(©, 0, sizeof(copy)); |
src_bo = (struct kgem_bo*)src_bitmap->handle; |
if( sna_device->render.copy(sna_device, GXcopy, |
&src, src_bo, |
&dst, sna_fb.fb_bo, ©) ) |
{ |
copy.blt(sna_device, ©, src_x, src_y, w, h, winx+dst_x, winy+dst_y); |
copy.done(sna_device, ©); |
} |
kgem_submit(&sna_device->kgem); |
return 0; |
// __asm__ __volatile__("int3"); |
}; |
typedef struct |
{ |
uint32_t width; |
uint32_t height; |
void *data; |
uint32_t pitch; |
struct kgem_bo *bo; |
uint32_t bo_size; |
uint32_t flags; |
}surface_t; |
int sna_create_bitmap(bitmap_t *bitmap) |
{ |
surface_t *sf; |
struct kgem_bo *bo; |
sf = malloc(sizeof(*sf)); |
if(sf == NULL) |
goto err_1; |
__lock_acquire_recursive(__sna_lock); |
bo = kgem_create_2d(&sna_device->kgem, bitmap->width, bitmap->height, |
32,I915_TILING_NONE, CREATE_CPU_MAP); |
if(bo == NULL) |
goto err_2; |
void *map = kgem_bo_map(&sna_device->kgem, bo); |
if(map == NULL) |
goto err_3; |
sf->width = bitmap->width; |
sf->height = bitmap->height; |
sf->data = map; |
sf->pitch = bo->pitch; |
sf->bo = bo; |
sf->bo_size = PAGE_SIZE * bo->size.pages.count; |
sf->flags = bitmap->flags; |
bitmap->handle = (uint32_t)sf; |
__lock_release_recursive(__sna_lock); |
return 0; |
err_3: |
kgem_bo_destroy(&sna_device->kgem, bo); |
err_2: |
__lock_release_recursive(__sna_lock); |
free(sf); |
err_1: |
return -1; |
}; |
int sna_bitmap_from_handle(bitmap_t *bitmap, uint32_t handle) |
{ |
surface_t *sf; |
struct kgem_bo *bo; |
sf = malloc(sizeof(*sf)); |
if(sf == NULL) |
goto err_1; |
__lock_acquire_recursive(__sna_lock); |
bo = kgem_bo_from_handle(&sna_device->kgem, handle, bitmap->pitch, bitmap->height); |
__lock_release_recursive(__sna_lock); |
sf->width = bitmap->width; |
sf->height = bitmap->height; |
sf->data = NULL; |
sf->pitch = bo->pitch; |
sf->bo = bo; |
sf->bo_size = PAGE_SIZE * bo->size.pages.count; |
sf->flags = bitmap->flags; |
bitmap->handle = (uint32_t)sf; |
return 0; |
err_2: |
__lock_release_recursive(__sna_lock); |
free(sf); |
err_1: |
return -1; |
}; |
void sna_set_bo_handle(bitmap_t *bitmap, int handle) |
{ |
surface_t *sf = to_surface(bitmap); |
struct kgem_bo *bo = sf->bo; |
bo->handle = handle; |
} |
int sna_destroy_bitmap(bitmap_t *bitmap) |
{ |
surface_t *sf = to_surface(bitmap); |
__lock_acquire_recursive(__sna_lock); |
kgem_bo_destroy(&sna_device->kgem, sf->bo); |
__lock_release_recursive(__sna_lock); |
free(sf); |
bitmap->handle = -1; |
bitmap->data = (void*)-1; |
bitmap->pitch = -1; |
return 0; |
}; |
int sna_lock_bitmap(bitmap_t *bitmap) |
{ |
surface_t *sf = to_surface(bitmap); |
// printf("%s\n", __FUNCTION__); |
__lock_acquire_recursive(__sna_lock); |
kgem_bo_sync__cpu(&sna_device->kgem, sf->bo); |
__lock_release_recursive(__sna_lock); |
bitmap->data = sf->data; |
bitmap->pitch = sf->pitch; |
return 0; |
}; |
int sna_resize_bitmap(bitmap_t *bitmap) |
{ |
surface_t *sf = to_surface(bitmap); |
struct kgem *kgem = &sna_device->kgem; |
struct kgem_bo *bo = sf->bo; |
uint32_t size; |
uint32_t pitch; |
bitmap->pitch = -1; |
bitmap->data = (void *) -1; |
size = kgem_surface_size(kgem,kgem->has_relaxed_fencing, CREATE_CPU_MAP, |
bitmap->width, bitmap->height, 32, I915_TILING_NONE, &pitch); |
assert(size && size <= kgem->max_object_size); |
if(sf->bo_size >= size) |
{ |
sf->width = bitmap->width; |
sf->height = bitmap->height; |
sf->pitch = pitch; |
bo->pitch = pitch; |
return 0; |
} |
else |
{ |
__lock_acquire_recursive(__sna_lock); |
sna_bo_destroy(kgem, bo); |
sf->bo = NULL; |
bo = kgem_create_2d(kgem, bitmap->width, bitmap->height, |
32, I915_TILING_NONE, CREATE_CPU_MAP); |
if(bo == NULL) |
{ |
__lock_release_recursive(__sna_lock); |
return -1; |
}; |
void *map = kgem_bo_map(kgem, bo); |
if(map == NULL) |
{ |
sna_bo_destroy(kgem, bo); |
__lock_release_recursive(__sna_lock); |
return -1; |
}; |
__lock_release_recursive(__sna_lock); |
sf->width = bitmap->width; |
sf->height = bitmap->height; |
sf->data = map; |
sf->pitch = bo->pitch; |
sf->bo = bo; |
sf->bo_size = PAGE_SIZE * bo->size.pages.count; |
} |
return 0; |
}; |
int sna_create_mask() |
{ |
struct kgem_bo *bo; |
// printf("%s width %d height %d\n", __FUNCTION__, sna_fb.width, sna_fb.height); |
__lock_acquire_recursive(__sna_lock); |
bo = kgem_create_2d(&sna_device->kgem, sna_fb.width, sna_fb.height, |
8,I915_TILING_NONE, CREATE_CPU_MAP); |
if(unlikely(bo == NULL)) |
goto err_1; |
int *map = kgem_bo_map(&sna_device->kgem, bo); |
if(map == NULL) |
goto err_2; |
__lock_release_recursive(__sna_lock); |
memset(map, 0, bo->pitch * sna_fb.height); |
tls_set(tls_mask, bo); |
return 0; |
err_2: |
kgem_bo_destroy(&sna_device->kgem, bo); |
err_1: |
__lock_release_recursive(__sna_lock); |
return -1; |
}; |
bool |
gen6_composite(struct sna *sna, |
uint8_t op, |
PixmapPtr src, struct kgem_bo *src_bo, |
PixmapPtr mask,struct kgem_bo *mask_bo, |
PixmapPtr dst, struct kgem_bo *dst_bo, |
int32_t src_x, int32_t src_y, |
int32_t msk_x, int32_t msk_y, |
int32_t dst_x, int32_t dst_y, |
int32_t width, int32_t height, |
struct sna_composite_op *tmp); |
#define MAP(ptr) ((void*)((uintptr_t)(ptr) & ~3)) |
int sna_blit_tex(bitmap_t *bitmap, bool scale, int dst_x, int dst_y, |
int w, int h, int src_x, int src_y) |
{ |
surface_t *sf = to_surface(bitmap); |
struct drm_i915_mask_update update; |
struct sna_composite_op composite; |
struct _Pixmap src, dst, mask; |
struct kgem_bo *src_bo, *mask_bo; |
int winx, winy; |
char proc_info[1024]; |
get_proc_info(proc_info); |
winx = *(uint32_t*)(proc_info+34); |
winy = *(uint32_t*)(proc_info+38); |
// winw = *(uint32_t*)(proc_info+42)+1; |
// winh = *(uint32_t*)(proc_info+46)+1; |
mask_bo = tls_get(tls_mask); |
if(unlikely(mask_bo == NULL)) |
{ |
sna_create_mask(); |
mask_bo = tls_get(tls_mask); |
if( mask_bo == NULL) |
return -1; |
}; |
if(kgem_update_fb(&sna_device->kgem, &sna_fb)) |
{ |
__lock_acquire_recursive(__sna_lock); |
kgem_bo_destroy(&sna_device->kgem, mask_bo); |
__lock_release_recursive(__sna_lock); |
sna_create_mask(); |
mask_bo = tls_get(tls_mask); |
if( mask_bo == NULL) |
return -1; |
} |
VG_CLEAR(update); |
update.handle = mask_bo->handle; |
update.bo_map = (int)kgem_bo_map__cpu(&sna_device->kgem, mask_bo); |
drmIoctl(sna_device->kgem.fd, SRV_MASK_UPDATE, &update); |
mask_bo->pitch = update.bo_pitch; |
memset(&src, 0, sizeof(src)); |
memset(&dst, 0, sizeof(dst)); |
memset(&mask, 0, sizeof(dst)); |
src.drawable.bitsPerPixel = 32; |
src.drawable.width = sf->width; |
src.drawable.height = sf->height; |
dst.drawable.bitsPerPixel = 32; |
dst.drawable.width = sna_fb.width; |
dst.drawable.height = sna_fb.height; |
mask.drawable.bitsPerPixel = 8; |
mask.drawable.width = update.width; |
mask.drawable.height = update.height; |
memset(&composite, 0, sizeof(composite)); |
src_bo = sf->bo; |
__lock_acquire_recursive(__sna_lock); |
if( sna_device->render.blit_tex(sna_device, PictOpSrc,scale, |
&src, src_bo, |
&mask, mask_bo, |
&dst, sna_fb.fb_bo, |
src_x, src_y, |
dst_x, dst_y, |
winx+dst_x, winy+dst_y, |
w, h, |
&composite) ) |
{ |
struct sna_composite_rectangles r; |
r.src.x = src_x; |
r.src.y = src_y; |
r.mask.x = dst_x; |
r.mask.y = dst_y; |
r.dst.x = winx+dst_x; |
r.dst.y = winy+dst_y; |
r.width = w; |
r.height = h; |
composite.blt(sna_device, &composite, &r); |
composite.done(sna_device, &composite); |
}; |
kgem_submit(&sna_device->kgem); |
__lock_release_recursive(__sna_lock); |
bitmap->data = (void*)-1; |
bitmap->pitch = -1; |
return 0; |
} |
static const struct intel_device_info intel_generic_info = { |
.gen = -1, |
}; |
static const struct intel_device_info intel_i915_info = { |
.gen = 030, |
}; |
static const struct intel_device_info intel_i945_info = { |
.gen = 031, |
}; |
static const struct intel_device_info intel_g33_info = { |
.gen = 033, |
}; |
static const struct intel_device_info intel_i965_info = { |
.gen = 040, |
}; |
static const struct intel_device_info intel_g4x_info = { |
.gen = 045, |
}; |
static const struct intel_device_info intel_ironlake_info = { |
.gen = 050, |
}; |
static const struct intel_device_info intel_sandybridge_info = { |
.gen = 060, |
}; |
static const struct intel_device_info intel_ivybridge_info = { |
.gen = 070, |
}; |
static const struct intel_device_info intel_valleyview_info = { |
.gen = 071, |
