Subversion Repositories Kolibri OS

Rev

Blame | Last modification | View Log | RSS feed

  1. /*
  2.  * Copyright © 2011 Intel Corporation
  3.  *
  4.  * Permission is hereby granted, free of charge, to any person obtaining a
  5.  * copy of this software and associated documentation files (the "Software"),
  6.  * to deal in the Software without restriction, including without limitation
  7.  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
  8.  * and/or sell copies of the Software, and to permit persons to whom the
  9.  * Software is furnished to do so, subject to the following conditions:
  10.  *
  11.  * The above copyright notice and this permission notice (including the next
  12.  * paragraph) shall be included in all copies or substantial portions of the
  13.  * Software.
  14.  *
  15.  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  16.  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  17.  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
  18.  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  19.  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
  20.  * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
  21.  * IN THE SOFTWARE.
  22.  */
  23. #include "main/mtypes.h"
  24. #include "main/blend.h"
  25. #include "main/samplerobj.h"
  26. #include "main/texformat.h"
  27. #include "main/teximage.h"
  28. #include "program/prog_parameter.h"
  29.  
  30. #include "intel_mipmap_tree.h"
  31. #include "intel_batchbuffer.h"
  32. #include "intel_tex.h"
  33. #include "intel_fbo.h"
  34. #include "intel_buffer_objects.h"
  35.  
  36. #include "brw_context.h"
  37. #include "brw_state.h"
  38. #include "brw_defines.h"
  39. #include "brw_wm.h"
  40.  
  41. /**
  42.  * Convert an swizzle enumeration (i.e. SWIZZLE_X) to one of the Gen7.5+
  43.  * "Shader Channel Select" enumerations (i.e. HSW_SCS_RED).  The mappings are
  44.  *
  45.  * SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_W, SWIZZLE_ZERO, SWIZZLE_ONE
  46.  *         0          1          2          3             4            5
  47.  *         4          5          6          7             0            1
  48.  *   SCS_RED, SCS_GREEN,  SCS_BLUE, SCS_ALPHA,     SCS_ZERO,     SCS_ONE
  49.  *
  50.  * which is simply adding 4 then modding by 8 (or anding with 7).
  51.  *
  52.  * We then may need to apply workarounds for textureGather hardware bugs.
  53.  */
  54. static unsigned
  55. swizzle_to_scs(GLenum swizzle, bool need_green_to_blue)
  56. {
  57.    unsigned scs = (swizzle + 4) & 7;
  58.  
  59.    return (need_green_to_blue && scs == HSW_SCS_GREEN) ? HSW_SCS_BLUE : scs;
  60. }
  61.  
  62. uint32_t
  63. gen7_surface_tiling_mode(uint32_t tiling)
  64. {
  65.    switch (tiling) {
  66.    case I915_TILING_X:
  67.       return GEN7_SURFACE_TILING_X;
  68.    case I915_TILING_Y:
  69.       return GEN7_SURFACE_TILING_Y;
  70.    default:
  71.       return GEN7_SURFACE_TILING_NONE;
  72.    }
  73. }
  74.  
  75.  
  76. uint32_t
  77. gen7_surface_msaa_bits(unsigned num_samples, enum intel_msaa_layout layout)
  78. {
  79.    uint32_t ss4 = 0;
  80.  
  81.    assert(num_samples <= 8);
  82.  
  83.    /* The SURFACE_MULTISAMPLECOUNT_X enums are simply log2(num_samples) << 3. */
  84.    ss4 |= (ffs(MAX2(num_samples, 1)) - 1) << 3;
  85.  
  86.    if (layout == INTEL_MSAA_LAYOUT_IMS)
  87.       ss4 |= GEN7_SURFACE_MSFMT_DEPTH_STENCIL;
  88.    else
  89.       ss4 |= GEN7_SURFACE_MSFMT_MSS;
  90.  
  91.    return ss4;
  92. }
  93.  
  94.  
