src/gallium/drivers/vc4/vc4_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 vc4_opt_dead_code.c
  26  *
  27  * This is a simple dead code eliminator for SSA values in QIR.
  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->QIR.
  36  */
  37
  38 #include "vc4_qir.h"
  39
  40 static bool debug;
  41
  42 static void
  43 dce(struct vc4_compile *c, struct qinst *inst)
  44 {
  45         if (debug) {
  46                 fprintf(stderr, "Removing: ");
  47                 qir_dump_inst(c, inst);
  48                 fprintf(stderr, "\n");
  49         }
  50         assert(!inst->sf);
  51         qir_remove_instruction(c, inst);
  52 }
  53
  54 static bool
  55 has_nonremovable_reads(struct vc4_compile *c, struct qinst *inst)
  56 {
  57         for (int i = 0; i < qir_get_nsrc(inst); i++) {
  58                 if (inst->src[i].file == QFILE_VPM) {
  59                         uint32_t attr = inst->src[i].index / 4;
  60                         uint32_t offset = (inst->src[i].index % 4) * 4;
  61
  62                         if (c->vattr_sizes[attr] != offset + 4)
  63                                 return true;
  64
  65                         /* Can't get rid of the last VPM read, or the
  66                          * simulator (at least) throws an error.
  67                          */
  68                         uint32_t total_size = 0;
  69                         for (uint32_t i = 0; i < ARRAY_SIZE(c->vattr_sizes); i++)
  70                                 total_size += c->vattr_sizes[i];
  71                         if (total_size == 4)
  72                                 return true;
  73                 }
  74
  75                 if (inst->src[i].file == QFILE_VARY &&
  76                     c->input_slots[inst->src[i].index].slot == 0xff) {
  77                         return true;
  78                 }
  79         }
  80
  81         return false;
  82 }
  83
  84 bool
  85 qir_opt_dead_code(struct vc4_compile *c)
  86 {
  87         bool progress = false;
  88         bool *used = calloc(c->num_temps, sizeof(bool));
  89
  90         qir_for_each_inst_inorder(inst, c) {
  91                 for (int i = 0; i < qir_get_nsrc(inst); i++) {
  92                         if (inst->src[i].file == QFILE_TEMP)
  93                                 used[inst->src[i].index] = true;
  94                 }
  95         }
  96
  97         qir_for_each_block(block, c) {
  98                 qir_for_each_inst_safe(inst, block) {
  99                         if (inst->dst.file != QFILE_NULL &&
 100                             !(inst->dst.file == QFILE_TEMP &&
 101                               !used[inst->dst.index])) {
 102                                 continue;
 103                         }
 104
 105                         if (qir_has_side_effects(c, inst))
 106                                 continue;
 107
 108                         if (inst->sf ||
 109                             has_nonremovable_reads(c, inst)) {
 110                                 /* If we can't remove the instruction, but we
 111                                  * don't need its destination value, just
 112                                  * remove the destination.  The register
 113                                  * allocator would trivially color it and it
 114                                  * wouldn't cause any register pressure, but
 115                                  * it's nicer to read the QIR code without
 116                                  * unused destination regs.
 117                                  */
 118                                 if (inst->dst.file == QFILE_TEMP) {
 119                                         if (debug) {
 120                                                 fprintf(stderr,
 121                                                         "Removing dst from: ");
 122                                                 qir_dump_inst(c, inst);
 123                                                 fprintf(stderr, "\n");
 124                                         }
 125                                         c->defs[inst->dst.index] = NULL;
 126                                         inst->dst.file = QFILE_NULL;
 127                                         progress = true;
 128                                 }
 129                                 continue;
 130                         }
 131
 132                         for (int i = 0; i < qir_get_nsrc(inst); i++) {
 133                                 if (inst->src[i].file != QFILE_VPM)
 134                                         continue;
 135                                 uint32_t attr = inst->src[i].index / 4;
 136                                 uint32_t offset = (inst->src[i].index % 4) * 4;
 137
 138                                 if (c->vattr_sizes[attr] == offset + 4) {
 139                                         c->num_inputs--;
 140                                         c->vattr_sizes[attr] -= 4;
 141                                 }
 142                         }
 143
 144                         dce(c, inst);
 145                         progress = true;
 146                         continue;
 147                 }
 148         }
 149
 150         free(used);
 151
 152         return progress;
 153 }