Details | Last modification | View Log | RSS feed
Rev | Author | Line No. | Line |
---|---|---|---|
5564 | serge | 1 | /* |
2 | * Copyright © 2014 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 | * Authors: |
||
24 | * Connor Abbott (cwabbott0@gmail.com) |
||
25 | * |
||
26 | */ |
||
27 | |||
28 | #include "nir.h" |
||
29 | #include "main/config.h" |
||
30 | #include |
||
31 | |||
32 | /* |
||
33 | * replace atomic counter intrinsics that use a variable with intrinsics |
||
34 | * that directly store the buffer index and byte offset |
||
35 | */ |
||
36 | |||
37 | static void |
||
38 | lower_instr(nir_intrinsic_instr *instr, nir_function_impl *impl) |
||
39 | { |
||
40 | nir_intrinsic_op op; |
||
41 | switch (instr->intrinsic) { |
||
42 | case nir_intrinsic_atomic_counter_read_var: |
||
43 | op = nir_intrinsic_atomic_counter_read; |
||
44 | break; |
||
45 | |||
46 | case nir_intrinsic_atomic_counter_inc_var: |
||
47 | op = nir_intrinsic_atomic_counter_inc; |
||
48 | break; |
||
49 | |||
50 | case nir_intrinsic_atomic_counter_dec_var: |
||
51 | op = nir_intrinsic_atomic_counter_dec; |
||
52 | break; |
||
53 | |||
54 | default: |
||
55 | return; |
||
56 | } |
||
57 | |||
58 | if (instr->variables[0]->var->data.mode != nir_var_uniform) |
||
59 | return; /* atomics passed as function arguments can't be lowered */ |
||
60 | |||
61 | void *mem_ctx = ralloc_parent(instr); |
||
62 | |||
63 | nir_intrinsic_instr *new_instr = nir_intrinsic_instr_create(mem_ctx, op); |
||
64 | new_instr->const_index[0] = |
||
65 | (int) instr->variables[0]->var->data.atomic.buffer_index; |
||
66 | |||
67 | nir_load_const_instr *offset_const = nir_load_const_instr_create(mem_ctx, 1); |
||
68 | offset_const->value.u[0] = instr->variables[0]->var->data.atomic.offset; |
||
69 | |||
70 | nir_instr_insert_before(&instr->instr, &offset_const->instr); |
||
71 | |||
72 | nir_ssa_def *offset_def = &offset_const->def; |
||
73 | |||
74 | if (instr->variables[0]->deref.child != NULL) { |
||
75 | assert(instr->variables[0]->deref.child->deref_type == |
||
76 | nir_deref_type_array); |
||
77 | nir_deref_array *deref_array = |
||
78 | nir_deref_as_array(instr->variables[0]->deref.child); |
||
79 | assert(deref_array->deref.child == NULL); |
||
80 | |||
81 | offset_const->value.u[0] += |
||
82 | deref_array->base_offset * ATOMIC_COUNTER_SIZE; |
||
83 | |||
84 | if (deref_array->deref_array_type == nir_deref_array_type_indirect) { |
||
85 | nir_load_const_instr *atomic_counter_size = |
||
86 | nir_load_const_instr_create(mem_ctx, 1); |
||
87 | atomic_counter_size->value.u[0] = ATOMIC_COUNTER_SIZE; |
||
88 | nir_instr_insert_before(&instr->instr, &atomic_counter_size->instr); |
||
89 | |||
90 | nir_alu_instr *mul = nir_alu_instr_create(mem_ctx, nir_op_imul); |
||
91 | nir_ssa_dest_init(&mul->instr, &mul->dest.dest, 1, NULL); |
||
92 | mul->dest.write_mask = 0x1; |
||
93 | nir_src_copy(&mul->src[0].src, &deref_array->indirect, mem_ctx); |
||
94 | mul->src[1].src.is_ssa = true; |
||
95 | mul->src[1].src.ssa = &atomic_counter_size->def; |
||
96 | nir_instr_insert_before(&instr->instr, &mul->instr); |
||
97 | |||
98 | nir_alu_instr *add = nir_alu_instr_create(mem_ctx, nir_op_iadd); |
||
99 | nir_ssa_dest_init(&add->instr, &add->dest.dest, 1, NULL); |
||
100 | add->dest.write_mask = 0x1; |
||
101 | add->src[0].src.is_ssa = true; |
||
102 | add->src[0].src.ssa = &mul->dest.dest.ssa; |
||
103 | add->src[1].src.is_ssa = true; |
||
104 | add->src[1].src.ssa = &offset_const->def; |
||
105 | nir_instr_insert_before(&instr->instr, &add->instr); |
||
106 | |||
107 | offset_def = &add->dest.dest.ssa; |
||
108 | } |
||
109 | } |
||
110 | |||
111 | new_instr->src[0].is_ssa = true; |
||
112 | new_instr->src[0].ssa = offset_def;; |
||
113 | |||
114 | if (instr->dest.is_ssa) { |
||
115 | nir_ssa_dest_init(&new_instr->instr, &new_instr->dest, |
||
116 | instr->dest.ssa.num_components, NULL); |
||
117 | nir_ssa_def_rewrite_uses(&instr->dest.ssa, |
||
118 | nir_src_for_ssa(&new_instr->dest.ssa), |
||
119 | mem_ctx); |
||
120 | } else { |
||
121 | nir_dest_copy(&new_instr->dest, &instr->dest, mem_ctx); |
||
122 | } |
||
123 | |||
124 | nir_instr_insert_before(&instr->instr, &new_instr->instr); |
||
125 | nir_instr_remove(&instr->instr); |
||
126 | } |
||
127 | |||
128 | static bool |
||
129 | lower_block(nir_block *block, void *state) |
||
130 | { |
||
131 | nir_foreach_instr_safe(block, instr) { |
||
132 | if (instr->type == nir_instr_type_intrinsic) |
||
133 | lower_instr(nir_instr_as_intrinsic(instr), state); |
||
134 | } |
||
135 | |||
136 | return true; |
||
137 | } |
||
138 | |||
139 | void |
||
140 | nir_lower_atomics(nir_shader *shader) |
||
141 | { |
||
142 | nir_foreach_overload(shader, overload) { |
||
143 | if (overload->impl) { |
||
144 | nir_foreach_block(overload->impl, lower_block, overload->impl); |
||
145 | nir_metadata_preserve(overload->impl, nir_metadata_block_index | |
||
146 | nir_metadata_dominance); |
||
147 | } |
||
148 | } |
||
149 | } |