  95. void
  96. gen7_set_surface_mcs_info(struct brw_context *brw,
  97.                           uint32_t *surf,
  98.                           uint32_t surf_offset,
  99.                           const struct intel_mipmap_tree *mcs_mt,
  100.                           bool is_render_target)
  101. {
  102.    /* From the Ivy Bridge PRM, Vol4 Part1 p76, "MCS Base Address":
  103.     *
  104.     *     "The MCS surface must be stored as Tile Y."
  105.     */
  106.    assert(mcs_mt->tiling == I915_TILING_Y);
  107.  
  108.    /* Compute the pitch in units of tiles.  To do this we need to divide the
  109.     * pitch in bytes by 128, since a single Y-tile is 128 bytes wide.
  110.     */
  111.    unsigned pitch_tiles = mcs_mt->pitch / 128;
  112.  
  113.    /* The upper 20 bits of surface state DWORD 6 are the upper 20 bits of the
  114.     * GPU address of the MCS buffer; the lower 12 bits contain other control
  115.     * information.  Since buffer addresses are always on 4k boundaries (and
  116.     * thus have their lower 12 bits zero), we can use an ordinary reloc to do
  117.     * the necessary address translation.
  118.     */
  119.    assert ((mcs_mt->bo->offset64 & 0xfff) == 0);
  120.  
  121.    surf[6] = GEN7_SURFACE_MCS_ENABLE |
  122.              SET_FIELD(pitch_tiles - 1, GEN7_SURFACE_MCS_PITCH) |
  123.              mcs_mt->bo->offset64;
  124.  
  125.    drm_intel_bo_emit_reloc(brw->batch.bo,
  126.                            surf_offset + 6 * 4,
  127.                            mcs_mt->bo,
  128.                            surf[6] & 0xfff,
  129.                            is_render_target ? I915_GEM_DOMAIN_RENDER
  130.                            : I915_GEM_DOMAIN_SAMPLER,
  131.                            is_render_target ? I915_GEM_DOMAIN_RENDER : 0);
  132. }
  133.  
  134.  
  135. void
  136. gen7_check_surface_setup(uint32_t *surf, bool is_render_target)
  137. {
  138.    unsigned num_multisamples = surf[4] & INTEL_MASK(5, 3);
  139.    unsigned multisampled_surface_storage_format = surf[4] & (1 << 6);
  140.    unsigned surface_array_spacing = surf[0] & (1 << 10);
  141.    bool is_multisampled = num_multisamples != GEN7_SURFACE_MULTISAMPLECOUNT_1;
  142.  
  143.    (void) surface_array_spacing;
  144.  
  145.    /* From the Ivybridge PRM, Volume 4 Part 1, page 66 (RENDER_SURFACE_STATE
  146.     * dword 0 bit 10 "Surface Array Spacing" Programming Notes):
  147.     *
  148.     *   If Multisampled Surface Storage Format is MSFMT_MSS and Number of
  149.     *   Multisamples is not MULTISAMPLECOUNT_1, this field must be set to
  150.     *   ARYSPC_LOD0.
  151.     */
  152.    if (multisampled_surface_storage_format == GEN7_SURFACE_MSFMT_MSS
  153.        && is_multisampled)
  154.       assert(surface_array_spacing == GEN7_SURFACE_ARYSPC_LOD0);
  155.  
  156.    /* From the Ivybridge PRM, Volume 4 Part 1, page 72 (RENDER_SURFACE_STATE
  157.     * dword 4 bit 6 "Multisampled Surface Storage" Programming Notes):
  158.     *
  159.     *   All multisampled render target surfaces must have this field set to
  160.     *   MSFMT_MSS.
  161.     *
  162.     * But also:
  163.     *
  164.     *   This field is ignored if Number of Multisamples is MULTISAMPLECOUNT_1.
  165.     */
  166.    if (is_render_target && is_multisampled) {
  167.       assert(multisampled_surface_storage_format == GEN7_SURFACE_MSFMT_MSS);
  168.    }
  169.  
