src/broadcom/compiler/vir_opt_dead_code.c

   1 /*
   2  * Copyright © 2014 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 /**
  25  * @file v3d_opt_dead_code.c
  26  *
  27  * This is a simple dead code eliminator for SSA values in VIR.
  28  *
  29  * It walks all the instructions finding what temps are used, then walks again
  30  * to remove instructions writing unused temps.
  31  *
  32  * This is an inefficient implementation if you have long chains of
  33  * instructions where the entire chain is dead, but we expect those to have
  34  * been eliminated at the NIR level, and here we're just cleaning up small
  35  * problems produced by NIR->VIR.
  36  */
  37
  38 #include "v3d_compiler.h"
  39
  40 static bool debug;
  41
  42 static void
  43 dce(struct v3d_compile *c, struct qinst *inst)
  44 {
  45         if (debug) {
  46                 fprintf(stderr, "Removing: ");
  47                 vir_dump_inst(c, inst);
  48                 fprintf(stderr, "\n");
  49         }
  50         assert(inst->qpu.flags.apf == V3D_QPU_PF_NONE);
  51         assert(inst->qpu.flags.mpf == V3D_QPU_PF_NONE);
  52         vir_remove_instruction(c, inst);
  53 }
  54
  55 static bool
  56 has_nonremovable_reads(struct v3d_compile *c, struct qinst *inst)
  57 {
  58         for (int i = 0; i < vir_get_nsrc(inst); i++) {
  59                 if (inst->src[i].file == QFILE_VPM) {
  60                         /* Instance ID, Vertex ID: Should have been removed at
  61                          * the NIR level
  62                          */
  63                         if (inst->src[i].index == ~0)
  64                                 return true;
  65
  66                         uint32_t attr = inst->src[i].index / 4;
  67                         uint32_t offset = inst->src[i].index % 4;
  68
  69                         if (c->vattr_sizes[attr] != offset)
  70                                 return true;
  71
  72                         /* Can't get rid of the last VPM read, or the
  73                          * simulator (at least) throws an error.
  74                          */
  75                         uint32_t total_size = 0;
  76                         for (uint32_t i = 0; i < ARRAY_SIZE(c->vattr_sizes); i++)
  77                                 total_size += c->vattr_sizes[i];
  78                         if (total_size == 1)
  79                                 return true;
  80                 }
  81
  82                 /* Dead code removal of varyings is tricky, so just assert
  83                  * that it all happened at the NIR level.
  84                  */
  85                 if (inst->src[i].file == QFILE_VARY)
  86                         return true;
  87         }
  88
  89         return false;
  90 }
  91
  92 bool
  93 vir_opt_dead_code(struct v3d_compile *c)
  94 {
  95         bool progress = false;
  96         bool *used = calloc(c->num_temps, sizeof(bool));
  97
  98         vir_for_each_inst_inorder(inst, c) {
  99                 for (int i = 0; i < vir_get_nsrc(inst); i++) {
 100                         if (inst->src[i].file == QFILE_TEMP)
 101                                 used[inst->src[i].index] = true;
 102                 }
 103         }
 104
 105         vir_for_each_block(block, c) {
 106                 vir_for_each_inst_safe(inst, block) {
 107                         if (inst->dst.file != QFILE_NULL &&
 108                             !(inst->dst.file == QFILE_TEMP &&
 109                               !used[inst->dst.index])) {
 110                                 continue;
 111                         }
 112
 113                         if (vir_has_side_effects(c, inst))
 114                                 continue;
 115
 116                         if (inst->qpu.flags.apf != V3D_QPU_PF_NONE ||
 117                             inst->qpu.flags.mpf != V3D_QPU_PF_NONE||
 118                             has_nonremovable_reads(c, inst)) {
 119                                 /* If we can't remove the instruction, but we
 120                                  * don't need its destination value, just
 121                                  * remove the destination.  The register
 122                                  * allocator would trivially color it and it
 123                                  * wouldn't cause any register pressure, but
 124                                  * it's nicer to read the VIR code without
 125                                  * unused destination regs.
 126                                  */
 127                                 if (inst->dst.file == QFILE_TEMP) {
 128                                         if (debug) {
 129                                                 fprintf(stderr,
 130                                                         "Removing dst from: ");
 131                                                 vir_dump_inst(c, inst);
 132                                                 fprintf(stderr, "\n");
 133                                         }
 134                                         c->defs[inst->dst.index] = NULL;
 135                                         inst->dst.file = QFILE_NULL;
 136                                         progress = true;
 137                                 }
 138                                 continue;
 139                         }
 140
 141                         for (int i = 0; i < vir_get_nsrc(inst); i++) {
 142                                 if (inst->src[i].file != QFILE_VPM)
 143                                         continue;
 144                                 uint32_t attr = inst->src[i].index / 4;
 145                                 uint32_t offset = (inst->src[i].index % 4);
 146
 147                                 if (c->vattr_sizes[attr] == offset) {
 148                                         c->num_inputs--;
 149                                         c->vattr_sizes[attr]--;
 150                                 }
 151                         }
 152
 153                         dce(c, inst);
 154                         progress = true;
 155                         continue;
 156                 }
 157         }
 158
 159         free(used);
 160
 161         return progress;
 162 }