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 exclusively consume a value read from the
  28  * VPM to directly read the VPM if other operands allow it.
  29  */
  30
  31 #include "vc4_qir.h"
  32
  33 bool
  34 qir_opt_vpm(struct vc4_compile *c)
  35 {
  36         if (c->stage == QSTAGE_FRAG)
  37                 return false;
  38
  39         /* For now, only do this pass when we don't have control flow. */
  40         struct qblock *block = qir_entry_block(c);
  41         if (block != qir_exit_block(c))
  42                 return false;
  43
  44         bool progress = false;
  45         uint32_t use_count[c->num_temps];
  46         memset(&use_count, 0, sizeof(use_count));
  47
  48         qir_for_each_inst_inorder(inst, c) {
  49                 for (int i = 0; i < qir_get_nsrc(inst); i++) {
  50                         if (inst->src[i].file == QFILE_TEMP) {
  51                                 uint32_t temp = inst->src[i].index;
  52                                 use_count[temp]++;
  53                         }
  54                 }
  55         }
  56
  57         /* For instructions reading from a temporary that contains a VPM read
  58          * result, try to move the instruction up in place of the VPM read.
  59          */
  60         qir_for_each_inst_inorder(inst, c) {
  61                 if (!inst)
  62                         continue;
  63
  64                 if (qir_depends_on_flags(inst) || inst->sf)
  65                         continue;
  66
  67                 if (qir_has_side_effects(c, inst) ||
  68                     qir_has_side_effect_reads(c, inst) ||
  69                     qir_is_tex(inst))
  70                         continue;
  71
  72                 for (int j = 0; j < qir_get_nsrc(inst); j++) {
  73                         if (inst->src[j].file != QFILE_TEMP ||
  74                             inst->src[j].pack)
  75                                 continue;
  76
  77                         uint32_t temp = inst->src[j].index;
  78
  79                         /* Since VPM reads pull from a FIFO, we only get to
  80                          * read each VPM entry once (unless we reset the read
  81                          * pointer).  That means we can't copy-propagate a VPM
  82                          * read to multiple locations.
  83                          */
  84                         if (use_count[temp] != 1)
  85                                 continue;
  86
  87                         struct qinst *mov = c->defs[temp];
  88                         if (!mov ||
  89                             (mov->op != QOP_MOV &&
  90                              mov->op != QOP_FMOV &&
  91                              mov->op != QOP_MMOV) ||
  92                             mov->src[0].file != QFILE_VPM) {
  93                                 continue;
  94                         }
  95
  96                         uint32_t temps = 0;
  97                         for (int k = 0; k < qir_get_nsrc(inst); k++) {
  98                                 if (inst->src[k].file == QFILE_TEMP)
  99                                         temps++;
 100                         }
 101
 102                         /* The instruction is safe to reorder if its other
 103                          * sources are independent of previous instructions
 104                          */
 105                         if (temps == 1) {
 106                                 inst->src[j] = mov->src[0];
 107
 108                                 list_del(&inst->link);
 109                                 list_addtail(&inst->link, &mov->link);
 110                                 qir_remove_instruction(c, mov);
 111
 112                                 progress = true;
 113                                 break;
 114                         }
 115                 }
 116         }
 117
 118         return progress;
 119 }