}; |
static const struct intel_device_info intel_haswell_info = { |
.gen = 075, |
}; |
#define INTEL_DEVICE_MATCH(d,i) \ |
{ 0x8086, (d), PCI_MATCH_ANY, PCI_MATCH_ANY, 0x3 << 16, 0xff << 16, (intptr_t)(i) } |
static const struct pci_id_match intel_device_match[] = { |
INTEL_I915G_IDS(&intel_i915_info), |
INTEL_I915GM_IDS(&intel_i915_info), |
INTEL_I945G_IDS(&intel_i945_info), |
INTEL_I945GM_IDS(&intel_i945_info), |
INTEL_G33_IDS(&intel_g33_info), |
INTEL_PINEVIEW_IDS(&intel_g33_info), |
INTEL_I965G_IDS(&intel_i965_info), |
INTEL_I965GM_IDS(&intel_i965_info), |
INTEL_G45_IDS(&intel_g4x_info), |
INTEL_GM45_IDS(&intel_g4x_info), |
INTEL_IRONLAKE_D_IDS(&intel_ironlake_info), |
INTEL_IRONLAKE_M_IDS(&intel_ironlake_info), |
INTEL_SNB_D_IDS(&intel_sandybridge_info), |
INTEL_SNB_M_IDS(&intel_sandybridge_info), |
INTEL_IVB_D_IDS(&intel_ivybridge_info), |
INTEL_IVB_M_IDS(&intel_ivybridge_info), |
INTEL_HSW_D_IDS(&intel_haswell_info), |
INTEL_HSW_M_IDS(&intel_haswell_info), |
INTEL_VLV_D_IDS(&intel_valleyview_info), |
INTEL_VLV_M_IDS(&intel_valleyview_info), |
INTEL_VGA_DEVICE(PCI_MATCH_ANY, &intel_generic_info), |
{ 0, 0, 0 }, |
}; |
const struct pci_id_match *PciDevMatch(uint16_t dev,const struct pci_id_match *list) |
{ |
while(list->device_id) |
{ |
if(dev==list->device_id) |
return list; |
list++; |
} |
return NULL; |
} |
const struct intel_device_info * |
intel_detect_chipset(struct pci_device *pci) |
{ |
const struct pci_id_match *ent = NULL; |
ent = PciDevMatch(pci->device_id, intel_device_match); |
if(ent != NULL) |
return (const struct intel_device_info*)ent->match_data; |
else |
return &intel_generic_info; |
#if 0 |
for (i = 0; intel_chipsets[i].name != NULL; i++) { |
if (DEVICE_ID(pci) == intel_chipsets[i].token) { |
name = intel_chipsets[i].name; |
break; |
} |
} |
if (name == NULL) { |
xf86DrvMsg(scrn->scrnIndex, X_WARNING, "unknown chipset\n"); |
name = "unknown"; |
} else { |
xf86DrvMsg(scrn->scrnIndex, from, |
"Integrated Graphics Chipset: Intel(R) %s\n", |
name); |
} |
scrn->chipset = name; |
#endif |
} |
int intel_get_device_id(int fd) |
{ |
struct drm_i915_getparam gp; |
int devid = 0; |
memset(&gp, 0, sizeof(gp)); |
gp.param = I915_PARAM_CHIPSET_ID; |
gp.value = &devid; |
if (drmIoctl(fd, DRM_IOCTL_I915_GETPARAM, &gp)) |
return 0; |
return devid; |
} |
int drmIoctl(int fd, unsigned long request, void *arg) |
{ |
ioctl_t io; |
io.handle = fd; |
io.io_code = request; |
io.input = arg; |
io.inp_size = 64; |
io.output = NULL; |
io.out_size = 0; |
return call_service(&io); |
} |
/drivers/video/Intel-2D/sna/sna.h |
---|
0,0 → 1,607 |
/************************************************************************** |
Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas. |
Copyright © 2002 David Dawes |
All Rights Reserved. |
Permission is hereby granted, free of charge, to any person obtaining a |
copy of this software and associated documentation files (the |
"Software"), to deal in the Software without restriction, including |
without limitation the rights to use, copy, modify, merge, publish, |
distribute, sub license, and/or sell copies of the Software, and to |
permit persons to whom the Software is furnished to do so, subject to |
the following conditions: |
The above copyright notice and this permission notice (including the |
next paragraph) shall be included in all copies or substantial portions |
of the Software. |
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS |
OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF |
MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. |
IN NO EVENT SHALL PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR |
ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, |
TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE |
SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. |
**************************************************************************/ |
/* |
* Authors: |
* Keith Whitwell <keith@tungstengraphics.com> |
* David Dawes <dawes@xfree86.org> |
* |
*/ |
#ifndef _SNA_H_ |
#define _SNA_H_ |
#include <stdint.h> |
#ifdef HAVE_CONFIG_H |
#include "config.h" |
#endif |
#include "compiler.h" |
#include <memory.h> |
#include <malloc.h> |
#include <errno.h> |
#include <kos32sys.h> |
#include "intel_driver.h" |
#include "pciaccess.h" |
#include <drm.h> |
#include <i915_drm.h> |
#ifdef HAVE_DRI2_H |
#include <dri2.h> |
#endif |
#if HAVE_UDEV |
#include <libudev.h> |
#endif |
#if 0 |
#include <xorg-server.h> |
#include <xf86Crtc.h> |
#if XF86_CRTC_VERSION >= 5 |
#define HAS_PIXMAP_SHARING 1 |
#endif |
#include <xf86str.h> |
#include <windowstr.h> |
#include <glyphstr.h> |
#include <picturestr.h> |
#include <gcstruct.h> |
#include <xvdix.h> |
#include <pciaccess.h> |
#include <xf86drmMode.h> |
#include "../compat-api.h" |
#endif |
#include <assert.h> |
#define ErrorF printf |
#if HAS_DEBUG_FULL |
#define DBG(x) ErrorF x |
#else |
#define DBG(x) |
#endif |
#define DEBUG_NO_BLT 0 |
#define DEBUG_FLUSH_BATCH 0 |
#define TEST_ALL 0 |
#define TEST_ACCEL (TEST_ALL || 0) |
#define TEST_BATCH (TEST_ALL || 0) |
#define TEST_BLT (TEST_ALL || 0) |
#define TEST_COMPOSITE (TEST_ALL || 0) |
#define TEST_DAMAGE (TEST_ALL || 0) |
#define TEST_GRADIENT (TEST_ALL || 0) |
#define TEST_GLYPHS (TEST_ALL || 0) |
#define TEST_IO (TEST_ALL || 0) |
#define TEST_KGEM (TEST_ALL || 0) |
#define TEST_RENDER (TEST_ALL || 0) |
int drmIoctl(int fd, unsigned long request, void *arg); |
#define PIXMAN_FORMAT(bpp,type,a,r,g,b) (((bpp) << 24) | \ |
((type) << 16) | \ |
((a) << 12) | \ |
((r) << 8) | \ |
((g) << 4) | \ |
((b))) |
#define PIXMAN_TYPE_OTHER 0 |
#define PIXMAN_TYPE_A 1 |
#define PIXMAN_TYPE_ARGB 2 |
#define PIXMAN_TYPE_ABGR 3 |
#define PIXMAN_TYPE_COLOR 4 |
#define PIXMAN_TYPE_GRAY 5 |
#define PIXMAN_TYPE_YUY2 6 |
#define PIXMAN_TYPE_YV12 7 |
#define PIXMAN_TYPE_BGRA 8 |
#define PIXMAN_TYPE_RGBA 9 |
#define PIXMAN_TYPE_ARGB_SRGB 10 |
/* 32bpp formats */ |
typedef enum { |
PIXMAN_a8r8g8b8 = PIXMAN_FORMAT(32,PIXMAN_TYPE_ARGB,8,8,8,8), |
PIXMAN_x8r8g8b8 = PIXMAN_FORMAT(32,PIXMAN_TYPE_ARGB,0,8,8,8), |
PIXMAN_a8b8g8r8 = PIXMAN_FORMAT(32,PIXMAN_TYPE_ABGR,8,8,8,8), |
PIXMAN_x8b8g8r8 = PIXMAN_FORMAT(32,PIXMAN_TYPE_ABGR,0,8,8,8), |
PIXMAN_b8g8r8a8 = PIXMAN_FORMAT(32,PIXMAN_TYPE_BGRA,8,8,8,8), |
PIXMAN_b8g8r8x8 = PIXMAN_FORMAT(32,PIXMAN_TYPE_BGRA,0,8,8,8), |
PIXMAN_r8g8b8a8 = PIXMAN_FORMAT(32,PIXMAN_TYPE_RGBA,8,8,8,8), |
PIXMAN_r8g8b8x8 = PIXMAN_FORMAT(32,PIXMAN_TYPE_RGBA,0,8,8,8), |
PIXMAN_x14r6g6b6 = PIXMAN_FORMAT(32,PIXMAN_TYPE_ARGB,0,6,6,6), |
PIXMAN_x2r10g10b10 = PIXMAN_FORMAT(32,PIXMAN_TYPE_ARGB,0,10,10,10), |
PIXMAN_a2r10g10b10 = PIXMAN_FORMAT(32,PIXMAN_TYPE_ARGB,2,10,10,10), |
PIXMAN_x2b10g10r10 = PIXMAN_FORMAT(32,PIXMAN_TYPE_ABGR,0,10,10,10), |
PIXMAN_a2b10g10r10 = PIXMAN_FORMAT(32,PIXMAN_TYPE_ABGR,2,10,10,10), |
/* sRGB formats */ |
PIXMAN_a8r8g8b8_sRGB = PIXMAN_FORMAT(32,PIXMAN_TYPE_ARGB_SRGB,8,8,8,8), |
/* 24bpp formats */ |
PIXMAN_r8g8b8 = PIXMAN_FORMAT(24,PIXMAN_TYPE_ARGB,0,8,8,8), |
PIXMAN_b8g8r8 = PIXMAN_FORMAT(24,PIXMAN_TYPE_ABGR,0,8,8,8), |
/* 16bpp formats */ |
PIXMAN_r5g6b5 = PIXMAN_FORMAT(16,PIXMAN_TYPE_ARGB,0,5,6,5), |
PIXMAN_b5g6r5 = PIXMAN_FORMAT(16,PIXMAN_TYPE_ABGR,0,5,6,5), |
PIXMAN_a1r5g5b5 = PIXMAN_FORMAT(16,PIXMAN_TYPE_ARGB,1,5,5,5), |
PIXMAN_x1r5g5b5 = PIXMAN_FORMAT(16,PIXMAN_TYPE_ARGB,0,5,5,5), |
PIXMAN_a1b5g5r5 = PIXMAN_FORMAT(16,PIXMAN_TYPE_ABGR,1,5,5,5), |
PIXMAN_x1b5g5r5 = PIXMAN_FORMAT(16,PIXMAN_TYPE_ABGR,0,5,5,5), |
PIXMAN_a4r4g4b4 = PIXMAN_FORMAT(16,PIXMAN_TYPE_ARGB,4,4,4,4), |
PIXMAN_x4r4g4b4 = PIXMAN_FORMAT(16,PIXMAN_TYPE_ARGB,0,4,4,4), |
PIXMAN_a4b4g4r4 = PIXMAN_FORMAT(16,PIXMAN_TYPE_ABGR,4,4,4,4), |
PIXMAN_x4b4g4r4 = PIXMAN_FORMAT(16,PIXMAN_TYPE_ABGR,0,4,4,4), |
/* 8bpp formats */ |
PIXMAN_a8 = PIXMAN_FORMAT(8,PIXMAN_TYPE_A,8,0,0,0), |
PIXMAN_r3g3b2 = PIXMAN_FORMAT(8,PIXMAN_TYPE_ARGB,0,3,3,2), |
PIXMAN_b2g3r3 = PIXMAN_FORMAT(8,PIXMAN_TYPE_ABGR,0,3,3,2), |
PIXMAN_a2r2g2b2 = PIXMAN_FORMAT(8,PIXMAN_TYPE_ARGB,2,2,2,2), |
PIXMAN_a2b2g2r2 = PIXMAN_FORMAT(8,PIXMAN_TYPE_ABGR,2,2,2,2), |
PIXMAN_c8 = PIXMAN_FORMAT(8,PIXMAN_TYPE_COLOR,0,0,0,0), |
PIXMAN_g8 = PIXMAN_FORMAT(8,PIXMAN_TYPE_GRAY,0,0,0,0), |
PIXMAN_x4a4 = PIXMAN_FORMAT(8,PIXMAN_TYPE_A,4,0,0,0), |