  170.    /* From the Ivybridge PRM, Volume 4 Part 1, page 72 (RENDER_SURFACE_STATE
  171.     * dword 4 bit 6 "Multisampled Surface Storage Format" Errata):
  172.     *
  173.     *   If the surface’s Number of Multisamples is MULTISAMPLECOUNT_8, Width
  174.     *   is >= 8192 (meaning the actual surface width is >= 8193 pixels), this
  175.     *   field must be set to MSFMT_MSS.
  176.     */
  177.    uint32_t width = GET_FIELD(surf[2], GEN7_SURFACE_WIDTH) + 1;
  178.    if (num_multisamples == GEN7_SURFACE_MULTISAMPLECOUNT_8 && width >= 8193) {
  179.       assert(multisampled_surface_storage_format == GEN7_SURFACE_MSFMT_MSS);
  180.    }
  181.  
  182.    /* From the Ivybridge PRM, Volume 4 Part 1, page 72 (RENDER_SURFACE_STATE
  183.     * dword 4 bit 6 "Multisampled Surface Storage Format" Errata):
  184.     *
  185.     *   If the surface’s Number of Multisamples is MULTISAMPLECOUNT_8,
  186.     *   ((Depth+1) * (Height+1)) is > 4,194,304, OR if the surface’s Number of
  187.     *   Multisamples is MULTISAMPLECOUNT_4, ((Depth+1) * (Height+1)) is >
  188.     *   8,388,608, this field must be set to MSFMT_DEPTH_STENCIL.This field
  189.     *   must be set to MSFMT_DEPTH_STENCIL if Surface Format is one of the
  190.     *   following: I24X8_UNORM, L24X8_UNORM, A24X8_UNORM, or
  191.     *   R24_UNORM_X8_TYPELESS.
  192.     *
  193.     * But also (from the Programming Notes):
  194.     *
  195.     *   This field is ignored if Number of Multisamples is MULTISAMPLECOUNT_1.
  196.     */
  197.    uint32_t depth = GET_FIELD(surf[3], BRW_SURFACE_DEPTH) + 1;
  198.    uint32_t height = GET_FIELD(surf[2], GEN7_SURFACE_HEIGHT) + 1;
  199.    if (num_multisamples == GEN7_SURFACE_MULTISAMPLECOUNT_8 &&
  200.        depth * height > 4194304) {
  201.       assert(multisampled_surface_storage_format ==
  202.              GEN7_SURFACE_MSFMT_DEPTH_STENCIL);
  203.    }
  204.    if (num_multisamples == GEN7_SURFACE_MULTISAMPLECOUNT_4 &&
  205.        depth * height > 8388608) {
  206.       assert(multisampled_surface_storage_format ==
  207.              GEN7_SURFACE_MSFMT_DEPTH_STENCIL);
  208.    }
  209.    if (is_multisampled) {
  210.       switch (GET_FIELD(surf[0], BRW_SURFACE_FORMAT)) {
  211.       case BRW_SURFACEFORMAT_I24X8_UNORM:
  212.       case BRW_SURFACEFORMAT_L24X8_UNORM:
  213.       case BRW_SURFACEFORMAT_A24X8_UNORM:
  214.       case BRW_SURFACEFORMAT_R24_UNORM_X8_TYPELESS:
  215.          assert(multisampled_surface_storage_format ==
  216.                 GEN7_SURFACE_MSFMT_DEPTH_STENCIL);
  217.       }
  218.    }
  219. }
  220.  
  221. static void
  222. gen7_emit_buffer_surface_state(struct brw_context *brw,
  223.                                uint32_t *out_offset,
  224.                                drm_intel_bo *bo,
  225.                                unsigned buffer_offset,
  226.                                unsigned surface_format,
  227.                                unsigned buffer_size,
  228.                                unsigned pitch,
  229.                                bool rw)
  230. {
  231.    uint32_t *surf = brw_state_batch(brw, AUB_TRACE_SURFACE_STATE,
  232.                                     8 * 4, 32, out_offset);
  233.    memset(surf, 0, 8 * 4);
  234.  
