src/gallium/drivers/vc4/vc4_opt_peephole_sf.c

   1 /*
   2  * Copyright © 2016 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_peephole_sf.c
  26  *
  27  * Quick optimization to eliminate unused or identical SF updates.
  28  */
  29
  30 #include "vc4_qir.h"
  31 #include "util/u_math.h"
  32
  33 static bool debug;
  34
  35 static void
  36 dump_from(struct vc4_compile *c, struct qinst *inst, const char *type)
  37 {
  38         if (!debug)
  39                 return;
  40
  41         fprintf(stderr, "optimizing %s: ", type);
  42         qir_dump_inst(c, inst);
  43         fprintf(stderr, "\n");
  44 }
  45
  46 static void
  47 dump_to(struct vc4_compile *c, struct qinst *inst)
  48 {
  49         if (!debug)
  50                 return;
  51
  52         fprintf(stderr, "to: ");
  53         qir_dump_inst(c, inst);
  54         fprintf(stderr, "\n");
  55 }
  56
  57 static bool
  58 inst_srcs_updated(struct qinst *inst, struct qinst *writer)
  59 {
  60         /* If the sources get overwritten, stop tracking the
  61          * last instruction writing SF.
  62          */
  63         switch (writer->dst.file) {
  64         case QFILE_TEMP:
  65                 for (int i = 0; i < qir_get_op_nsrc(inst->op); i++) {
  66                         if (inst->src[i].file == QFILE_TEMP &&
  67                             inst->src[i].index == writer->dst.index) {
  68                                 return true;
  69                         }
  70                 }
  71                 return false;
  72         default:
  73                 return false;
  74         }
  75 }
  76
  77 static bool
  78 src_file_varies_on_reread(struct qreg reg)
  79 {
  80         switch (reg.file) {
  81         case QFILE_VARY:
  82         case QFILE_VPM:
  83                 return true;
  84         default:
  85                 return false;
  86         }
  87 }
  88
  89 static bool
  90 inst_result_equals(struct qinst *a, struct qinst *b)
  91 {
  92         if (a->op != b->op ||
  93             qir_depends_on_flags(a) ||
  94             qir_depends_on_flags(b)) {
  95                 return false;
  96         }
  97
  98         for (int i = 0; i < qir_get_op_nsrc(a->op); i++) {
  99                 if (!qir_reg_equals(a->src[i], b->src[i]) ||
 100                     src_file_varies_on_reread(a->src[i]) ||
 101                     src_file_varies_on_reread(b->src[i])) {
 102                         return false;
 103                 }
 104         }
 105
 106         return true;
 107 }
 108
 109 static bool
 110 qir_opt_peephole_sf_block(struct vc4_compile *c, struct qblock *block)
 111 {
 112         bool progress = false;
 113         /* We don't have liveness dataflow analysis for flags, but we also
 114          * never generate a use of flags across control flow, so just treat
 115          * them as unused at block exit.
 116          */
 117         bool sf_live = false;
 118         struct qinst *last_sf = NULL;
 119
 120         /* Walk the block from bottom to top, tracking if the SF is used, and
 121          * removing unused or repeated ones.
 122          */
 123         qir_for_each_inst_rev(inst, block) {
 124                 if (inst->sf) {
 125                         if (!sf_live) {
 126                                 /* Our instruction's SF isn't read, so drop it.
 127                                  */
 128                                 dump_from(c, inst, "dead SF");
 129                                 inst->sf = false;
 130                                 dump_to(c, inst);
 131                                 progress = true;
 132                         } else if (last_sf &&
 133                                    inst_result_equals(last_sf, inst)) {
 134                                 /* The last_sf sets up same value as inst, so
 135                                  * just drop the later one.
 136                                  */
 137                                 dump_from(c, last_sf, "repeated SF");
 138                                 last_sf->sf = false;
 139                                 dump_to(c, last_sf);
 140                                 progress = true;
 141                                 last_sf = inst;
 142                         } else {
 143                                 last_sf = inst;
 144                         }
 145                         sf_live = false;
 146                 }
 147
 148                 if (last_sf) {
 149                         if (inst_srcs_updated(last_sf, inst))
 150                                 last_sf = NULL;
 151                 }
 152
 153                 if (qir_depends_on_flags(inst))
 154                         sf_live = true;
 155         }
 156
 157         return progress;
 158 }
 159
 160 bool
 161 qir_opt_peephole_sf(struct vc4_compile *c)
 162 {
 163         bool progress = false;
 164
 165         qir_for_each_block(block, c)
 166                 progress = qir_opt_peephole_sf_block(c, block) || progress;
 167
 168         return progress;
 169 }