PIXMAN_x4c4 = PIXMAN_FORMAT(8,PIXMAN_TYPE_COLOR,0,0,0,0), |
PIXMAN_x4g4 = PIXMAN_FORMAT(8,PIXMAN_TYPE_GRAY,0,0,0,0), |
/* 4bpp formats */ |
PIXMAN_a4 = PIXMAN_FORMAT(4,PIXMAN_TYPE_A,4,0,0,0), |
PIXMAN_r1g2b1 = PIXMAN_FORMAT(4,PIXMAN_TYPE_ARGB,0,1,2,1), |
PIXMAN_b1g2r1 = PIXMAN_FORMAT(4,PIXMAN_TYPE_ABGR,0,1,2,1), |
PIXMAN_a1r1g1b1 = PIXMAN_FORMAT(4,PIXMAN_TYPE_ARGB,1,1,1,1), |
PIXMAN_a1b1g1r1 = PIXMAN_FORMAT(4,PIXMAN_TYPE_ABGR,1,1,1,1), |
PIXMAN_c4 = PIXMAN_FORMAT(4,PIXMAN_TYPE_COLOR,0,0,0,0), |
PIXMAN_g4 = PIXMAN_FORMAT(4,PIXMAN_TYPE_GRAY,0,0,0,0), |
/* 1bpp formats */ |
PIXMAN_a1 = PIXMAN_FORMAT(1,PIXMAN_TYPE_A,1,0,0,0), |
PIXMAN_g1 = PIXMAN_FORMAT(1,PIXMAN_TYPE_GRAY,0,0,0,0), |
/* YUV formats */ |
PIXMAN_yuy2 = PIXMAN_FORMAT(16,PIXMAN_TYPE_YUY2,0,0,0,0), |
PIXMAN_yv12 = PIXMAN_FORMAT(12,PIXMAN_TYPE_YV12,0,0,0,0) |
} pixman_format_code_t; |
typedef enum _PictFormatShort { |
PICT_a2r10g10b10 = PIXMAN_a2r10g10b10, |
PICT_x2r10g10b10 = PIXMAN_x2r10g10b10, |
PICT_a2b10g10r10 = PIXMAN_a2b10g10r10, |
PICT_x2b10g10r10 = PIXMAN_x2b10g10r10, |
PICT_a8r8g8b8 = PIXMAN_a8r8g8b8, |
PICT_x8r8g8b8 = PIXMAN_x8r8g8b8, |
PICT_a8b8g8r8 = PIXMAN_a8b8g8r8, |
PICT_x8b8g8r8 = PIXMAN_x8b8g8r8, |
PICT_b8g8r8a8 = PIXMAN_b8g8r8a8, |
PICT_b8g8r8x8 = PIXMAN_b8g8r8x8, |
/* 24bpp formats */ |
PICT_r8g8b8 = PIXMAN_r8g8b8, |
PICT_b8g8r8 = PIXMAN_b8g8r8, |
/* 16bpp formats */ |
PICT_r5g6b5 = PIXMAN_r5g6b5, |
PICT_b5g6r5 = PIXMAN_b5g6r5, |
PICT_a1r5g5b5 = PIXMAN_a1r5g5b5, |
PICT_x1r5g5b5 = PIXMAN_x1r5g5b5, |
PICT_a1b5g5r5 = PIXMAN_a1b5g5r5, |
PICT_x1b5g5r5 = PIXMAN_x1b5g5r5, |
PICT_a4r4g4b4 = PIXMAN_a4r4g4b4, |
PICT_x4r4g4b4 = PIXMAN_x4r4g4b4, |
PICT_a4b4g4r4 = PIXMAN_a4b4g4r4, |
PICT_x4b4g4r4 = PIXMAN_x4b4g4r4, |
/* 8bpp formats */ |
PICT_a8 = PIXMAN_a8, |
PICT_r3g3b2 = PIXMAN_r3g3b2, |
PICT_b2g3r3 = PIXMAN_b2g3r3, |
PICT_a2r2g2b2 = PIXMAN_a2r2g2b2, |
PICT_a2b2g2r2 = PIXMAN_a2b2g2r2, |
PICT_c8 = PIXMAN_c8, |
PICT_g8 = PIXMAN_g8, |
PICT_x4a4 = PIXMAN_x4a4, |
PICT_x4c4 = PIXMAN_x4c4, |
PICT_x4g4 = PIXMAN_x4g4, |
/* 4bpp formats */ |
PICT_a4 = PIXMAN_a4, |
PICT_r1g2b1 = PIXMAN_r1g2b1, |
PICT_b1g2r1 = PIXMAN_b1g2r1, |
PICT_a1r1g1b1 = PIXMAN_a1r1g1b1, |
PICT_a1b1g1r1 = PIXMAN_a1b1g1r1, |
PICT_c4 = PIXMAN_c4, |
PICT_g4 = PIXMAN_g4, |
/* 1bpp formats */ |
PICT_a1 = PIXMAN_a1, |
PICT_g1 = PIXMAN_g1 |
} PictFormatShort; |
#define PIXMAN_FORMAT_A(f) (((f) >> 12) & 0x0f) |
#define PIXMAN_FORMAT_RGB(f) (((f) ) & 0xfff) |
#define PICT_FORMAT_A(f) PIXMAN_FORMAT_A(f) |
#define RepeatNone 0 |
#define RepeatNormal 1 |
#define RepeatPad 2 |
#define RepeatReflect 3 |
#define PictFilterNearest 0 |
#define PictFilterBilinear 1 |
#define PictFilterFast 2 |
#define PictFilterGood 3 |
#define PictFilterBest 4 |
#define PictFilterConvolution 5 |
typedef int32_t pixman_fixed_16_16_t; |
typedef pixman_fixed_16_16_t pixman_fixed_t; |
struct pixman_transform |
{ |
pixman_fixed_t matrix[3][3]; |
}; |
typedef unsigned long Picture; |
typedef unsigned long PictFormat; |
typedef struct _Pixmap *PixmapPtr; |
typedef struct _Picture *PicturePtr; |
typedef struct _Drawable *DrawablePtr; |
typedef struct _PictFormat *PictFormatPtr; |
typedef struct pixman_transform PictTransform, *PictTransformPtr; |
typedef struct _Drawable { |
unsigned char type; /* DRAWABLE_<type> */ |
unsigned char class; /* specific to type */ |
unsigned char depth; |
unsigned char bitsPerPixel; |
unsigned int id; /* resource id */ |
short x; /* window: screen absolute, pixmap: 0 */ |
short y; /* window: screen absolute, pixmap: 0 */ |
unsigned short width; |
unsigned short height; |
} DrawableRec; |
/* |
* PIXMAP -- device dependent |
*/ |
typedef struct _Pixmap { |
DrawableRec drawable; |
// PrivateRec *devPrivates; |
int refcnt; |
int devKind; /* This is the pitch of the pixmap, typically width*bpp/8. */ |
// DevUnion devPrivate; /* When !NULL, devPrivate.ptr points to the raw pixel data. */ |
#ifdef COMPOSITE |
short screen_x; |
short screen_y; |
#endif |
unsigned usage_hint; /* see CREATE_PIXMAP_USAGE_* */ |
PixmapPtr master_pixmap; /* pointer to master copy of pixmap for pixmap sharing */ |
} PixmapRec; |
typedef struct _PictFormat { |
uint32_t id; |
uint32_t format; /* except bpp */ |
unsigned char type; |
unsigned char depth; |
// DirectFormatRec direct; |
// IndexFormatRec index; |
} PictFormatRec; |
typedef struct _Picture { |
DrawablePtr pDrawable; |
// PictFormatPtr pFormat; |
PictFormatShort format; /* PICT_FORMAT */ |
int refcnt; |
uint32_t id; |
unsigned int repeat:1; |
unsigned int graphicsExposures:1; |
unsigned int subWindowMode:1; |
unsigned int polyEdge:1; |
unsigned int polyMode:1; |
unsigned int freeCompClip:1; |
unsigned int clientClipType:2; |
unsigned int componentAlpha:1; |
unsigned int repeatType:2; |
unsigned int filter:3; |
// unsigned int stateChanges:CPLastBit; |
// unsigned int unused:18 - CPLastBit; |
// PicturePtr alphaMap; |
// PictTransform *transform; |
// SourcePictPtr pSourcePict; |
// xFixed *filter_params; |
// int filter_nparams; |
} PictureRec; |
#define PolyModePrecise 0 |
#define PolyModeImprecise 1 |
struct sna_fb |
{ |
uint32_t width; |
uint32_t height; |
uint32_t pitch; |
uint32_t tiling; |
struct kgem_bo *fb_bo; |
}; |
struct pixman_box16 |
{ |
int16_t x1, y1, x2, y2; |
}; |
typedef struct pixman_box16 BoxRec; |
typedef unsigned int CARD32; |
typedef unsigned short CARD16; |
#include "sna_render.h" |
#include "kgem.h" |
#define GXclear 0x0 |
#define GXcopy 0x3 |
#define PictOpClear 0 |
#define PictOpSrc 1 |
#define PictOpDst 2 |
#define PictOpOver 3 |
#define PictOpOverReverse 4 |
#define PictOpIn 5 |
#define PictOpInReverse 6 |
#define PictOpOut 7 |
#define PictOpOutReverse 8 |
#define PictOpAtop 9 |
#define PictOpAtopReverse 10 |
#define PictOpXor 11 |
#define PictOpAdd 12 |
#define PictOpSaturate 13 |
#define PictOpMaximum 13 |
#define SNA_CURSOR_X 64 |
#define SNA_CURSOR_Y SNA_CURSOR_X |
struct sna_client { |
int is_compositor; /* only 4 bits used */ |
}; |
//#define assert(x) |
struct sna { |
struct kgem kgem; |
unsigned scrn; |
unsigned flags; |
#define SNA_NO_WAIT 0x1 |
#define SNA_NO_FLIP 0x2 |
#define SNA_TRIPLE_BUFFER 0x4 |
#define SNA_TEAR_FREE 0x10 |
#define SNA_FORCE_SHADOW 0x20 |
#define SNA_FLUSH_GTT 0x40 |
#define SNA_IS_HOSTED 0x80 |
#define SNA_PERFORMANCE 0x100 |
#define SNA_POWERSAVE 0x200 |
#define SNA_REPROBE 0x80000000 |
unsigned cpu_features; |
#define MMX 0x1 |
#define SSE 0x2 |
#define SSE2 0x4 |
#define SSE3 0x8 |
#define SSSE3 0x10 |
#define SSE4_1 0x20 |
#define SSE4_2 0x40 |
#define AVX 0x80 |
#define AVX2 0x100 |
struct list flush_pixmaps; |
struct list active_pixmaps; |
unsigned int tiling; |
#define SNA_TILING_FB 0x1 |
#define SNA_TILING_2D 0x2 |
#define SNA_TILING_ALL (~0) |
struct pci_device *PciInfo; |
const struct intel_device_info *info; |
// PicturePtr clear; |
struct { |
uint32_t fill_bo; |
uint32_t fill_pixel; |
uint32_t fill_alu; |
} blt_state; |
union { |
struct gen3_render_state gen3; |
struct gen4_render_state gen4; |
struct gen5_render_state gen5; |
struct gen6_render_state gen6; |
struct gen7_render_state gen7; |
} render_state; |
/* Broken-out options. */ |
// OptionInfoPtr Options; |
/* Driver phase/state information */ |
// Bool suspended; |
struct sna_render render; |
#if DEBUG_MEMORY |
struct { |
int pixmap_allocs; |
int cpu_bo_allocs; |
size_t shadow_pixels_bytes; |
size_t cpu_bo_bytes; |
} debug_memory; |
#endif |
}; |
static inline struct sna * |
to_sna_from_kgem(struct kgem *kgem) |
{ |
return container_of(kgem, struct sna, kgem); |
} |
#ifndef ARRAY_SIZE |
#define ARRAY_SIZE(x) (sizeof(x) / sizeof(x[0])) |
#endif |
#ifndef ALIGN |
#define ALIGN(i,m) (((i) + (m) - 1) & ~((m) - 1)) |
#endif |
#ifndef MIN |
#define MIN(a,b) ((a) <= (b) ? (a) : (b)) |
#endif |
#ifndef MAX |
#define MAX(a,b) ((a) >= (b) ? (a) : (b)) |
#endif |
static inline bool |
_sna_transform_point(const PictTransform *transform, |
int64_t x, int64_t y, int64_t result[3]) |
{ |
int j; |
for (j = 0; j < 3; j++) |
result[j] = (transform->matrix[j][0] * x + |
transform->matrix[j][1] * y + |
transform->matrix[j][2]); |
return result[2] != 0; |
} |
static inline void |
_sna_get_transformed_coordinates(int x, int y, |
const PictTransform *transform, |
float *x_out, float *y_out) |
{ |
int64_t result[3]; |
