src/compiler/nir/nir_lower_to_source_mods.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  *    Jason Ekstrand (jason@jlekstrand.net)
  25  *
  26  */
  27
  28 #include "nir.h"
  29
  30 /*
  31  * This pass lowers the neg, abs, and sat operations to source modifiers on
  32  * ALU operations to make things nicer for the backend.  It's just much
  33  * easier to not have them when we're doing optimizations.
  34  */
  35
  36 static bool
  37 nir_lower_to_source_mods_block(nir_block *block,
  38                                nir_lower_to_source_mods_flags options)
  39 {
  40    bool progress = false;
  41
  42    nir_foreach_instr(instr, block) {
  43       if (instr->type != nir_instr_type_alu)
  44          continue;
  45
  46       nir_alu_instr *alu = nir_instr_as_alu(instr);
  47
  48       for (unsigned i = 0; i < nir_op_infos[alu->op].num_inputs; i++) {
  49          if (!alu->src[i].src.is_ssa)
  50             continue;
  51
  52          if (alu->src[i].src.ssa->parent_instr->type != nir_instr_type_alu)
  53             continue;
  54
  55          nir_alu_instr *parent = nir_instr_as_alu(alu->src[i].src.ssa->parent_instr);
  56
  57          if (parent->dest.saturate)
  58             continue;
  59
  60          switch (nir_alu_type_get_base_type(nir_op_infos[alu->op].input_types[i])) {
  61          case nir_type_float:
  62             if (!(options & nir_lower_float_source_mods))
  63                continue;
  64             if (parent->op != nir_op_fmov)
  65                continue;
  66             break;
  67          case nir_type_int:
  68             if (!(options & nir_lower_int_source_mods))
  69                continue;
  70             if (parent->op != nir_op_imov)
  71                continue;
  72             break;
  73          default:
  74             continue;
  75          }
  76
  77          /* We can only do a rewrite if the source we are copying is SSA.
  78           * Otherwise, moving the read might invalidly reorder reads/writes
  79           * on a register.
  80           */
  81          if (!parent->src[0].src.is_ssa)
  82             continue;
  83
  84          nir_instr_rewrite_src(instr, &alu->src[i].src, parent->src[0].src);
  85          if (alu->src[i].abs) {
  86             /* abs trumps both neg and abs, do nothing */
  87          } else {
  88             alu->src[i].negate = (alu->src[i].negate != parent->src[0].negate);
  89             alu->src[i].abs |= parent->src[0].abs;
  90          }
  91
  92          for (int j = 0; j < 4; ++j) {
  93             if (!nir_alu_instr_channel_used(alu, i, j))
  94                continue;
  95             alu->src[i].swizzle[j] = parent->src[0].swizzle[alu->src[i].swizzle[j]];
  96          }
  97
  98          if (list_empty(&parent->dest.dest.ssa.uses) &&
  99              list_empty(&parent->dest.dest.ssa.if_uses))
 100             nir_instr_remove(&parent->instr);
 101
 102          progress = true;
 103       }
 104
 105       if (options & nir_lower_float_source_mods) {
 106          switch (alu->op) {
 107          case nir_op_fsat:
 108             alu->op = nir_op_fmov;
 109             alu->dest.saturate = true;
 110             break;
 111          case nir_op_fneg:
 112             alu->op = nir_op_fmov;
 113             alu->src[0].negate = !alu->src[0].negate;
 114             break;
 115          case nir_op_fabs:
 116             alu->op = nir_op_fmov;
 117             alu->src[0].abs = true;
 118             alu->src[0].negate = false;
 119             break;
 120          default:
 121             break;
 122          }
 123       }
 124
 125       if (options & nir_lower_int_source_mods) {
 126          switch (alu->op) {
 127          case nir_op_ineg:
 128             alu->op = nir_op_imov;
 129             alu->src[0].negate = !alu->src[0].negate;
 130             break;
 131          case nir_op_iabs:
 132             alu->op = nir_op_imov;
 133             alu->src[0].abs = true;
 134             alu->src[0].negate = false;
 135             break;
 136          default:
 137             break;
 138          }
 139       }
 140       /* We've covered sources.  Now we're going to try and saturate the
 141        * destination if we can.
 142        */
 143
 144       if (!alu->dest.dest.is_ssa)
 145          continue;
 146
 147       /* We can only saturate float destinations */
 148       if (nir_alu_type_get_base_type(nir_op_infos[alu->op].output_type) !=
 149           nir_type_float)
 150          continue;
 151
 152       if (!(options & nir_lower_float_source_mods))
 153          continue;
 154
 155       if (!list_empty(&alu->dest.dest.ssa.if_uses))
 156          continue;
 157
 158       bool all_children_are_sat = true;
 159       nir_foreach_use(child_src, &alu->dest.dest.ssa) {
 160          assert(child_src->is_ssa);
 161          nir_instr *child = child_src->parent_instr;
 162          if (child->type != nir_instr_type_alu) {
 163             all_children_are_sat = false;
 164             continue;
 165          }
 166
 167          nir_alu_instr *child_alu = nir_instr_as_alu(child);
 168          if (child_alu->src[0].negate || child_alu->src[0].abs) {
 169             all_children_are_sat = false;
 170             continue;
 171          }
 172
 173          if (child_alu->op != nir_op_fsat &&
 174              !(child_alu->op == nir_op_fmov && child_alu->dest.saturate)) {
 175             all_children_are_sat = false;
 176             continue;
 177          }
 178       }
 179
 180       if (!all_children_are_sat)
 181          continue;
 182
 183       alu->dest.saturate = true;
 184       progress = true;
 185
 186       nir_foreach_use(child_src, &alu->dest.dest.ssa) {
 187          assert(child_src->is_ssa);
 188          nir_alu_instr *child_alu = nir_instr_as_alu(child_src->parent_instr);
 189
 190          child_alu->op = nir_op_fmov;
 191          child_alu->dest.saturate = false;
 192          /* We could propagate the dest of our instruction to the
 193           * destinations of the uses here.  However, one quick round of
 194           * copy propagation will clean that all up and then we don't have
 195           * the complexity.
 196           */
 197       }
 198    }
 199
 200    return progress;
 201 }
 202
 203 static bool
 204 nir_lower_to_source_mods_impl(nir_function_impl *impl,
 205                               nir_lower_to_source_mods_flags options)
 206 {
 207    bool progress = false;
 208
 209    nir_foreach_block(block, impl) {
 210       progress |= nir_lower_to_source_mods_block(block, options);
 211    }
 212
 213    if (progress)
 214       nir_metadata_preserve(impl, nir_metadata_block_index |
 215                                   nir_metadata_dominance);
 216
 217    return progress;
 218 }
 219
 220 bool
 221 nir_lower_to_source_mods(nir_shader *shader,
 222                          nir_lower_to_source_mods_flags options)
 223 {
 224    bool progress = false;
 225
 226    nir_foreach_function(function, shader) {
 227       if (function->impl) {
 228          progress |= nir_lower_to_source_mods_impl(function->impl, options);
 229       }
 230    }
 231
 232    return progress;
 233 }