src/compiler/nir/nir_opt_intrinsics.c

   1 /*
   2  * Copyright © 2017 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
  24 #include "nir.h"
  25 #include "nir_builder.h"
  26
  27 /**
  28  * \file nir_opt_intrinsics.c
  29  */
  30
  31 static bool
  32 opt_intrinsics_impl(nir_function_impl *impl)
  33 {
  34    nir_builder b;
  35    nir_builder_init(&b, impl);
  36    bool progress = false;
  37
  38    nir_foreach_block(block, impl) {
  39       nir_foreach_instr_safe(instr, block) {
  40          if (instr->type != nir_instr_type_intrinsic)
  41             continue;
  42
  43          nir_intrinsic_instr *intrin = nir_instr_as_intrinsic(instr);
  44          nir_ssa_def *replacement = NULL;
  45          b.cursor = nir_before_instr(instr);
  46
  47          switch (intrin->intrinsic) {
  48          case nir_intrinsic_vote_any:
  49          case nir_intrinsic_vote_all: {
  50             nir_const_value *val = nir_src_as_const_value(intrin->src[0]);
  51             if (!val && !b.shader->options->lower_vote_trivial)
  52                continue;
  53
  54             replacement = nir_ssa_for_src(&b, intrin->src[0], 1);
  55             break;
  56          }
  57          case nir_intrinsic_vote_eq: {
  58             nir_const_value *val = nir_src_as_const_value(intrin->src[0]);
  59             if (!val && !b.shader->options->lower_vote_trivial)
  60                continue;
  61
  62             replacement = nir_imm_int(&b, NIR_TRUE);
  63             break;
  64          }
  65          case nir_intrinsic_ballot: {
  66             assert(b.shader->options->max_subgroup_size != 0);
  67             if (b.shader->options->max_subgroup_size > 32 ||
  68                 intrin->dest.ssa.bit_size <= 32)
  69                continue;
  70
  71             nir_intrinsic_instr *ballot =
  72                nir_intrinsic_instr_create(b.shader, nir_intrinsic_ballot);
  73             nir_ssa_dest_init(&ballot->instr, &ballot->dest, 1, 32, NULL);
  74             nir_src_copy(&ballot->src[0], &intrin->src[0], ballot);
  75
  76             nir_builder_instr_insert(&b, &ballot->instr);
  77
  78             replacement = nir_pack_64_2x32_split(&b,
  79                                                  &ballot->dest.ssa,
  80                                                  nir_imm_int(&b, 0));
  81             break;
  82          }
  83          case nir_intrinsic_load_subgroup_eq_mask:
  84          case nir_intrinsic_load_subgroup_ge_mask:
  85          case nir_intrinsic_load_subgroup_gt_mask:
  86          case nir_intrinsic_load_subgroup_le_mask:
  87          case nir_intrinsic_load_subgroup_lt_mask: {
  88             if (!b.shader->options->lower_subgroup_masks)
  89                break;
  90
  91             nir_ssa_def *count = nir_load_subgroup_invocation(&b);
  92
  93             switch (intrin->intrinsic) {
  94             case nir_intrinsic_load_subgroup_eq_mask:
  95                replacement = nir_ishl(&b, nir_imm_int64(&b, 1ull), count);
  96                break;
  97             case nir_intrinsic_load_subgroup_ge_mask:
  98                replacement = nir_ishl(&b, nir_imm_int64(&b, ~0ull), count);
  99                break;
 100             case nir_intrinsic_load_subgroup_gt_mask:
 101                replacement = nir_ishl(&b, nir_imm_int64(&b, ~1ull), count);
 102                break;
 103             case nir_intrinsic_load_subgroup_le_mask:
 104                replacement = nir_inot(&b, nir_ishl(&b, nir_imm_int64(&b, ~1ull), count));
 105                break;
 106             case nir_intrinsic_load_subgroup_lt_mask:
 107                replacement = nir_inot(&b, nir_ishl(&b, nir_imm_int64(&b, ~0ull), count));
 108                break;
 109             default:
 110                unreachable("you seriously can't tell this is unreachable?");
 111             }
 112             break;
 113          }
 114          default:
 115             break;
 116          }
 117
 118          if (!replacement)
 119             continue;
 120
 121          nir_ssa_def_rewrite_uses(&intrin->dest.ssa,
 122                                   nir_src_for_ssa(replacement));
 123          nir_instr_remove(instr);
 124          progress = true;
 125       }
 126    }
 127
 128    return progress;
 129 }
 130
 131 bool
 132 nir_opt_intrinsics(nir_shader *shader)
 133 {
 134    bool progress = false;
 135
 136    nir_foreach_function(function, shader) {
 137       if (!function->impl)
 138          continue;
 139
 140       if (opt_intrinsics_impl(function->impl)) {
 141          progress = true;
 142          nir_metadata_preserve(function->impl, nir_metadata_block_index |
 143                                                nir_metadata_dominance);
 144       }
 145    }
 146
 147    return progress;
 148 }