_sna_transform_point(transform, x, y, result); |
*x_out = result[0] / (double)result[2]; |
*y_out = result[1] / (double)result[2]; |
} |
static inline void |
_sna_get_transformed_scaled(int x, int y, |
const PictTransform *transform, const float *sf, |
float *x_out, float *y_out) |
{ |
*x_out = sf[0] * (transform->matrix[0][0] * x + |
transform->matrix[0][1] * y + |
transform->matrix[0][2]); |
*y_out = sf[1] * (transform->matrix[1][0] * x + |
transform->matrix[1][1] * y + |
transform->matrix[1][2]); |
} |
void |
sna_get_transformed_coordinates(int x, int y, |
const PictTransform *transform, |
float *x_out, float *y_out); |
void |
sna_get_transformed_coordinates_3d(int x, int y, |
const PictTransform *transform, |
float *x_out, float *y_out, float *z_out); |
bool sna_transform_is_affine(const PictTransform *t); |
bool sna_transform_is_integer_translation(const PictTransform *t, |
int16_t *tx, int16_t *ty); |
bool sna_transform_is_translation(const PictTransform *t, |
pixman_fixed_t *tx, pixman_fixed_t *ty); |
static inline bool |
sna_affine_transform_is_rotation(const PictTransform *t) |
{ |
assert(sna_transform_is_affine(t)); |
return t->matrix[0][1] | t->matrix[1][0]; |
} |
static inline bool |
sna_transform_equal(const PictTransform *a, const PictTransform *b) |
{ |
if (a == b) |
return true; |
if (a == NULL || b == NULL) |
return false; |
return memcmp(a, b, sizeof(*a)) == 0; |
} |
#endif /* _SNA_H */ |
/drivers/video/Intel-2D/sna/sna_cpu.c |
---|
0,0 → 1,116 |
/* |
* Copyright (c) 2013 Intel Corporation |
* |
* Permission is hereby granted, free of charge, to any person obtaining a |
* copy of this software and associated documentation files (the "Software"), |
* to deal in the Software without restriction, including without limitation |
* the rights to use, copy, modify, merge, publish, distribute, sublicense, |
* and/or sell copies of the Software, and to permit persons to whom the |
* Software is furnished to do so, subject to the following conditions: |
* |
* The above copyright notice and this permission notice (including the next |
* paragraph) shall be included in all copies or substantial portions of the |
* Software. |
* |
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL |
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, |
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE |
* SOFTWARE. |
* |
* Authors: |
* Chris Wilson <chris@chris-wilson.co.uk> |
* |
*/ |
#ifdef HAVE_CONFIG_H |
#include "config.h" |
#endif |
#include "sna.h" |
#include "sna_cpuid.h" |
#define xgetbv(index,eax,edx) \ |
__asm__ ("xgetbv" : "=a"(eax), "=d"(edx) : "c" (index)) |
#define has_YMM 0x1 |
unsigned sna_cpu_detect(void) |
{ |
unsigned max = __get_cpuid_max(BASIC_CPUID, NULL); |
unsigned int eax, ebx, ecx, edx; |
unsigned features = 0; |
unsigned extra = 0; |
if (max >= 1) { |
__cpuid(1, eax, ebx, ecx, edx); |
if (ecx & bit_SSE3) |
features |= SSE3; |
if (ecx & bit_SSSE3) |
features |= SSSE3; |
if (ecx & bit_SSE4_1) |
features |= SSE4_1; |
if (ecx & bit_SSE4_2) |
features |= SSE4_2; |
if (ecx & bit_OSXSAVE) { |
unsigned int bv_eax, bv_ecx; |
xgetbv(0, bv_eax, bv_ecx); |
if ((bv_eax & 6) == 6) |
extra |= has_YMM; |
} |
if ((extra & has_YMM) && (ecx & bit_AVX)) |
features |= AVX; |
if (edx & bit_MMX) |
features |= MMX; |
if (edx & bit_SSE) |
features |= SSE; |
if (edx & bit_SSE2) |
features |= SSE2; |
} |
if (max >= 7) { |
__cpuid_count(7, 0, eax, ebx, ecx, edx); |
if ((extra & has_YMM) && (ebx & bit_AVX2)) |
features |= AVX2; |
} |
return features; |
} |
char *sna_cpu_features_to_string(unsigned features, char *line) |
{ |
char *ret = line; |
#ifdef __x86_64__ |
line += sprintf (line, "x86-64"); |
#else |
line += sprintf (line, "x86"); |
#endif |
if (features & SSE2) |
line += sprintf (line, ", sse2"); |
if (features & SSE3) |
line += sprintf (line, ", sse3"); |
if (features & SSSE3) |
line += sprintf (line, ", ssse3"); |
if (features & SSE4_1) |
line += sprintf (line, ", sse4.1"); |
if (features & SSE4_2) |
line += sprintf (line, ", sse4.2"); |
if (features & AVX) |
line += sprintf (line, ", avx"); |
if (features & AVX2) |
line += sprintf (line, ", avx2"); |
return ret; |
} |
/drivers/video/Intel-2D/sna/sna_cpuid.h |
---|
0,0 → 1,86 |
/* |
* Copyright (c) 2013 Intel Corporation |
* |
* Permission is hereby granted, free of charge, to any person obtaining a |
* copy of this software and associated documentation files (the "Software"), |
* to deal in the Software without restriction, including without limitation |
* the rights to use, copy, modify, merge, publish, distribute, sublicense, |
* and/or sell copies of the Software, and to permit persons to whom the |
* Software is furnished to do so, subject to the following conditions: |
* |
* The above copyright notice and this permission notice (including the next |
* paragraph) shall be included in all copies or substantial portions of the |
* Software. |
* |
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL |
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, |
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE |
* SOFTWARE. |
* |
* Authors: |
* Chris Wilson <chris@chris-wilson.co.uk> |
* |
*/ |
/* Small wrapper around compiler specific implementation details of cpuid */ |
#ifndef SNA_CPUID_H |
#define SNA_CPUID_H |
#include "compiler.h" |
#if HAS_GCC(4, 4) /* for __cpuid_count() */ |
#include <cpuid.h> |
#else |
#define __get_cpuid_max(x, y) 0 |
#define __cpuid(level, a, b, c, d) |
#define __cpuid_count(level, count, a, b, c, d) |
#endif |
#define BASIC_CPUID 0x0 |
#define EXTENDED_CPUID 0x80000000 |
#ifndef bit_MMX |
#define bit_MMX (1 << 23) |
#endif |
#ifndef bit_SSE |
#define bit_SSE (1 << 25) |
#endif |
#ifndef bit_SSE2 |
#define bit_SSE2 (1 << 26) |
#endif |
#ifndef bit_SSE3 |
#define bit_SSE3 (1 << 0) |
#endif |
#ifndef bit_SSSE3 |
#define bit_SSSE3 (1 << 9) |
#endif |
#ifndef bit_SSE4_1 |
#define bit_SSE4_1 (1 << 19) |
#endif |
#ifndef bit_SSE4_2 |
#define bit_SSE4_2 (1 << 20) |
#endif |
#ifndef bit_OSXSAVE |
#define bit_OSXSAVE (1 << 27) |
#endif |
#ifndef bit_AVX |
#define bit_AVX (1 << 28) |
#endif |
#ifndef bit_AVX2 |
#define bit_AVX2 (1<<5) |
#endif |
#endif /* SNA_CPUID_H */ |
/drivers/video/Intel-2D/sna/sna_reg.h |
---|
0,0 → 1,82 |
#ifndef SNA_REG_H |
#define SNA_REG_H |
/* Flush */ |
#define MI_FLUSH (0x04<<23) |
#define MI_FLUSH_DW (0x26<<23) |
#define MI_WRITE_DIRTY_STATE (1<<4) |
#define MI_END_SCENE (1<<3) |
#define MI_GLOBAL_SNAPSHOT_COUNT_RESET (1<<3) |
#define MI_INHIBIT_RENDER_CACHE_FLUSH (1<<2) |
#define MI_STATE_INSTRUCTION_CACHE_FLUSH (1<<1) |
#define MI_INVALIDATE_MAP_CACHE (1<<0) |
/* broadwater flush bits */ |
#define BRW_MI_GLOBAL_SNAPSHOT_RESET (1 << 3) |
#define MI_BATCH_BUFFER_END (0xA << 23) |
/* Noop */ |
#define MI_NOOP 0x00 |
#define MI_NOOP_WRITE_ID (1<<22) |
#define MI_NOOP_ID_MASK (1<<22 - 1) |
/* Wait for Events */ |
#define MI_WAIT_FOR_EVENT (0x03<<23) |
#define MI_WAIT_FOR_PIPEB_SVBLANK (1<<18) |
#define MI_WAIT_FOR_PIPEA_SVBLANK (1<<17) |
#define MI_WAIT_FOR_OVERLAY_FLIP (1<<16) |
#define MI_WAIT_FOR_PIPEB_VBLANK (1<<7) |
#define MI_WAIT_FOR_PIPEB_SCAN_LINE_WINDOW (1<<5) |
#define MI_WAIT_FOR_PIPEA_VBLANK (1<<3) |
#define MI_WAIT_FOR_PIPEA_SCAN_LINE_WINDOW (1<<1) |
/* Set the scan line for MI_WAIT_FOR_PIPE?_SCAN_LINE_WINDOW */ |
#define MI_LOAD_SCAN_LINES_INCL (0x12<<23) |
#define MI_LOAD_SCAN_LINES_DISPLAY_PIPEA (0) |
#define MI_LOAD_SCAN_LINES_DISPLAY_PIPEB (0x1<<20) |
/* BLT commands */ |
#define BLT_WRITE_ALPHA (1<<21) |
#define BLT_WRITE_RGB (1<<20) |
#define BLT_SRC_TILED (1<<15) |
#define BLT_DST_TILED (1<<11) |
#define COLOR_BLT_CMD ((2<<29)|(0x40<<22)|(0x3)) |
#define XY_COLOR_BLT ((2<<29)|(0x50<<22)|(0x4)) |
#define XY_SETUP_BLT ((2<<29)|(1<<22)|6) |
#define XY_SETUP_MONO_PATTERN_SL_BLT ((2<<29)|(0x11<<22)|7) |
#define XY_SETUP_CLIP ((2<<29)|(3<<22)|1) |
#define XY_SCANLINE_BLT ((2<<29)|(0x25<<22)|1) |
#define XY_TEXT_IMMEDIATE_BLT ((2<<29)|(0x31<<22)|(1<<16)) |
#define XY_SRC_COPY_BLT_CMD ((2<<29)|(0x53<<22)|6) |
#define SRC_COPY_BLT_CMD ((2<<29)|(0x43<<22)|0x4) |
#define XY_PAT_BLT ((2<<29)|(0x51<<22)|0x4) |
#define XY_PAT_BLT_IMMEDIATE ((2<<29)|(0x72<<22)) |
#define XY_MONO_PAT ((0x2<<29)|(0x52<<22)|0x7) |