  235.    surf[0] = BRW_SURFACE_BUFFER << BRW_SURFACE_TYPE_SHIFT |
  236.              surface_format << BRW_SURFACE_FORMAT_SHIFT |
  237.              BRW_SURFACE_RC_READ_WRITE;
  238.    surf[1] = (bo ? bo->offset64 : 0) + buffer_offset; /* reloc */
  239.    surf[2] = SET_FIELD((buffer_size - 1) & 0x7f, GEN7_SURFACE_WIDTH) |
  240.              SET_FIELD(((buffer_size - 1) >> 7) & 0x3fff, GEN7_SURFACE_HEIGHT);
  241.    if (surface_format == BRW_SURFACEFORMAT_RAW)
  242.       surf[3] = SET_FIELD(((buffer_size - 1) >> 21) & 0x3ff, BRW_SURFACE_DEPTH);
  243.    else
  244.       surf[3] = SET_FIELD(((buffer_size - 1) >> 21) & 0x3f, BRW_SURFACE_DEPTH);
  245.    surf[3] |= (pitch - 1);
  246.  
  247.    surf[5] = SET_FIELD(GEN7_MOCS_L3, GEN7_SURFACE_MOCS);
  248.  
  249.    if (brw->is_haswell) {
  250.       surf[7] |= (SET_FIELD(HSW_SCS_RED,   GEN7_SURFACE_SCS_R) |
  251.                   SET_FIELD(HSW_SCS_GREEN, GEN7_SURFACE_SCS_G) |
  252.                   SET_FIELD(HSW_SCS_BLUE,  GEN7_SURFACE_SCS_B) |
  253.                   SET_FIELD(HSW_SCS_ALPHA, GEN7_SURFACE_SCS_A));
  254.    }
  255.  
  256.    /* Emit relocation to surface contents */
  257.    if (bo) {
  258.       drm_intel_bo_emit_reloc(brw->batch.bo, *out_offset + 4,
  259.                               bo, buffer_offset, I915_GEM_DOMAIN_SAMPLER,
  260.                               (rw ? I915_GEM_DOMAIN_SAMPLER : 0));
  261.    }
  262.  
  263.    gen7_check_surface_setup(surf, false /* is_render_target */);
  264. }
  265.  
  266. static void
  267. gen7_emit_texture_surface_state(struct brw_context *brw,
  268.                                 struct intel_mipmap_tree *mt,
  269.                                 GLenum target,
  270.                                 unsigned min_layer, unsigned max_layer,
  271.                                 unsigned min_level, unsigned max_level,
  272.                                 unsigned format,
  273.                                 unsigned swizzle,
  274.                                 uint32_t *surf_offset,
  275.                                 bool rw, bool for_gather)
  276. {
  277.    const unsigned depth = max_layer - min_layer;
  278.    uint32_t *surf = brw_state_batch(brw, AUB_TRACE_SURFACE_STATE,
  279.                                     8 * 4, 32, surf_offset);
  280.  
  281.    memset(surf, 0, 8 * 4);
  282.  
  283.    surf[0] = translate_tex_target(target) << BRW_SURFACE_TYPE_SHIFT |
  284.              format << BRW_SURFACE_FORMAT_SHIFT |
  285.              gen7_surface_tiling_mode(mt->tiling);
  286.  
  287.    /* mask of faces present in cube map; for other surfaces MBZ. */
  288.    if (target == GL_TEXTURE_CUBE_MAP || target == GL_TEXTURE_CUBE_MAP_ARRAY)
  289.       surf[0] |= BRW_SURFACE_CUBEFACE_ENABLES;
  290.  
