src/compiler/nir/nir_opt_undef.c

   1 /*
   2  * Copyright © 2015 Broadcom
   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
  24 #include "nir.h"
  25 #include "nir_builder.h"
  26
  27 /** @file nir_opt_undef.c
  28  *
  29  * Handles optimization of operations involving ssa_undef.
  30  */
  31
  32 /**
  33  * Turn conditional selects between an undef and some other value into a move
  34  * of that other value (on the assumption that the condition's going to be
  35  * choosing the defined value).  This reduces work after if flattening when
  36  * each side of the if is defining a variable.
  37  */
  38 static bool
  39 opt_undef_csel(nir_alu_instr *instr)
  40 {
  41    if (instr->op != nir_op_bcsel && instr->op != nir_op_fcsel)
  42       return false;
  43
  44    assert(instr->dest.dest.is_ssa);
  45
  46    for (int i = 1; i <= 2; i++) {
  47       if (!instr->src[i].src.is_ssa)
  48          continue;
  49
  50       nir_instr *parent = instr->src[i].src.ssa->parent_instr;
  51       if (parent->type != nir_instr_type_ssa_undef)
  52          continue;
  53
  54       /* We can't just use nir_alu_src_copy, because we need the def/use
  55        * updated.
  56        */
  57       nir_instr_rewrite_src(&instr->instr, &instr->src[0].src,
  58                             instr->src[i == 1 ? 2 : 1].src);
  59       nir_alu_src_copy(&instr->src[0], &instr->src[i == 1 ? 2 : 1],
  60                        ralloc_parent(instr));
  61
  62       nir_src empty_src;
  63       memset(&empty_src, 0, sizeof(empty_src));
  64       nir_instr_rewrite_src(&instr->instr, &instr->src[1].src, empty_src);
  65       nir_instr_rewrite_src(&instr->instr, &instr->src[2].src, empty_src);
  66       instr->op = nir_op_mov;
  67
  68       return true;
  69    }
  70
  71    return false;
  72 }
  73
  74 /**
  75  * Replace vecN(undef, undef, ...) with a single undef.
  76  */
  77 static bool
  78 opt_undef_vecN(nir_builder *b, nir_alu_instr *alu)
  79 {
  80    if (alu->op != nir_op_vec2 &&
  81        alu->op != nir_op_vec3 &&
  82        alu->op != nir_op_vec4 &&
  83        alu->op != nir_op_mov)
  84       return false;
  85
  86    assert(alu->dest.dest.is_ssa);
  87
  88    for (unsigned i = 0; i < nir_op_infos[alu->op].num_inputs; i++) {
  89       if (!alu->src[i].src.is_ssa ||
  90           alu->src[i].src.ssa->parent_instr->type != nir_instr_type_ssa_undef)
  91          return false;
  92    }
  93
  94    b->cursor = nir_before_instr(&alu->instr);
  95    nir_ssa_def *undef = nir_ssa_undef(b, alu->dest.dest.ssa.num_components,
  96                                       nir_dest_bit_size(alu->dest.dest));
  97    nir_ssa_def_rewrite_uses(&alu->dest.dest.ssa, nir_src_for_ssa(undef));
  98
  99    return true;
 100 }
 101
 102 /**
 103  * Remove any store intrinsics whose value is undefined (the existing
 104  * value is a fine representation of "undefined").
 105  */
 106 static bool
 107 opt_undef_store(nir_intrinsic_instr *intrin)
 108 {
 109    int arg_index;
 110    switch (intrin->intrinsic) {
 111    case nir_intrinsic_store_deref:
 112       arg_index = 1;
 113       break;
 114    case nir_intrinsic_store_output:
 115    case nir_intrinsic_store_per_vertex_output:
 116    case nir_intrinsic_store_ssbo:
 117    case nir_intrinsic_store_shared:
 118       arg_index =  0;
 119       break;
 120    default:
 121       return false;
 122    }
 123
 124    if (!intrin->src[arg_index].is_ssa ||
 125        intrin->src[arg_index].ssa->parent_instr->type != nir_instr_type_ssa_undef)
 126       return false;
 127
 128    nir_instr_remove(&intrin->instr);
 129
 130    return true;
 131 }
 132
 133 bool
 134 nir_opt_undef(nir_shader *shader)
 135 {
 136    nir_builder b;
 137    bool progress = false;
 138
 139    nir_foreach_function(function, shader) {
 140       if (function->impl) {
 141          nir_builder_init(&b, function->impl);
 142          nir_foreach_block(block, function->impl) {
 143             nir_foreach_instr_safe(instr, block) {
 144                if (instr->type == nir_instr_type_alu) {
 145                   nir_alu_instr *alu = nir_instr_as_alu(instr);
 146
 147                   progress = opt_undef_csel(alu) || progress;
 148                   progress = opt_undef_vecN(&b, alu) || progress;
 149                } else if (instr->type == nir_instr_type_intrinsic) {
 150                   nir_intrinsic_instr *intrin = nir_instr_as_intrinsic(instr);
 151                   progress = opt_undef_store(intrin) || progress;
 152                }
 153             }
 154          }
 155
 156          if (progress) {
 157             nir_metadata_preserve(function->impl,
 158                                   nir_metadata_block_index |
 159                                   nir_metadata_dominance);
 160          } else {
 161 #ifndef NDEBUG
 162             function->impl->valid_metadata &= ~nir_metadata_not_properly_reset;
 163 #endif
 164          }
 165       }
 166    }
 167
 168    return progress;
 169 }