#define XY_MONO_SRC_COPY ((0x2<<29)|(0x54<<22)|(0x6)) |
#define XY_MONO_SRC_COPY_IMM ((0x2<<29)|(0x71<<22)) |
#define XY_FULL_MONO_PATTERN_BLT ((0x2<<29)|(0x57<<22)|0xa) |
#define XY_FULL_MONO_PATTERN_MONO_SRC_BLT ((0x2<<29)|(0x58<<22)|0xa) |
/* FLUSH commands */ |
#define BRW_3D(Pipeline,Opcode,Subopcode) \ |
((3 << 29) | \ |
((Pipeline) << 27) | \ |
((Opcode) << 24) | \ |
((Subopcode) << 16)) |
#define PIPE_CONTROL BRW_3D(3, 2, 0) |
#define PIPE_CONTROL_NOWRITE (0 << 14) |
#define PIPE_CONTROL_WRITE_QWORD (1 << 14) |
#define PIPE_CONTROL_WRITE_DEPTH (2 << 14) |
#define PIPE_CONTROL_WRITE_TIME (3 << 14) |
#define PIPE_CONTROL_DEPTH_STALL (1 << 13) |
#define PIPE_CONTROL_WC_FLUSH (1 << 12) |
#define PIPE_CONTROL_IS_FLUSH (1 << 11) |
#define PIPE_CONTROL_TC_FLUSH (1 << 10) |
#define PIPE_CONTROL_NOTIFY_ENABLE (1 << 8) |
#define PIPE_CONTROL_GLOBAL_GTT (1 << 2) |
#define PIPE_CONTROL_LOCAL_PGTT (0 << 2) |
#define PIPE_CONTROL_DEPTH_CACHE_FLUSH (1 << 0) |
#endif |
/drivers/video/Intel-2D/sna/sna_render.h |
---|
0,0 → 1,786 |
#ifndef SNA_RENDER_H |
#define SNA_RENDER_H |
#include "compiler.h" |
#include <stdbool.h> |
#include <stdint.h> |
#define GRADIENT_CACHE_SIZE 16 |
#define GXinvalid 0xff |
#define HW_BIT_BLIT (1<<0) /* BGRX blitter */ |
#define HW_TEX_BLIT (1<<1) /* stretch blit */ |
#define HW_VID_BLIT (1<<2) /* planar and packed video */ |
struct sna; |
struct sna_glyph; |
struct sna_video; |
struct sna_video_frame; |
struct brw_compile; |
struct sna_composite_rectangles { |
struct sna_coordinate { |
int16_t x, y; |
} src, mask, dst; |
int16_t width, height; |
}; |
struct sna_composite_op { |
fastcall void (*blt)(struct sna *sna, const struct sna_composite_op *op, |
const struct sna_composite_rectangles *r); |
void (*done)(struct sna *sna, const struct sna_composite_op *op); |
struct sna_damage **damage; |
uint32_t op; |
struct { |
PixmapPtr pixmap; |
CARD32 format; |
struct kgem_bo *bo; |
int16_t x, y; |
uint16_t width, height; |
} dst; |
struct sna_composite_channel { |
struct kgem_bo *bo; |
PictTransform *transform; |
uint16_t width; |
uint16_t height; |
uint32_t pict_format; |
uint32_t card_format; |
uint32_t filter; |
uint32_t repeat; |
uint32_t is_affine : 1; |
uint32_t is_solid : 1; |
uint32_t is_linear : 1; |
uint32_t is_opaque : 1; |
uint32_t alpha_fixup : 1; |
uint32_t rb_reversed : 1; |
int16_t offset[2]; |
float scale[2]; |
// pixman_transform_t embedded_transform; |
union { |
struct { |
float dx, dy, offset; |
} linear; |
struct { |
uint32_t pixel; |
} gen2; |
struct gen3_shader_channel { |
int type; |
uint32_t mode; |
uint32_t constants; |
} gen3; |
} u; |
} src, mask; |
uint32_t is_affine : 1; |
uint32_t has_component_alpha : 1; |
uint32_t need_magic_ca_pass : 1; |
uint32_t rb_reversed : 1; |
int16_t floats_per_vertex; |
int16_t floats_per_rect; |
fastcall void (*prim_emit)(struct sna *sna, |
const struct sna_composite_op *op, |
const struct sna_composite_rectangles *r); |
struct sna_composite_redirect { |
struct kgem_bo *real_bo; |
struct sna_damage **real_damage, *damage; |
BoxRec box; |
} redirect; |
union { |
struct sna_blt_state { |
PixmapPtr src_pixmap; |
int16_t sx, sy; |
uint32_t inplace :1; |
uint32_t overwrites:1; |
uint32_t bpp : 6; |
uint32_t cmd; |
uint32_t br13; |
uint32_t pitch[2]; |
uint32_t pixel; |
struct kgem_bo *bo[2]; |
} blt; |
struct { |
float constants[8]; |
uint32_t num_constants; |
} gen3; |
struct { |
int wm_kernel; |
int ve_id; |
} gen4; |
struct { |
int16_t wm_kernel; |
int16_t ve_id; |
} gen5; |
struct { |
uint32_t flags; |
} gen6; |
struct { |
uint32_t flags; |
} gen7; |
} u; |
void *priv; |
}; |
struct sna_copy_op { |
struct sna_composite_op base; |
void (*blt)(struct sna *sna, const struct sna_copy_op *op, |
int16_t sx, int16_t sy, |
int16_t w, int16_t h, |
int16_t dx, int16_t dy); |
void (*done)(struct sna *sna, const struct sna_copy_op *op); |
}; |
struct sna_render { |
int active; |
int caps; |
int max_3d_size; |
int max_3d_pitch; |
unsigned prefer_gpu; |
#define PREFER_GPU_BLT 0x1 |
#define PREFER_GPU_RENDER 0x2 |
#define PREFER_GPU_SPANS 0x4 |
bool (*composite)(struct sna *sna, uint8_t op, |
PicturePtr dst, PicturePtr src, PicturePtr mask, |
int16_t src_x, int16_t src_y, |
int16_t msk_x, int16_t msk_y, |
int16_t dst_x, int16_t dst_y, |
int16_t w, int16_t h, |
struct sna_composite_op *tmp); |
#if 0 |
bool (*check_composite_spans)(struct sna *sna, uint8_t op, |
PicturePtr dst, PicturePtr src, |
int16_t w, int16_t h, unsigned flags); |
bool (*composite_spans)(struct sna *sna, uint8_t op, |
PicturePtr dst, PicturePtr src, |
int16_t src_x, int16_t src_y, |
int16_t dst_x, int16_t dst_y, |
int16_t w, int16_t h, |
unsigned flags, |
struct sna_composite_spans_op *tmp); |
#define COMPOSITE_SPANS_RECTILINEAR 0x1 |
#define COMPOSITE_SPANS_INPLACE_HINT 0x2 |
bool (*video)(struct sna *sna, |
struct sna_video *video, |
struct sna_video_frame *frame, |
RegionPtr dstRegion, |
PixmapPtr pixmap); |
bool (*fill_boxes)(struct sna *sna, |
CARD8 op, |
PictFormat format, |
const xRenderColor *color, |
PixmapPtr dst, struct kgem_bo *dst_bo, |
const BoxRec *box, int n); |
bool (*fill)(struct sna *sna, uint8_t alu, |
PixmapPtr dst, struct kgem_bo *dst_bo, |
uint32_t color, |
struct sna_fill_op *tmp); |
bool (*fill_one)(struct sna *sna, PixmapPtr dst, struct kgem_bo *dst_bo, |
uint32_t color, |
int16_t x1, int16_t y1, int16_t x2, int16_t y2, |
uint8_t alu); |
bool (*clear)(struct sna *sna, PixmapPtr dst, struct kgem_bo *dst_bo); |
bool (*copy_boxes)(struct sna *sna, uint8_t alu, |
PixmapPtr src, struct kgem_bo *src_bo, int16_t src_dx, int16_t src_dy, |
PixmapPtr dst, struct kgem_bo *dst_bo, int16_t dst_dx, int16_t dst_dy, |
const BoxRec *box, int n, unsigned flags); |
#define COPY_LAST 0x1 |
#define COPY_SYNC 0x2 |
#endif |
bool (*blit_tex)(struct sna *sna, |
uint8_t op, bool scale, |
PixmapPtr src, struct kgem_bo *src_bo, |
PixmapPtr mask,struct kgem_bo *mask_bo, |
PixmapPtr dst, struct kgem_bo *dst_bo, |
int32_t src_x, int32_t src_y, |
int32_t msk_x, int32_t msk_y, |
int32_t dst_x, int32_t dst_y, |
int32_t width, int32_t height, |
struct sna_composite_op *tmp); |
bool (*copy)(struct sna *sna, uint8_t alu, |
PixmapPtr src, struct kgem_bo *src_bo, |
PixmapPtr dst, struct kgem_bo *dst_bo, |
struct sna_copy_op *op); |
void (*flush)(struct sna *sna); |
void (*reset)(struct sna *sna); |
void (*fini)(struct sna *sna); |
#if 0 |
struct sna_alpha_cache { |
struct kgem_bo *cache_bo; |
struct kgem_bo *bo[256+7]; |
} alpha_cache; |
struct sna_solid_cache { |
struct kgem_bo *cache_bo; |
struct kgem_bo *bo[1024]; |
uint32_t color[1025]; |
int last; |
int size; |
int dirty; |
} solid_cache; |
struct { |
struct sna_gradient_cache { |
struct kgem_bo *bo; |
int nstops; |
PictGradientStop *stops; |
} cache[GRADIENT_CACHE_SIZE]; |
int size; |
} gradient_cache; |
struct sna_glyph_cache{ |
PicturePtr picture; |
struct sna_glyph **glyphs; |
uint16_t count; |
uint16_t evict; |
} glyph[2]; |
pixman_image_t *white_image; |
PicturePtr white_picture; |
#if HAS_PIXMAN_GLYPHS |
pixman_glyph_cache_t *glyph_cache; |
#endif |
#endif |
uint16_t vb_id; |
uint16_t vertex_offset; |
uint16_t vertex_start; |
uint16_t vertex_index; |
uint16_t vertex_used; |
uint16_t vertex_size; |
uint16_t vertex_reloc[16]; |
int nvertex_reloc; |
struct kgem_bo *vbo; |
float *vertices; |
float vertex_data[1024]; |
}; |
struct gen2_render_state { |
uint32_t target; |
bool need_invariant; |
uint32_t logic_op_enabled; |
uint32_t ls1, ls2, vft; |
uint32_t diffuse; |
uint32_t specular; |
}; |
struct gen3_render_state { |
uint32_t current_dst; |
bool need_invariant; |
uint32_t tex_count; |
uint32_t last_drawrect_limit; |
uint32_t last_target; |
uint32_t last_blend; |
uint32_t last_constants; |
uint32_t last_sampler; |
uint32_t last_shader; |
uint32_t last_diffuse; |
uint32_t last_specular; |
uint16_t last_vertex_offset; |
uint16_t floats_per_vertex; |
uint16_t last_floats_per_vertex; |
uint32_t tex_map[4]; |
uint32_t tex_handle[2]; |
uint32_t tex_delta[2]; |
}; |
struct gen4_render_state { |
struct kgem_bo *general_bo; |
uint32_t vs; |
uint32_t sf; |
uint32_t wm; |
uint32_t cc; |
int ve_id; |
uint32_t drawrect_offset; |
uint32_t drawrect_limit; |
uint32_t last_pipelined_pointers; |