  291.    if (mt->align_h == 4)
  292.       surf[0] |= GEN7_SURFACE_VALIGN_4;
  293.    if (mt->align_w == 8)
  294.       surf[0] |= GEN7_SURFACE_HALIGN_8;
  295.  
  296.    if (_mesa_is_array_texture(target) || target == GL_TEXTURE_CUBE_MAP)
  297.       surf[0] |= GEN7_SURFACE_IS_ARRAY;
  298.  
  299.    if (mt->array_layout == ALL_SLICES_AT_EACH_LOD)
  300.       surf[0] |= GEN7_SURFACE_ARYSPC_LOD0;
  301.  
  302.    surf[1] = mt->bo->offset64 + mt->offset; /* reloc */
  303.  
  304.    surf[2] = SET_FIELD(mt->logical_width0 - 1, GEN7_SURFACE_WIDTH) |
  305.              SET_FIELD(mt->logical_height0 - 1, GEN7_SURFACE_HEIGHT);
  306.  
  307.    surf[3] = SET_FIELD(depth - 1, BRW_SURFACE_DEPTH) |
  308.              (mt->pitch - 1);
  309.  
  310.    if (brw->is_haswell && _mesa_is_format_integer(mt->format))
  311.       surf[3] |= HSW_SURFACE_IS_INTEGER_FORMAT;
  312.  
  313.    surf[4] = gen7_surface_msaa_bits(mt->num_samples, mt->msaa_layout) |
  314.              SET_FIELD(min_layer, GEN7_SURFACE_MIN_ARRAY_ELEMENT) |
  315.              SET_FIELD(depth - 1, GEN7_SURFACE_RENDER_TARGET_VIEW_EXTENT);
  316.  
  317.    surf[5] = (SET_FIELD(GEN7_MOCS_L3, GEN7_SURFACE_MOCS) |
  318.               SET_FIELD(min_level - mt->first_level, GEN7_SURFACE_MIN_LOD) |
  319.               /* mip count */
  320.               (max_level - min_level - 1));
  321.  
  322.    surf[7] = mt->fast_clear_color_value;
  323.  
  324.    if (brw->is_haswell) {
  325.       const bool need_scs_green_to_blue = for_gather && format == BRW_SURFACEFORMAT_R32G32_FLOAT_LD;
  326.  
  327.       surf[7] |=
  328.          SET_FIELD(swizzle_to_scs(GET_SWZ(swizzle, 0), need_scs_green_to_blue), GEN7_SURFACE_SCS_R) |
  329.          SET_FIELD(swizzle_to_scs(GET_SWZ(swizzle, 1), need_scs_green_to_blue), GEN7_SURFACE_SCS_G) |
  330.          SET_FIELD(swizzle_to_scs(GET_SWZ(swizzle, 2), need_scs_green_to_blue), GEN7_SURFACE_SCS_B) |
  331.          SET_FIELD(swizzle_to_scs(GET_SWZ(swizzle, 3), need_scs_green_to_blue), GEN7_SURFACE_SCS_A);
  332.    }
  333.  
  334.    if (mt->mcs_mt) {
  335.       gen7_set_surface_mcs_info(brw, surf, *surf_offset,
  336.                                 mt->mcs_mt, false /* is RT */);
  337.    }
  338.  
  339.    /* Emit relocation to surface contents */
  340.    drm_intel_bo_emit_reloc(brw->batch.bo,
  341.                            *surf_offset + 4,
  342.                            mt->bo,
  343.                            surf[1] - mt->bo->offset64,
  344.                            I915_GEM_DOMAIN_SAMPLER,
  345.                            (rw ? I915_GEM_DOMAIN_SAMPLER : 0));
  346.  
  347.    gen7_check_surface_setup(surf, false /* is_render_target */);
  348. }
  349.  
  350. static void
  351. gen7_update_texture_surface(struct gl_context *ctx,
  352.                             unsigned unit,
  353.                             uint32_t *surf_offset,
  354.                             bool for_gather)
  355. {
  356.    struct brw_context *brw = brw_context(ctx);
  357.    struct gl_texture_object *obj = ctx->Texture.Unit[unit]._Current;
  358.  
