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 void
  37 alu_src_consume_abs(nir_alu_src *src)
  38 {
  39    src->abs = true;
  40 }
  41
  42 static void
  43 alu_src_consume_negate(nir_alu_src *src)
  44 {
  45    /* If abs is set on the source, the negate goes away */
  46    if (!src->abs)
  47       src->negate = !src->negate;
  48 }
  49
  50 static bool
  51 nir_lower_to_source_mods_block(nir_block *block,
  52                                nir_lower_to_source_mods_flags options)
  53 {
  54    bool progress = false;
  55
  56    nir_foreach_instr(instr, block) {
  57       if (instr->type != nir_instr_type_alu)
  58          continue;
  59
  60       nir_alu_instr *alu = nir_instr_as_alu(instr);
  61
  62       bool lower_abs = (nir_op_infos[alu->op].num_inputs < 3) ||
  63             (options & nir_lower_triop_abs);
  64
  65       for (unsigned i = 0; i < nir_op_infos[alu->op].num_inputs; i++) {
  66          if (!alu->src[i].src.is_ssa)
  67             continue;
  68
  69          if (alu->src[i].src.ssa->parent_instr->type != nir_instr_type_alu)
  70             continue;
  71
  72          nir_alu_instr *parent = nir_instr_as_alu(alu->src[i].src.ssa->parent_instr);
  73
  74          if (parent->dest.saturate)
  75             continue;
  76
  77          switch (nir_alu_type_get_base_type(nir_op_infos[alu->op].input_types[i])) {
  78          case nir_type_float:
  79             if (!(options & nir_lower_float_source_mods))
  80                continue;
  81             if (parent->op != nir_op_fabs && parent->op != nir_op_fneg)
  82                continue;
  83             break;
  84          case nir_type_int:
  85             if (!(options & nir_lower_int_source_mods))
  86                continue;
  87             if (parent->op != nir_op_iabs && parent->op != nir_op_ineg)
  88                continue;
  89             break;
  90          default:
  91             continue;
  92          }
  93
  94          /* We can only do a rewrite if the source we are copying is SSA.
  95           * Otherwise, moving the read might invalidly reorder reads/writes
  96           * on a register.
  97           */
  98          if (!parent->src[0].src.is_ssa)
  99             continue;
 100
 101          if (!lower_abs && (parent->op == nir_op_fabs ||
 102                             parent->op == nir_op_iabs))
 103             continue;
 104
 105          nir_instr_rewrite_src(instr, &alu->src[i].src, parent->src[0].src);
 106
 107          /* Apply any modifiers that come from the parent opcode */
 108          if (parent->op == nir_op_fneg || parent->op == nir_op_ineg)
 109             alu_src_consume_negate(&alu->src[i]);
 110          if (parent->op == nir_op_fabs || parent->op == nir_op_iabs)
 111             alu_src_consume_abs(&alu->src[i]);
 112
 113          /* Apply modifiers from the parent source */
 114          if (parent->src[0].negate)
 115             alu_src_consume_negate(&alu->src[i]);
 116          if (parent->src[0].abs)
 117             alu_src_consume_abs(&alu->src[i]);
 118
 119          for (int j = 0; j < 4; ++j) {
 120             if (!nir_alu_instr_channel_used(alu, i, j))
 121                continue;
 122             alu->src[i].swizzle[j] = parent->src[0].swizzle[alu->src[i].swizzle[j]];
 123          }
 124
 125          if (list_is_empty(&parent->dest.dest.ssa.uses) &&
 126              list_is_empty(&parent->dest.dest.ssa.if_uses))
 127             nir_instr_remove(&parent->instr);
 128
 129          progress = true;
 130       }
 131
 132       /* We've covered sources.  Now we're going to try and saturate the
 133        * destination if we can.
 134        */
 135
 136       if (!alu->dest.dest.is_ssa)
 137          continue;
 138
 139       /* We can only saturate float destinations */
 140       if (nir_alu_type_get_base_type(nir_op_infos[alu->op].output_type) !=
 141           nir_type_float)
 142          continue;
 143
 144       if (!(options & nir_lower_float_source_mods))
 145          continue;
 146
 147       if (!list_is_empty(&alu->dest.dest.ssa.if_uses))
 148          continue;
 149
 150       bool all_children_are_sat = true;
 151       nir_foreach_use(child_src, &alu->dest.dest.ssa) {
 152          assert(child_src->is_ssa);
 153          nir_instr *child = child_src->parent_instr;
 154          if (child->type != nir_instr_type_alu) {
 155             all_children_are_sat = false;
 156             continue;
 157          }
 158
 159          nir_alu_instr *child_alu = nir_instr_as_alu(child);
 160          if (child_alu->src[0].negate || child_alu->src[0].abs) {
 161             all_children_are_sat = false;
 162             continue;
 163          }
 164
 165          if (child_alu->op != nir_op_fsat) {
 166             all_children_are_sat = false;
 167             continue;
 168          }
 169       }
 170
 171       if (!all_children_are_sat)
 172          continue;
 173
 174       alu->dest.saturate = true;
 175       progress = true;
 176
 177       nir_foreach_use(child_src, &alu->dest.dest.ssa) {
 178          assert(child_src->is_ssa);
 179          nir_alu_instr *child_alu = nir_instr_as_alu(child_src->parent_instr);
 180
 181          child_alu->op = nir_op_mov;
 182          child_alu->dest.saturate = false;
 183          /* We could propagate the dest of our instruction to the
 184           * destinations of the uses here.  However, one quick round of
 185           * copy propagation will clean that all up and then we don't have
 186           * the complexity.
 187           */
 188       }
 189    }
 190
 191    return progress;
 192 }
 193
 194 static bool
 195 nir_lower_to_source_mods_impl(nir_function_impl *impl,
 196                               nir_lower_to_source_mods_flags options)
 197 {
 198    bool progress = false;
 199
 200    nir_foreach_block(block, impl) {
 201       progress |= nir_lower_to_source_mods_block(block, options);
 202    }
 203
 204    if (progress)
 205       nir_metadata_preserve(impl, nir_metadata_block_index |
 206                                   nir_metadata_dominance);
 207
 208    return progress;
 209 }
 210
 211 bool
 212 nir_lower_to_source_mods(nir_shader *shader,
 213                          nir_lower_to_source_mods_flags options)
 214 {
 215    bool progress = false;
 216
 217    nir_foreach_function(function, shader) {
 218       if (function->impl) {
 219          progress |= nir_lower_to_source_mods_impl(function->impl, options);
 220       }
 221    }
 222
 223    return progress;
 224 }