uint16_t last_primitive; |
int16_t floats_per_vertex; |
uint16_t surface_table; |
bool needs_invariant; |
bool needs_urb; |
}; |
struct gen5_render_state { |
struct kgem_bo *general_bo; |
uint32_t vs; |
uint32_t sf[2]; |
uint32_t wm; |
uint32_t cc; |
int ve_id; |
uint32_t drawrect_offset; |
uint32_t drawrect_limit; |
uint32_t last_pipelined_pointers; |
uint16_t last_primitive; |
int16_t floats_per_vertex; |
uint16_t surface_table; |
bool needs_invariant; |
}; |
enum { |
GEN6_WM_KERNEL_NOMASK = 0, |
GEN6_WM_KERNEL_NOMASK_P, |
GEN6_WM_KERNEL_MASK, |
GEN6_WM_KERNEL_MASK_P, |
GEN6_WM_KERNEL_MASKCA, |
GEN6_WM_KERNEL_MASKCA_P, |
GEN6_WM_KERNEL_MASKSA, |
GEN6_WM_KERNEL_MASKSA_P, |
GEN6_WM_KERNEL_OPACITY, |
GEN6_WM_KERNEL_OPACITY_P, |
GEN6_WM_KERNEL_VIDEO_PLANAR, |
GEN6_WM_KERNEL_VIDEO_PACKED, |
GEN6_KERNEL_COUNT |
}; |
struct gen6_render_state { |
const struct gt_info *info; |
struct kgem_bo *general_bo; |
uint32_t vs_state; |
uint32_t sf_state; |
uint32_t sf_mask_state; |
uint32_t wm_state; |
uint32_t wm_kernel[GEN6_KERNEL_COUNT][3]; |
uint32_t cc_blend; |
uint32_t drawrect_offset; |
uint32_t drawrect_limit; |
uint32_t blend; |
uint32_t samplers; |
uint32_t kernel; |
uint16_t num_sf_outputs; |
uint16_t ve_id; |
uint16_t last_primitive; |
int16_t floats_per_vertex; |
uint16_t surface_table; |
bool needs_invariant; |
bool first_state_packet; |
}; |
enum { |
GEN7_WM_KERNEL_NOMASK = 0, |
GEN7_WM_KERNEL_NOMASK_P, |
GEN7_WM_KERNEL_MASK, |
GEN7_WM_KERNEL_MASK_P, |
GEN7_WM_KERNEL_MASKCA, |
GEN7_WM_KERNEL_MASKCA_P, |
GEN7_WM_KERNEL_MASKSA, |
GEN7_WM_KERNEL_MASKSA_P, |
GEN7_WM_KERNEL_OPACITY, |
GEN7_WM_KERNEL_OPACITY_P, |
GEN7_WM_KERNEL_VIDEO_PLANAR, |
GEN7_WM_KERNEL_VIDEO_PACKED, |
GEN7_WM_KERNEL_COUNT |
}; |
struct gen7_render_state { |
const struct gt_info *info; |
struct kgem_bo *general_bo; |
uint32_t vs_state; |
uint32_t sf_state; |
uint32_t sf_mask_state; |
uint32_t wm_state; |
uint32_t wm_kernel[GEN7_WM_KERNEL_COUNT][3]; |
uint32_t cc_blend; |
uint32_t drawrect_offset; |
uint32_t drawrect_limit; |
uint32_t blend; |
uint32_t samplers; |
uint32_t kernel; |
uint16_t num_sf_outputs; |
uint16_t ve_id; |
uint16_t last_primitive; |
int16_t floats_per_vertex; |
uint16_t surface_table; |
bool needs_invariant; |
bool emit_flush; |
}; |
struct sna_static_stream { |
uint32_t size, used; |
uint8_t *data; |
}; |
int sna_static_stream_init(struct sna_static_stream *stream); |
uint32_t sna_static_stream_add(struct sna_static_stream *stream, |
const void *data, uint32_t len, uint32_t align); |
void *sna_static_stream_map(struct sna_static_stream *stream, |
uint32_t len, uint32_t align); |
uint32_t sna_static_stream_offsetof(struct sna_static_stream *stream, |
void *ptr); |
unsigned sna_static_stream_compile_sf(struct sna *sna, |
struct sna_static_stream *stream, |
bool (*compile)(struct brw_compile *)); |
unsigned sna_static_stream_compile_wm(struct sna *sna, |
struct sna_static_stream *stream, |
bool (*compile)(struct brw_compile *, int), |
int width); |
struct kgem_bo *sna_static_stream_fini(struct sna *sna, |
struct sna_static_stream *stream); |
struct kgem_bo * |
sna_render_get_solid(struct sna *sna, |
uint32_t color); |
void |
sna_render_flush_solid(struct sna *sna); |
uint32_t sna_rgba_for_color(uint32_t color, int depth); |
uint32_t sna_rgba_to_color(uint32_t rgba, uint32_t format); |
bool sna_get_rgba_from_pixel(uint32_t pixel, |
uint16_t *red, |
uint16_t *green, |
uint16_t *blue, |
uint16_t *alpha, |
uint32_t format); |
bool sna_picture_is_solid(PicturePtr picture, uint32_t *color); |
const char *no_render_init(struct sna *sna); |
const char *gen2_render_init(struct sna *sna, const char *backend); |
const char *gen3_render_init(struct sna *sna, const char *backend); |
const char *gen4_render_init(struct sna *sna, const char *backend); |
const char *gen5_render_init(struct sna *sna, const char *backend); |
const char *gen6_render_init(struct sna *sna, const char *backend); |
const char *gen7_render_init(struct sna *sna, const char *backend); |
#if 0 |
bool sna_tiling_composite(uint32_t op, |
PicturePtr src, |
PicturePtr mask, |
PicturePtr dst, |
int16_t src_x, int16_t src_y, |
int16_t mask_x, int16_t mask_y, |
int16_t dst_x, int16_t dst_y, |
int16_t width, int16_t height, |
struct sna_composite_op *tmp); |
bool sna_tiling_fill_boxes(struct sna *sna, |
CARD8 op, |
PictFormat format, |
const xRenderColor *color, |
PixmapPtr dst, struct kgem_bo *dst_bo, |
const BoxRec *box, int n); |
bool sna_tiling_copy_boxes(struct sna *sna, uint8_t alu, |
PixmapPtr src, struct kgem_bo *src_bo, int16_t src_dx, int16_t src_dy, |
PixmapPtr dst, struct kgem_bo *dst_bo, int16_t dst_dx, int16_t dst_dy, |
const BoxRec *box, int n); |
bool sna_tiling_blt_copy_boxes(struct sna *sna, uint8_t alu, |
struct kgem_bo *src_bo, int16_t src_dx, int16_t src_dy, |
struct kgem_bo *dst_bo, int16_t dst_dx, int16_t dst_dy, |
int bpp, const BoxRec *box, int nbox); |
bool sna_blt_composite(struct sna *sna, |
uint32_t op, |
PicturePtr src, |
PicturePtr dst, |
int16_t src_x, int16_t src_y, |
int16_t dst_x, int16_t dst_y, |
int16_t width, int16_t height, |
struct sna_composite_op *tmp, |
bool fallback); |
bool sna_blt_composite__convert(struct sna *sna, |
int x, int y, |
int width, int height, |
struct sna_composite_op *tmp); |
bool sna_blt_fill(struct sna *sna, uint8_t alu, |
struct kgem_bo *bo, |
int bpp, |
uint32_t pixel, |
struct sna_fill_op *fill); |
bool sna_blt_copy(struct sna *sna, uint8_t alu, |
struct kgem_bo *src, |
struct kgem_bo *dst, |
int bpp, |
struct sna_copy_op *copy); |
bool sna_blt_fill_boxes(struct sna *sna, uint8_t alu, |
struct kgem_bo *bo, |
int bpp, |
uint32_t pixel, |
const BoxRec *box, int n); |
bool sna_blt_copy_boxes(struct sna *sna, uint8_t alu, |
struct kgem_bo *src_bo, int16_t src_dx, int16_t src_dy, |
struct kgem_bo *dst_bo, int16_t dst_dx, int16_t dst_dy, |
int bpp, |
const BoxRec *box, int n); |
bool sna_blt_copy_boxes_fallback(struct sna *sna, uint8_t alu, |
PixmapPtr src, struct kgem_bo *src_bo, int16_t src_dx, int16_t src_dy, |
PixmapPtr dst, struct kgem_bo *dst_bo, int16_t dst_dx, int16_t dst_dy, |
const BoxRec *box, int nbox); |
bool _sna_get_pixel_from_rgba(uint32_t *pixel, |
uint16_t red, |
uint16_t green, |
uint16_t blue, |
uint16_t alpha, |
uint32_t format); |
static inline bool |
sna_get_pixel_from_rgba(uint32_t * pixel, |
uint16_t red, |
uint16_t green, |
uint16_t blue, |
uint16_t alpha, |
uint32_t format) |
{ |
switch (format) { |
case PICT_x8r8g8b8: |
alpha = 0xffff; |
/* fall through to re-use a8r8g8b8 expansion */ |
case PICT_a8r8g8b8: |
*pixel = ((alpha >> 8 << 24) | |
(red >> 8 << 16) | |
(green & 0xff00) | |
(blue >> 8)); |
return TRUE; |
case PICT_a8: |
*pixel = alpha >> 8; |
return TRUE; |
} |
return _sna_get_pixel_from_rgba(pixel, red, green, blue, alpha, format); |
} |
struct kgem_bo * |
__sna_render_pixmap_bo(struct sna *sna, |
PixmapPtr pixmap, |
const BoxRec *box, |
bool blt); |
int |
sna_render_pixmap_bo(struct sna *sna, |
struct sna_composite_channel *channel, |
PixmapPtr pixmap, |
int16_t x, int16_t y, |
int16_t w, int16_t h, |
int16_t dst_x, int16_t dst_y); |
bool |
sna_render_pixmap_partial(struct sna *sna, |
PixmapPtr pixmap, |
struct kgem_bo *bo, |
struct sna_composite_channel *channel, |
int16_t x, int16_t y, |
int16_t w, int16_t h); |
int |
sna_render_picture_extract(struct sna *sna, |
PicturePtr picture, |
struct sna_composite_channel *channel, |
int16_t x, int16_t y, |
int16_t w, int16_t h, |
int16_t dst_x, int16_t dst_y); |
int |
sna_render_picture_approximate_gradient(struct sna *sna, |
PicturePtr picture, |
struct sna_composite_channel *channel, |
int16_t x, int16_t y, |
int16_t w, int16_t h, |
int16_t dst_x, int16_t dst_y); |
int |
sna_render_picture_fixup(struct sna *sna, |
PicturePtr picture, |
struct sna_composite_channel *channel, |
int16_t x, int16_t y, |
int16_t w, int16_t h, |
int16_t dst_x, int16_t dst_y); |
int |
sna_render_picture_convert(struct sna *sna, |
PicturePtr picture, |
struct sna_composite_channel *channel, |
PixmapPtr pixmap, |
int16_t x, int16_t y, |
int16_t w, int16_t h, |
int16_t dst_x, int16_t dst_y, |
bool fixup_alpha); |
inline static void sna_render_composite_redirect_init(struct sna_composite_op *op) |