  359.    if (obj->Target == GL_TEXTURE_BUFFER) {
  360.       brw_update_buffer_texture_surface(ctx, unit, surf_offset);
  361.  
  362.    } else {
  363.       struct intel_texture_object *intel_obj = intel_texture_object(obj);
  364.       struct intel_mipmap_tree *mt = intel_obj->mt;
  365.       struct gl_sampler_object *sampler = _mesa_get_samplerobj(ctx, unit);
  366.       /* If this is a view with restricted NumLayers, then our effective depth
  367.        * is not just the miptree depth.
  368.        */
  369.       const unsigned depth = (obj->Immutable && obj->Target != GL_TEXTURE_3D ?
  370.                               obj->NumLayers : mt->logical_depth0);
  371.  
  372.       /* Handling GL_ALPHA as a surface format override breaks 1.30+ style
  373.        * texturing functions that return a float, as our code generation always
  374.        * selects the .x channel (which would always be 0).
  375.        */
  376.       struct gl_texture_image *firstImage = obj->Image[0][obj->BaseLevel];
  377.       const bool alpha_depth = obj->DepthMode == GL_ALPHA &&
  378.          (firstImage->_BaseFormat == GL_DEPTH_COMPONENT ||
  379.           firstImage->_BaseFormat == GL_DEPTH_STENCIL);
  380.       const unsigned swizzle = (unlikely(alpha_depth) ? SWIZZLE_XYZW :
  381.                                 brw_get_texture_swizzle(&brw->ctx, obj));
  382.  
  383.       unsigned format = translate_tex_format(
  384.          brw, intel_obj->_Format, sampler->sRGBDecode);
  385.  
  386.       if (for_gather && format == BRW_SURFACEFORMAT_R32G32_FLOAT)
  387.          format = BRW_SURFACEFORMAT_R32G32_FLOAT_LD;
  388.  
  389.       gen7_emit_texture_surface_state(brw, mt, obj->Target,
  390.                                       obj->MinLayer, obj->MinLayer + depth,
  391.                                       obj->MinLevel + obj->BaseLevel,
  392.                                       obj->MinLevel + intel_obj->_MaxLevel + 1,
  393.                                       format, swizzle,
  394.                                       surf_offset, false, for_gather);
  395.    }
  396. }
  397.  
  398. /**
  399.  * Creates a null surface.
  400.  *
  401.  * This is used when the shader doesn't write to any color output.  An FB
  402.  * write to target 0 will still be emitted, because that's how the thread is
  403.  * terminated (and computed depth is returned), so we need to have the
  404.  * hardware discard the target 0 color output..
  405.  */
  406. static void
  407. gen7_emit_null_surface_state(struct brw_context *brw,
  408.                              unsigned width,
  409.                              unsigned height,
  410.                              unsigned samples,
  411.                              uint32_t *out_offset)
  412. {
  413.    /* From the Ivy bridge PRM, Vol4 Part1 p62 (Surface Type: Programming
  414.     * Notes):
  415.     *
  416.     *     A null surface is used in instances where an actual surface is not
  417.     *     bound. When a write message is generated to a null surface, no
  418.     *     actual surface is written to. When a read message (including any
  419.     *     sampling engine message) is generated to a null surface, the result
  420.     *     is all zeros. Note that a null surface type is allowed to be used
  421.     *     with all messages, even if it is not specificially indicated as
  422.     *     supported. All of the remaining fields in surface state are ignored
  423.     *     for null surfaces, with the following exceptions: Width, Height,
  424.     *     Depth, LOD, and Render Target View Extent fields must match the
  425.     *     depth buffer’s corresponding state for all render target surfaces,
  426.     *     including null.
  427.     */
  428.    uint32_t *surf = brw_state_batch(brw, AUB_TRACE_SURFACE_STATE, 8 * 4, 32,
  429.                                     out_offset);
  430.    memset(surf, 0, 8 * 4);
  431.  
