src/mesa/drivers/dri/i965/brw_vec4_dead_code_eliminate.cpp

   1 /*
   2  * Copyright © 2014 Intel Corporation
   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 #include "brw_vec4.h"
  25 #include "brw_vec4_live_variables.h"
  26 #include "brw_cfg.h"
  27
  28 /** @file brw_vec4_dead_code_eliminate.cpp
  29  *
  30  * Dataflow-aware dead code elimination.
  31  *
  32  * Walks the instruction list from the bottom, removing instructions that
  33  * have results that both aren't used in later blocks and haven't been read
  34  * yet in the tail end of this block.
  35  */
  36
  37 using namespace brw;
  38
  39 bool
  40 vec4_visitor::dead_code_eliminate()
  41 {
  42    bool progress = false;
  43
  44    calculate_live_intervals();
  45
  46    int num_vars = live_intervals->num_vars;
  47    BITSET_WORD *live = ralloc_array(NULL, BITSET_WORD, BITSET_WORDS(num_vars));
  48    BITSET_WORD *flag_live = ralloc_array(NULL, BITSET_WORD, 1);
  49
  50    foreach_block_reverse_safe(block, cfg) {
  51       memcpy(live, live_intervals->block_data[block->num].liveout,
  52              sizeof(BITSET_WORD) * BITSET_WORDS(num_vars));
  53       memcpy(flag_live, live_intervals->block_data[block->num].flag_liveout,
  54              sizeof(BITSET_WORD));
  55
  56       foreach_inst_in_block_reverse_safe(vec4_instruction, inst, block) {
  57          if ((inst->dst.file == VGRF && !inst->has_side_effects()) ||
  58              (inst->dst.is_null() && inst->writes_flag())){
  59             bool result_live[4] = { false };
  60
  61             if (inst->dst.file == VGRF) {
  62                for (unsigned i = 0; i < regs_written(inst); i++) {
  63                   for (int c = 0; c < 4; c++)
  64                      result_live[c] |= BITSET_TEST(live,
  65                         var_from_reg(alloc,
  66                                      byte_offset(inst->dst, i * REG_SIZE), c));
  67                }
  68             } else {
  69                for (unsigned c = 0; c < 4; c++)
  70                   result_live[c] = BITSET_TEST(flag_live, c);
  71             }
  72
  73             /* If the instruction can't do writemasking, then it's all or
  74              * nothing.
  75              */
  76             if (!inst->can_do_writemask(devinfo)) {
  77                bool result = result_live[0] | result_live[1] |
  78                              result_live[2] | result_live[3];
  79                result_live[0] = result;
  80                result_live[1] = result;
  81                result_live[2] = result;
  82                result_live[3] = result;
  83             }
  84
  85             for (int c = 0; c < 4; c++) {
  86                if (!result_live[c] && inst->dst.writemask & (1 << c)) {
  87                   inst->dst.writemask &= ~(1 << c);
  88                   progress = true;
  89
  90                   if (inst->dst.writemask == 0) {
  91                      if (inst->writes_accumulator || inst->writes_flag()) {
  92                         inst->dst = dst_reg(retype(brw_null_reg(), inst->dst.type));
  93                      } else {
  94                         inst->opcode = BRW_OPCODE_NOP;
  95                         break;
  96                      }
  97                   }
  98                }
  99             }
 100          }
 101
 102          if (inst->dst.is_null() && inst->writes_flag()) {
 103             bool combined_live = false;
 104             for (unsigned c = 0; c < 4; c++)
 105                combined_live |= BITSET_TEST(flag_live, c);
 106
 107             if (!combined_live) {
 108                inst->opcode = BRW_OPCODE_NOP;
 109                progress = true;
 110             }
 111          }
 112
 113          if (inst->dst.file == VGRF && !inst->predicate) {
 114             for (unsigned i = 0; i < regs_written(inst); i++) {
 115                for (int c = 0; c < 4; c++) {
 116                   if (inst->dst.writemask & (1 << c)) {
 117                      BITSET_CLEAR(live,
 118                                   var_from_reg(alloc,
 119                                                byte_offset(inst->dst,
 120                                                            i * REG_SIZE),
 121                                                c));
 122                   }
 123                }
 124             }
 125          }
 126
 127          if (inst->writes_flag() && !inst->predicate) {
 128             for (unsigned c = 0; c < 4; c++)
 129                BITSET_CLEAR(flag_live, c);
 130          }
 131
 132          if (inst->opcode == BRW_OPCODE_NOP) {
 133             inst->remove(block);
 134             continue;
 135          }
 136
 137          for (int i = 0; i < 3; i++) {
 138             if (inst->src[i].file == VGRF) {
 139                for (unsigned j = 0; j < regs_read(inst, i); j++) {
 140                   for (int c = 0; c < 4; c++) {
 141                      BITSET_SET(live, var_from_reg(alloc,
 142                                                    byte_offset(inst->src[i],
 143                                                                j * REG_SIZE),
 144                                                    c));
 145                   }
 146                }
 147             }
 148          }
 149
 150          for (unsigned c = 0; c < 4; c++) {
 151             if (inst->reads_flag(c)) {
 152                BITSET_SET(flag_live, c);
 153             }
 154          }
 155       }
 156    }
 157
 158    ralloc_free(live);
 159    ralloc_free(flag_live);
 160
 161    if (progress)
 162       invalidate_live_intervals();
 163
 164    return progress;
 165 }