{ |
struct sna_composite_redirect *t = &op->redirect; |
t->real_bo = NULL; |
t->damage = NULL; |
} |
bool |
sna_render_composite_redirect(struct sna *sna, |
struct sna_composite_op *op, |
int x, int y, int width, int height, |
bool partial); |
void |
sna_render_composite_redirect_done(struct sna *sna, |
const struct sna_composite_op *op); |
bool |
sna_composite_mask_is_opaque(PicturePtr mask); |
#endif |
void sna_vertex_init(struct sna *sna); |
static inline void sna_vertex_lock(struct sna_render *r) |
{ |
// pthread_mutex_lock(&r->lock); |
} |
static inline void sna_vertex_acquire__locked(struct sna_render *r) |
{ |
r->active++; |
} |
static inline void sna_vertex_unlock(struct sna_render *r) |
{ |
// pthread_mutex_unlock(&r->lock); |
} |
static inline void sna_vertex_release__locked(struct sna_render *r) |
{ |
assert(r->active > 0); |
--r->active; |
// if (--r->active == 0) |
// pthread_cond_signal(&r->wait); |
} |
static inline bool sna_vertex_wait__locked(struct sna_render *r) |
{ |
bool was_active = r->active; |
// while (r->active) |
// pthread_cond_wait(&r->wait, &r->lock); |
return was_active; |
} |
#define alphaless(format) PICT_FORMAT(PICT_FORMAT_BPP(format), \ |
PICT_FORMAT_TYPE(format), \ |
0, \ |
PICT_FORMAT_R(format), \ |
PICT_FORMAT_G(format), \ |
PICT_FORMAT_B(format)) |
static bool |
gen3_blit_tex(struct sna *sna, |
uint8_t op, bool scale, |
PixmapPtr src, struct kgem_bo *src_bo, |
PixmapPtr mask,struct kgem_bo *mask_bo, |
PixmapPtr dst, struct kgem_bo *dst_bo, |
int32_t src_x, int32_t src_y, |
int32_t msk_x, int32_t msk_y, |
int32_t dst_x, int32_t dst_y, |
int32_t width, int32_t height, |
struct sna_composite_op *tmp); |
static bool |
gen4_blit_tex(struct sna *sna, |
uint8_t op, bool scale, |
PixmapPtr src, struct kgem_bo *src_bo, |
PixmapPtr mask,struct kgem_bo *mask_bo, |
PixmapPtr dst, struct kgem_bo *dst_bo, |
int32_t src_x, int32_t src_y, |
int32_t msk_x, int32_t msk_y, |
int32_t dst_x, int32_t dst_y, |
int32_t width, int32_t height, |
struct sna_composite_op *tmp); |
static bool |
gen5_blit_tex(struct sna *sna, |
uint8_t op, bool scale, |
PixmapPtr src, struct kgem_bo *src_bo, |
PixmapPtr mask,struct kgem_bo *mask_bo, |
PixmapPtr dst, struct kgem_bo *dst_bo, |
int32_t src_x, int32_t src_y, |
int32_t msk_x, int32_t msk_y, |
int32_t dst_x, int32_t dst_y, |
int32_t width, int32_t height, |
struct sna_composite_op *tmp); |
static bool |
gen6_blit_tex(struct sna *sna, |
uint8_t op, bool scale, |
PixmapPtr src, struct kgem_bo *src_bo, |
PixmapPtr mask,struct kgem_bo *mask_bo, |
PixmapPtr dst, struct kgem_bo *dst_bo, |
int32_t src_x, int32_t src_y, |
int32_t msk_x, int32_t msk_y, |
int32_t dst_x, int32_t dst_y, |
int32_t width, int32_t height, |
struct sna_composite_op *tmp); |
static bool |
gen7_blit_tex(struct sna *sna, |
uint8_t op, bool scale, |
PixmapPtr src, struct kgem_bo *src_bo, |
PixmapPtr mask,struct kgem_bo *mask_bo, |
PixmapPtr dst, struct kgem_bo *dst_bo, |
int32_t src_x, int32_t src_y, |
int32_t msk_x, int32_t msk_y, |
int32_t dst_x, int32_t dst_y, |
int32_t width, int32_t height, |
struct sna_composite_op *tmp); |
#endif /* SNA_RENDER_H */ |
/drivers/video/Intel-2D/sna/sna_render_inline.h |
---|
0,0 → 1,70 |
#ifndef SNA_RENDER_INLINE_H |
#define SNA_RENDER_INLINE_H |
static inline bool need_tiling(struct sna *sna, int16_t width, int16_t height) |
{ |
/* Is the damage area too large to fit in 3D pipeline, |
* and so do we need to split the operation up into tiles? |
*/ |
return (width > sna->render.max_3d_size || |
height > sna->render.max_3d_size); |
} |
static inline bool need_redirect(struct sna *sna, PixmapPtr dst) |
{ |
/* Is the pixmap too large to render to? */ |
return (dst->drawable.width > sna->render.max_3d_size || |
dst->drawable.height > sna->render.max_3d_size); |
} |
static force_inline float pack_2s(int16_t x, int16_t y) |
{ |
union { |
struct sna_coordinate p; |
float f; |
} u; |
u.p.x = x; |
u.p.y = y; |
return u.f; |
} |
static force_inline int vertex_space(struct sna *sna) |
{ |
return sna->render.vertex_size - sna->render.vertex_used; |
} |
static force_inline void vertex_emit(struct sna *sna, float v) |
{ |
assert(sna->render.vertex_used < sna->render.vertex_size); |
sna->render.vertices[sna->render.vertex_used++] = v; |
} |
static force_inline void vertex_emit_2s(struct sna *sna, int16_t x, int16_t y) |
{ |
vertex_emit(sna, pack_2s(x, y)); |
} |
static force_inline int batch_space(struct sna *sna) |
{ |
assert(sna->kgem.nbatch <= KGEM_BATCH_SIZE(&sna->kgem)); |
assert(sna->kgem.nbatch + KGEM_BATCH_RESERVED <= sna->kgem.surface); |
return sna->kgem.surface - sna->kgem.nbatch - KGEM_BATCH_RESERVED; |
} |
static force_inline void batch_emit(struct sna *sna, uint32_t dword) |
{ |
assert(sna->kgem.mode != KGEM_NONE); |
assert(sna->kgem.nbatch + KGEM_BATCH_RESERVED < sna->kgem.surface); |
sna->kgem.batch[sna->kgem.nbatch++] = dword; |
} |
static force_inline void batch_emit_float(struct sna *sna, float f) |
{ |
union { |
uint32_t dw; |
float f; |
} u; |
u.f = f; |
batch_emit(sna, u.dw); |
} |
#endif /* SNA_RENDER_INLINE_H */ |
/drivers/video/Intel-2D/sna/sna_stream.c |
---|
0,0 → 1,156 |
/* |
* Copyright © 2011 Intel Corporation |
* |
* Permission is hereby granted, free of charge, to any person obtaining a |
* copy of this software and associated documentation files (the "Software"), |
* to deal in the Software without restriction, including without limitation |
* the rights to use, copy, modify, merge, publish, distribute, sublicense, |
* and/or sell copies of the Software, and to permit persons to whom the |
* Software is furnished to do so, subject to the following conditions: |
* |
* The above copyright notice and this permission notice (including the next |
* paragraph) shall be included in all copies or substantial portions of the |
* Software. |
* |
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL |
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, |
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE |
* SOFTWARE. |
* |
* Authors: |
* Chris Wilson <chris@chris-wilson.co.uk> |
* |
*/ |
#include "sna.h" |
#include "sna_render.h" |
#include "brw/brw.h" |
#include <fcntl.h> |
int sna_static_stream_init(struct sna_static_stream *stream) |
{ |
stream->used = 0; |
stream->size = 64*1024; |
stream->data = malloc(stream->size); |
return stream->data != NULL; |
} |
static uint32_t sna_static_stream_alloc(struct sna_static_stream *stream, |
uint32_t len, uint32_t align) |
{ |
uint32_t offset = ALIGN(stream->used, align); |
uint32_t size = offset + len; |
if (size > stream->size) { |
do |
stream->size *= 2; |
while (stream->size < size); |
stream->data = realloc(stream->data, stream->size); |
} |
stream->used = size; |
return offset; |
} |
uint32_t sna_static_stream_add(struct sna_static_stream *stream, |
const void *data, uint32_t len, uint32_t align) |
{ |
uint32_t offset = sna_static_stream_alloc(stream, len, align); |
memcpy(stream->data + offset, data, len); |
return offset; |
} |
void *sna_static_stream_map(struct sna_static_stream *stream, |
uint32_t len, uint32_t align) |
{ |
uint32_t offset = sna_static_stream_alloc(stream, len, align); |
return memset(stream->data + offset, 0, len); |
} |
uint32_t sna_static_stream_offsetof(struct sna_static_stream *stream, void *ptr) |
{ |
return (uint8_t *)ptr - stream->data; |
} |
struct kgem_bo *sna_static_stream_fini(struct sna *sna, |
struct sna_static_stream *stream) |
{ |
struct kgem_bo *bo; |
DBG(("uploaded %d bytes of static state\n", stream->used)); |
if (DEBUG_DUMP) |
{ |
int fd = open("/tmp1/1/static.bin", O_CREAT|O_WRONLY|O_BINARY); |
if (fd != -1) { |
write(fd, stream->data, stream->used); |
close(fd); |
} |
else |
{ |
printf("SNA: failed to write static stream\n"); |
asm volatile("int3"); |
} |
} |
bo = kgem_create_linear(&sna->kgem, stream->used, 0); |
if (bo && !kgem_bo_write(&sna->kgem, bo, stream->data, stream->used)) { |
kgem_bo_destroy(&sna->kgem, bo); |
return NULL; |
} |
free(stream->data); |
return bo; |
} |
unsigned |
sna_static_stream_compile_sf(struct sna *sna, |
struct sna_static_stream *stream, |
bool (*compile)(struct brw_compile *)) |
{ |
struct brw_compile p; |
brw_compile_init(&p, sna->kgem.gen, |
