src/gallium/drivers/vc4/vc4_opt_vpm.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_vpm.c
  26  *
  27  * This modifies instructions that:
  28  * 1. exclusively consume a value read from the VPM to directly read the VPM if
  29  *    other operands allow it.
  30  * 2. generate the value consumed by a VPM write to write directly into the VPM.
  31  */
  32
  33 #include "vc4_qir.h"
  34
  35 bool
  36 qir_opt_vpm(struct vc4_compile *c)
  37 {
  38         if (c->stage == QSTAGE_FRAG)
  39                 return false;
  40
  41         /* For now, only do this pass when we don't have control flow. */
  42         struct qblock *block = qir_entry_block(c);
  43         if (block != qir_exit_block(c))
  44                 return false;
  45
  46         bool progress = false;
  47         struct qinst *vpm_writes[64] = { 0 };
  48         uint32_t use_count[c->num_temps];
  49         uint32_t vpm_write_count = 0;
  50         memset(&use_count, 0, sizeof(use_count));
  51
  52         qir_for_each_inst_inorder(inst, c) {
  53                 switch (inst->dst.file) {
  54                 case QFILE_VPM:
  55                         vpm_writes[vpm_write_count++] = inst;
  56                         break;
  57                 default:
  58                         break;
  59                 }
  60
  61                 for (int i = 0; i < qir_get_op_nsrc(inst->op); i++) {
  62                         if (inst->src[i].file == QFILE_TEMP) {
  63                                 uint32_t temp = inst->src[i].index;
  64                                 use_count[temp]++;
  65                         }
  66                 }
  67         }
  68
  69         /* For instructions reading from a temporary that contains a VPM read
  70          * result, try to move the instruction up in place of the VPM read.
  71          */
  72         qir_for_each_inst_inorder(inst, c) {
  73                 if (!inst)
  74                         continue;
  75
  76                 if (qir_depends_on_flags(inst) || inst->sf)
  77                         continue;
  78
  79                 if (qir_has_side_effects(c, inst) ||
  80                     qir_has_side_effect_reads(c, inst) ||
  81                     qir_is_tex(inst))
  82                         continue;
  83
  84                 for (int j = 0; j < qir_get_op_nsrc(inst->op); j++) {
  85                         if (inst->src[j].file != QFILE_TEMP ||
  86                             inst->src[j].pack)
  87                                 continue;
  88
  89                         uint32_t temp = inst->src[j].index;
  90
  91                         /* Since VPM reads pull from a FIFO, we only get to
  92                          * read each VPM entry once (unless we reset the read
  93                          * pointer).  That means we can't copy-propagate a VPM
  94                          * read to multiple locations.
  95                          */
  96                         if (use_count[temp] != 1)
  97                                 continue;
  98
  99                         struct qinst *mov = c->defs[temp];
 100                         if (!mov ||
 101                             (mov->op != QOP_MOV &&
 102                              mov->op != QOP_FMOV &&
 103                              mov->op != QOP_MMOV) ||
 104                             mov->src[0].file != QFILE_VPM) {
 105                                 continue;
 106                         }
 107
 108                         uint32_t temps = 0;
 109                         for (int k = 0; k < qir_get_op_nsrc(inst->op); k++) {
 110                                 if (inst->src[k].file == QFILE_TEMP)
 111                                         temps++;
 112                         }
 113
 114                         /* The instruction is safe to reorder if its other
 115                          * sources are independent of previous instructions
 116                          */
 117                         if (temps == 1) {
 118                                 inst->src[j] = mov->src[0];
 119
 120                                 list_del(&inst->link);
 121                                 list_addtail(&inst->link, &mov->link);
 122                                 qir_remove_instruction(c, mov);
 123
 124                                 progress = true;
 125                                 break;
 126                         }
 127                 }
 128         }
 129
 130         for (int i = 0; i < vpm_write_count; i++) {
 131                 if (!qir_is_raw_mov(vpm_writes[i]) ||
 132                     vpm_writes[i]->src[0].file != QFILE_TEMP) {
 133                         continue;
 134                 }
 135
 136                 uint32_t temp = vpm_writes[i]->src[0].index;
 137                 if (use_count[temp] != 1)
 138                         continue;
 139
 140                 struct qinst *inst = c->defs[temp];
 141                 if (!inst)
 142                         continue;
 143
 144                 if (qir_depends_on_flags(inst) || inst->sf)
 145                         continue;
 146
 147                 if (qir_has_side_effects(c, inst) ||
 148                     qir_has_side_effect_reads(c, inst)) {
 149                         continue;
 150                 }
 151
 152                 /* Move the generating instruction to the end of the program
 153                  * to maintain the order of the VPM writes.
 154                  */
 155                 assert(!vpm_writes[i]->sf);
 156                 list_del(&inst->link);
 157                 list_addtail(&inst->link, &vpm_writes[i]->link);
 158                 qir_remove_instruction(c, vpm_writes[i]);
 159
 160                 c->defs[inst->dst.index] = NULL;
 161                 inst->dst.file = QFILE_VPM;
 162                 inst->dst.index = 0;
 163
 164                 progress = true;
 165         }
 166
 167         return progress;
 168 }