  432.    /* From the Ivybridge PRM, Volume 4, Part 1, page 65,
  433.     * Tiled Surface: Programming Notes:
  434.     * "If Surface Type is SURFTYPE_NULL, this field must be TRUE."
  435.     */
  436.    surf[0] = BRW_SURFACE_NULL << BRW_SURFACE_TYPE_SHIFT |
  437.              BRW_SURFACEFORMAT_B8G8R8A8_UNORM << BRW_SURFACE_FORMAT_SHIFT |
  438.              GEN7_SURFACE_TILING_Y;
  439.  
  440.    surf[2] = SET_FIELD(width - 1, GEN7_SURFACE_WIDTH) |
  441.              SET_FIELD(height - 1, GEN7_SURFACE_HEIGHT);
  442.  
  443.    gen7_check_surface_setup(surf, true /* is_render_target */);
  444. }
  445.  
  446. /**
  447.  * Sets up a surface state structure to point at the given region.
  448.  * While it is only used for the front/back buffer currently, it should be
  449.  * usable for further buffers when doing ARB_draw_buffer support.
  450.  */
  451. static uint32_t
  452. gen7_update_renderbuffer_surface(struct brw_context *brw,
  453.                                  struct gl_renderbuffer *rb,
  454.                                  bool layered, unsigned unit /* unused */,
  455.                                  uint32_t surf_index)
  456. {
  457.    struct gl_context *ctx = &brw->ctx;
  458.    struct intel_renderbuffer *irb = intel_renderbuffer(rb);
  459.    struct intel_mipmap_tree *mt = irb->mt;
  460.    uint32_t format;
  461.    /* _NEW_BUFFERS */
  462.    mesa_format rb_format = _mesa_get_render_format(ctx, intel_rb_format(irb));
  463.    uint32_t surftype;
  464.    bool is_array = false;
  465.    int depth = MAX2(irb->layer_count, 1);
  466.    const uint8_t mocs = GEN7_MOCS_L3;
  467.    uint32_t offset;
  468.  
  469.    int min_array_element = irb->mt_layer / MAX2(mt->num_samples, 1);
  470.  
  471.    GLenum gl_target = rb->TexImage ?
  472.                          rb->TexImage->TexObject->Target : GL_TEXTURE_2D;
  473.  
  474.    uint32_t *surf = brw_state_batch(brw, AUB_TRACE_SURFACE_STATE, 8 * 4, 32,
  475.                                     &offset);
  476.    memset(surf, 0, 8 * 4);
  477.  
  478.    intel_miptree_used_for_rendering(irb->mt);
  479.  
  480.    /* Render targets can't use IMS layout */
  481.    assert(irb->mt->msaa_layout != INTEL_MSAA_LAYOUT_IMS);
  482.  
  483.    assert(brw_render_target_supported(brw, rb));
  484.    format = brw->render_target_format[rb_format];
  485.    if (unlikely(!brw->format_supported_as_render_target[rb_format])) {
  486.       _mesa_problem(ctx, "%s: renderbuffer format %s unsupported\n",
  487.                     __func__, _mesa_get_format_name(rb_format));
  488.    }
  489.  
  490.    switch (gl_target) {
  491.    case GL_TEXTURE_CUBE_MAP_ARRAY:
  492.    case GL_TEXTURE_CUBE_MAP:
  493.       surftype = BRW_SURFACE_2D;
  494.       is_array = true;
  495.       depth *= 6;
  496.       break;
  497.    case GL_TEXTURE_3D:
  498.       depth = MAX2(irb->mt->logical_depth0, 1);
  499.       /* fallthrough */
  500.    default:
  501.       surftype = translate_tex_target(gl_target);
  502.       is_array = _mesa_tex_target_is_array(gl_target);
  503.       break;
  504.    }
  505.  