sna_static_stream_map(stream, |
64*sizeof(uint32_t), 64)); |
if (!compile(&p)) { |
stream->used -= 64*sizeof(uint32_t); |
return 0; |
} |
assert(p.nr_insn*sizeof(struct brw_instruction) <= 64*sizeof(uint32_t)); |
stream->used -= 64*sizeof(uint32_t) - p.nr_insn*sizeof(struct brw_instruction); |
return sna_static_stream_offsetof(stream, p.store); |
} |
unsigned |
sna_static_stream_compile_wm(struct sna *sna, |
struct sna_static_stream *stream, |
bool (*compile)(struct brw_compile *, int), |
int dispatch_width) |
{ |
struct brw_compile p; |
brw_compile_init(&p, sna->kgem.gen, |
sna_static_stream_map(stream, |
256*sizeof(uint32_t), 64)); |
if (!compile(&p, dispatch_width)) { |
stream->used -= 256*sizeof(uint32_t); |
return 0; |
} |
assert(p.nr_insn*sizeof(struct brw_instruction) <= 256*sizeof(uint32_t)); |
stream->used -= 256*sizeof(uint32_t) - p.nr_insn*sizeof(struct brw_instruction); |
return sna_static_stream_offsetof(stream, p.store); |
} |
/drivers/video/Intel-2D/sna/sna_transform.c |
---|
0,0 → 1,147 |
/* |
* Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas. All Rights Reserved. |
* Copyright (c) 2005 Jesse Barnes <jbarnes@virtuousgeek.org> |
* Copyright © 2010 Intel Corporation |
* |
* Permission is hereby granted, free of charge, to any person obtaining a |
* copy of this software and associated documentation files (the "Software"), |
* to deal in the Software without restriction, including without limitation |
* the rights to use, copy, modify, merge, publish, distribute, sublicense, |
* and/or sell copies of the Software, and to permit persons to whom the |
* Software is furnished to do so, subject to the following conditions: |
* |
* The above copyright notice and this permission notice (including the next |
* paragraph) shall be included in all copies or substantial portions of the |
* Software. |
* |
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL |
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, |
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE |
* SOFTWARE. |
* |
* Authors: |
* Jesse Barns <jbarnes@virtuousgeek.org> |
* Chris Wilson <chris@chris-wilson.co.uk> |
*/ |
#ifdef HAVE_CONFIG_H |
#include "config.h" |
#endif |
#include "sna.h" |
#define pixman_fixed_e ((pixman_fixed_t) 1) |
#define pixman_fixed_1 (pixman_int_to_fixed(1)) |
#define pixman_fixed_1_minus_e (pixman_fixed_1 - pixman_fixed_e) |
#define pixman_fixed_to_int(f) ((int) ((f) >> 16)) |
#define pixman_int_to_fixed(i) ((pixman_fixed_t) ((i) << 16)) |
#define pixman_fixed_fraction(f) ((f) & pixman_fixed_1_minus_e) |
#define IntToxFixed(i) pixman_int_to_fixed(i) |
/** |
* Returns whether the provided transform is affine. |
* |
* transform may be null. |
*/ |
bool sna_transform_is_affine(const PictTransform *t) |
{ |
if (t == NULL) |
return true; |
return t->matrix[2][0] == 0 && t->matrix[2][1] == 0; |
} |
bool |
sna_transform_is_translation(const PictTransform *t, |
pixman_fixed_t *tx, |
pixman_fixed_t *ty) |
{ |
if (t == NULL) { |
*tx = *ty = 0; |
return true; |
} |
if (t->matrix[0][0] != IntToxFixed(1) || |
t->matrix[0][1] != 0 || |
t->matrix[1][0] != 0 || |
t->matrix[1][1] != IntToxFixed(1) || |
t->matrix[2][0] != 0 || |
t->matrix[2][1] != 0 || |
t->matrix[2][2] != IntToxFixed(1)) |
return false; |
*tx = t->matrix[0][2]; |
*ty = t->matrix[1][2]; |
return true; |
} |
bool |
sna_transform_is_integer_translation(const PictTransform *t, int16_t *tx, int16_t *ty) |
{ |
if (t == NULL) { |
*tx = *ty = 0; |
return true; |
} |
if (t->matrix[0][0] != IntToxFixed(1) || |
t->matrix[0][1] != 0 || |
t->matrix[1][0] != 0 || |
t->matrix[1][1] != IntToxFixed(1) || |
t->matrix[2][0] != 0 || |
t->matrix[2][1] != 0 || |
t->matrix[2][2] != IntToxFixed(1)) |
return false; |
if (pixman_fixed_fraction(t->matrix[0][2]) || |
pixman_fixed_fraction(t->matrix[1][2])) |
return false; |
*tx = pixman_fixed_to_int(t->matrix[0][2]); |
*ty = pixman_fixed_to_int(t->matrix[1][2]); |
return true; |
} |
/** |
* Returns the floating-point coordinates transformed by the given transform. |
*/ |
void |
sna_get_transformed_coordinates(int x, int y, |
const PictTransform *transform, |
float *x_out, float *y_out) |
{ |
if (transform == NULL) { |
*x_out = x; |
*y_out = y; |
} else |
_sna_get_transformed_coordinates(x, y, transform, x_out, y_out); |
} |
/** |
* Returns the un-normalized floating-point coordinates transformed by the given transform. |
*/ |
void |
sna_get_transformed_coordinates_3d(int x, int y, |
const PictTransform *transform, |
float *x_out, float *y_out, float *w_out) |
{ |
if (transform == NULL) { |
*x_out = x; |
*y_out = y; |
*w_out = 1; |
} else { |
int64_t result[3]; |
if (_sna_transform_point(transform, x, y, result)) { |
*x_out = result[0] / 65536.; |
*y_out = result[1] / 65536.; |
*w_out = result[2] / 65536.; |
} else { |
*x_out = *y_out = 0; |
*w_out = 1.; |
} |
} |
} |
/drivers/video/Intel-2D/sna/utils.c |
---|
0,0 → 1,150 |
#include <stdint.h> |
#include <stdio.h> |
static inline void native_cpuid(unsigned int *eax, unsigned int *ebx, |
unsigned int *ecx, unsigned int *edx) |
{ |
/* ecx is often an input as well as an output. */ |
asm volatile("cpuid" |
: "=a" (*eax), |
"=b" (*ebx), |
"=c" (*ecx), |
"=d" (*edx) |
: "0" (*eax), "2" (*ecx) |
: "memory"); |
} |
/* Some CPUID calls want 'count' to be placed in ecx */ |
static inline void cpuid_count(unsigned int op, int count, |
unsigned int *eax, unsigned int *ebx, |
unsigned int *ecx, unsigned int *edx) |
{ |
*eax = op; |
*ecx = count; |
native_cpuid(eax, ebx, ecx, edx); |
} |
enum _cache_type { |
CACHE_TYPE_NULL = 0, |
CACHE_TYPE_DATA = 1, |
CACHE_TYPE_INST = 2, |
CACHE_TYPE_UNIFIED = 3 |
}; |
union _cpuid4_leaf_eax { |
struct { |
enum _cache_type type:5; |
unsigned int level:3; |
unsigned int is_self_initializing:1; |
unsigned int is_fully_associative:1; |
unsigned int reserved:4; |
unsigned int num_threads_sharing:12; |
unsigned int num_cores_on_die:6; |
} split; |
uint32_t full; |
}; |
union _cpuid4_leaf_ebx { |
struct { |
unsigned int coherency_line_size:12; |
unsigned int physical_line_partition:10; |
unsigned int ways_of_associativity:10; |
} split; |
uint32_t full; |
}; |
union _cpuid4_leaf_ecx { |
struct { |
unsigned int number_of_sets:32; |
} split; |
uint32_t full; |
}; |
struct _cpuid4_info_regs { |
union _cpuid4_leaf_eax eax; |
union _cpuid4_leaf_ebx ebx; |
union _cpuid4_leaf_ecx ecx; |
unsigned long size; |
}; |
static int |
cpuid4_cache_lookup_regs(int index, |
struct _cpuid4_info_regs *this_leaf) |
{ |
union _cpuid4_leaf_eax eax; |
union _cpuid4_leaf_ebx ebx; |
union _cpuid4_leaf_ecx ecx; |
unsigned edx; |
cpuid_count(4, index, &eax.full, &ebx.full, &ecx.full, &edx); |
if (eax.split.type == CACHE_TYPE_NULL) |
return -1; /* better error ? */ |
this_leaf->eax = eax; |
this_leaf->ebx = ebx; |
this_leaf->ecx = ecx; |
this_leaf->size = (ecx.split.number_of_sets + 1) * |
(ebx.split.coherency_line_size + 1) * |
(ebx.split.physical_line_partition + 1) * |
(ebx.split.ways_of_associativity + 1); |
return 0; |
} |
static int find_num_cache_leaves() |
{ |
unsigned int eax, ebx, ecx, edx, op; |
union _cpuid4_leaf_eax cache_eax; |
int i = -1; |
do { |
++i; |
/* Do cpuid(op) loop to find out num_cache_leaves */ |
cpuid_count(4, i, &eax, &ebx, &ecx, &edx); |
cache_eax.full = eax; |
} while (cache_eax.split.type != CACHE_TYPE_NULL); |
return i; |
}; |
unsigned int cpu_cache_size() |
{ |
unsigned int new_l1d = 0, new_l1i = 0; /* Cache sizes from cpuid(4) */ |
unsigned int new_l2 = 0, new_l3 = 0, i; /* Cache sizes from cpuid(4) */ |
unsigned int num_cache_leaves; |
num_cache_leaves = find_num_cache_leaves(); |
for (i = 0; i < num_cache_leaves; i++) |
{ |
struct _cpuid4_info_regs this_leaf; |
int retval; |
retval = cpuid4_cache_lookup_regs(i, &this_leaf); |
if (retval >= 0) { |
switch (this_leaf.eax.split.level) |
{ |
case 1: |
if (this_leaf.eax.split.type == CACHE_TYPE_DATA) |
new_l1d = this_leaf.size; |
else if (this_leaf.eax.split.type == CACHE_TYPE_INST) |
new_l1i = this_leaf.size; |
break; |
case 2: |
new_l2 = this_leaf.size; |
break; |
case 3: |
new_l3 = this_leaf.size; |
break; |
default: |
break; |
} |
} |
} |
printf("l2 cache %d l3 cache %d\n", new_l2, new_l3); |
return new_l3 != 0 ? new_l3 : new_l2; |
}; |