src/compiler/nir/nir_lower_atomics.c

   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 "compiler/glsl/ir_uniform.h"
  29 #include "nir.h"
  30 #include "main/config.h"
  31 #include <assert.h>
  32
  33 typedef struct {
  34    const struct gl_shader_program *shader_program;
  35    nir_shader   *shader;
  36 } lower_atomic_state;
  37
  38 /*
  39  * replace atomic counter intrinsics that use a variable with intrinsics
  40  * that directly store the buffer index and byte offset
  41  */
  42
  43 static void
  44 lower_instr(nir_intrinsic_instr *instr,
  45             lower_atomic_state *state)
  46 {
  47    nir_intrinsic_op op;
  48    switch (instr->intrinsic) {
  49    case nir_intrinsic_atomic_counter_read_var:
  50       op = nir_intrinsic_atomic_counter_read;
  51       break;
  52
  53    case nir_intrinsic_atomic_counter_inc_var:
  54       op = nir_intrinsic_atomic_counter_inc;
  55       break;
  56
  57    case nir_intrinsic_atomic_counter_dec_var:
  58       op = nir_intrinsic_atomic_counter_dec;
  59       break;
  60
  61    default:
  62       return;
  63    }
  64
  65    if (instr->variables[0]->var->data.mode != nir_var_uniform &&
  66        instr->variables[0]->var->data.mode != nir_var_shader_storage)
  67       return; /* atomics passed as function arguments can't be lowered */
  68
  69    void *mem_ctx = ralloc_parent(instr);
  70    unsigned uniform_loc = instr->variables[0]->var->data.location;
  71
  72    nir_intrinsic_instr *new_instr = nir_intrinsic_instr_create(mem_ctx, op);
  73    new_instr->const_index[0] =
  74       state->shader_program->UniformStorage[uniform_loc].opaque[state->shader->stage].index;
  75
  76    nir_load_const_instr *offset_const = nir_load_const_instr_create(mem_ctx, 1);
  77    offset_const->value.u[0] = instr->variables[0]->var->data.offset;
  78
  79    nir_instr_insert_before(&instr->instr, &offset_const->instr);
  80
  81    nir_ssa_def *offset_def = &offset_const->def;
  82
  83    nir_deref *tail = &instr->variables[0]->deref;
  84    while (tail->child != NULL) {
  85       assert(tail->child->deref_type == nir_deref_type_array);
  86       nir_deref_array *deref_array = nir_deref_as_array(tail->child);
  87       tail = tail->child;
  88
  89       unsigned child_array_elements = tail->child != NULL ?
  90          glsl_get_aoa_size(tail->type) : 1;
  91
  92       offset_const->value.u[0] += deref_array->base_offset *
  93          child_array_elements * ATOMIC_COUNTER_SIZE;
  94
  95       if (deref_array->deref_array_type == nir_deref_array_type_indirect) {
  96          nir_load_const_instr *atomic_counter_size =
  97                nir_load_const_instr_create(mem_ctx, 1);
  98          atomic_counter_size->value.u[0] = child_array_elements * ATOMIC_COUNTER_SIZE;
  99          nir_instr_insert_before(&instr->instr, &atomic_counter_size->instr);
 100
 101          nir_alu_instr *mul = nir_alu_instr_create(mem_ctx, nir_op_imul);
 102          nir_ssa_dest_init(&mul->instr, &mul->dest.dest, 1, NULL);
 103          mul->dest.write_mask = 0x1;
 104          nir_src_copy(&mul->src[0].src, &deref_array->indirect, mul);
 105          mul->src[1].src.is_ssa = true;
 106          mul->src[1].src.ssa = &atomic_counter_size->def;
 107          nir_instr_insert_before(&instr->instr, &mul->instr);
 108
 109          nir_alu_instr *add = nir_alu_instr_create(mem_ctx, nir_op_iadd);
 110          nir_ssa_dest_init(&add->instr, &add->dest.dest, 1, NULL);
 111          add->dest.write_mask = 0x1;
 112          add->src[0].src.is_ssa = true;
 113          add->src[0].src.ssa = &mul->dest.dest.ssa;
 114          add->src[1].src.is_ssa = true;
 115          add->src[1].src.ssa = offset_def;
 116          nir_instr_insert_before(&instr->instr, &add->instr);
 117
 118          offset_def = &add->dest.dest.ssa;
 119       }
 120    }
 121
 122    new_instr->src[0].is_ssa = true;
 123    new_instr->src[0].ssa = offset_def;
 124
 125    if (instr->dest.is_ssa) {
 126       nir_ssa_dest_init(&new_instr->instr, &new_instr->dest,
 127                         instr->dest.ssa.num_components, NULL);
 128       nir_ssa_def_rewrite_uses(&instr->dest.ssa,
 129                                nir_src_for_ssa(&new_instr->dest.ssa));
 130    } else {
 131       nir_dest_copy(&new_instr->dest, &instr->dest, mem_ctx);
 132    }
 133
 134    nir_instr_insert_before(&instr->instr, &new_instr->instr);
 135    nir_instr_remove(&instr->instr);
 136 }
 137
 138 static bool
 139 lower_block(nir_block *block, void *state)
 140 {
 141    nir_foreach_instr_safe(block, instr) {
 142       if (instr->type == nir_instr_type_intrinsic)
 143          lower_instr(nir_instr_as_intrinsic(instr),
 144                      (lower_atomic_state *) state);
 145    }
 146
 147    return true;
 148 }
 149
 150 void
 151 nir_lower_atomics(nir_shader *shader,
 152                   const struct gl_shader_program *shader_program)
 153 {
 154    lower_atomic_state state = {
 155       .shader = shader,
 156       .shader_program = shader_program,
 157    };
 158
 159    nir_foreach_function(shader, function) {
 160       if (function->impl) {
 161          nir_foreach_block(function->impl, lower_block, (void *) &state);
 162          nir_metadata_preserve(function->impl, nir_metadata_block_index |
 163                                                nir_metadata_dominance);
 164       }
 165    }
 166 }