  506.    surf[0] = surftype << BRW_SURFACE_TYPE_SHIFT |
  507.              format << BRW_SURFACE_FORMAT_SHIFT |
  508.              (irb->mt->array_layout == ALL_SLICES_AT_EACH_LOD ?
  509.                  GEN7_SURFACE_ARYSPC_LOD0 : GEN7_SURFACE_ARYSPC_FULL) |
  510.              gen7_surface_tiling_mode(mt->tiling);
  511.  
  512.    if (irb->mt->align_h == 4)
  513.       surf[0] |= GEN7_SURFACE_VALIGN_4;
  514.    if (irb->mt->align_w == 8)
  515.       surf[0] |= GEN7_SURFACE_HALIGN_8;
  516.  
  517.    if (is_array) {
  518.       surf[0] |= GEN7_SURFACE_IS_ARRAY;
  519.    }
  520.  
  521.    assert(mt->offset % mt->cpp == 0);
  522.    surf[1] = mt->bo->offset64 + mt->offset;
  523.  
  524.    assert(brw->has_surface_tile_offset);
  525.  
  526.    surf[5] = SET_FIELD(mocs, GEN7_SURFACE_MOCS) |
  527.              (irb->mt_level - irb->mt->first_level);
  528.  
  529.    surf[2] = SET_FIELD(irb->mt->logical_width0 - 1, GEN7_SURFACE_WIDTH) |
  530.              SET_FIELD(irb->mt->logical_height0 - 1, GEN7_SURFACE_HEIGHT);
  531.  
  532.    surf[3] = ((depth - 1) << BRW_SURFACE_DEPTH_SHIFT) |
  533.              (mt->pitch - 1);
  534.  
  535.    surf[4] = gen7_surface_msaa_bits(irb->mt->num_samples, irb->mt->msaa_layout) |
  536.              min_array_element << GEN7_SURFACE_MIN_ARRAY_ELEMENT_SHIFT |
  537.              (depth - 1) << GEN7_SURFACE_RENDER_TARGET_VIEW_EXTENT_SHIFT;
  538.  
  539.    if (irb->mt->mcs_mt) {
  540.       gen7_set_surface_mcs_info(brw, surf, offset,
  541.                                 irb->mt->mcs_mt, true /* is RT */);
  542.    }
  543.  
  544.    surf[7] = irb->mt->fast_clear_color_value;
  545.  
  546.    if (brw->is_haswell) {
  547.       surf[7] |= (SET_FIELD(HSW_SCS_RED,   GEN7_SURFACE_SCS_R) |
  548.                   SET_FIELD(HSW_SCS_GREEN, GEN7_SURFACE_SCS_G) |
  549.                   SET_FIELD(HSW_SCS_BLUE,  GEN7_SURFACE_SCS_B) |
  550.                   SET_FIELD(HSW_SCS_ALPHA, GEN7_SURFACE_SCS_A));
  551.    }
  552.  
  553.    drm_intel_bo_emit_reloc(brw->batch.bo,
  554.                            offset + 4,
  555.                            mt->bo,
  556.                            surf[1] - mt->bo->offset64,
  557.                            I915_GEM_DOMAIN_RENDER,
  558.                            I915_GEM_DOMAIN_RENDER);
  559.  
  560.    gen7_check_surface_setup(surf, true /* is_render_target */);
  561.  
  562.    return offset;
  563. }
  564.  
  565. void
  566. gen7_init_vtable_surface_functions(struct brw_context *brw)
  567. {
  568.    brw->vtbl.update_texture_surface = gen7_update_texture_surface;
  569.    brw->vtbl.update_renderbuffer_surface = gen7_update_renderbuffer_surface;
  570.    brw->vtbl.emit_null_surface_state = gen7_emit_null_surface_state;
  571.    brw->vtbl.emit_texture_surface_state = gen7_emit_texture_surface_state;
  572.    brw->vtbl.emit_buffer_surface_state = gen7_emit_buffer_surface_state;
  573. }
  574.