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

   1 /*
   2  * Copyright © 2012 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_fs.h"
  25 #include "brw_fs_cfg.h"
  26
  27 namespace { /* avoid conflict with opt_copy_propagation_elements */
  28 struct acp_entry : public exec_node {
  29    fs_reg dst;
  30    fs_reg src;
  31 };
  32 }
  33
  34 bool
  35 fs_visitor::try_copy_propagate(fs_inst *inst, int arg, acp_entry *entry)
  36 {
  37    if (inst->src[arg].file != entry->dst.file ||
  38        inst->src[arg].reg != entry->dst.reg ||
  39        inst->src[arg].reg_offset != entry->dst.reg_offset) {
  40       return false;
  41    }
  42
  43    inst->src[arg].reg = entry->src.reg;
  44    inst->src[arg].reg_offset = entry->src.reg_offset;
  45
  46    return true;
  47 }
  48
  49 /** @file brw_fs_copy_propagation.cpp
  50  *
  51  * Support for local copy propagation by walking the list of instructions
  52  * and maintaining the ACP table of available copies for propagation.
  53  *
  54  * See Muchnik's Advanced Compiler Design and Implementation, section
  55  * 12.5 (p356).
  56  */
  57
  58 /* Walks a basic block and does copy propagation on it using the acp
  59  * list.
  60  */
  61 bool
  62 fs_visitor::opt_copy_propagate_local(void *mem_ctx,
  63                                      fs_bblock *block, exec_list *acp)
  64 {
  65    bool progress = false;
  66
  67    for (fs_inst *inst = block->start;
  68         inst != block->end->next;
  69         inst = (fs_inst *)inst->next) {
  70
  71       /* Try propagating into this instruction. */
  72       foreach_list(entry_node, acp) {
  73          acp_entry *entry = (acp_entry *)entry_node;
  74
  75          for (int i = 0; i < 3; i++) {
  76             if (try_copy_propagate(inst, i, entry))
  77                progress = true;
  78          }
  79       }
  80
  81       /* kill the destination from the ACP */
  82       if (inst->dst.file == GRF) {
  83          int start_offset = inst->dst.reg_offset;
  84          int end_offset = start_offset + inst->regs_written();
  85
  86          foreach_list_safe(entry_node, acp) {
  87             acp_entry *entry = (acp_entry *)entry_node;
  88
  89             if (entry->dst.file == GRF &&
  90                 entry->dst.reg == inst->dst.reg &&
  91                 entry->dst.reg_offset >= start_offset &&
  92                 entry->dst.reg_offset < end_offset) {
  93                entry->remove();
  94                continue;
  95             }
  96             if (entry->src.file == GRF &&
  97                 entry->src.reg == inst->dst.reg &&
  98                 entry->src.reg_offset >= start_offset &&
  99                 entry->src.reg_offset < end_offset) {
 100                entry->remove();
 101             }
 102          }
 103       }
 104
 105       /* If this instruction is a raw copy, add it to the ACP. */
 106       if (inst->opcode == BRW_OPCODE_MOV &&
 107           inst->dst.file == GRF &&
 108           inst->src[0].file == GRF &&
 109           (inst->src[0].reg != inst->dst.reg ||
 110            inst->src[0].reg_offset != inst->dst.reg_offset) &&
 111           inst->src[0].type == inst->dst.type &&
 112           !inst->saturate &&
 113           !inst->predicated &&
 114           !inst->force_uncompressed &&
 115           !inst->force_sechalf &&
 116           inst->src[0].smear == -1 &&
 117           !inst->src[0].abs &&
 118           !inst->src[0].negate) {
 119          acp_entry *entry = ralloc(mem_ctx, acp_entry);
 120          entry->dst = inst->dst;
 121          entry->src = inst->src[0];
 122          acp->push_tail(entry);
 123       }
 124    }
 125
 126    return progress;
 127 }
 128
 129 bool
 130 fs_visitor::opt_copy_propagate()
 131 {
 132    bool progress = false;
 133    void *mem_ctx = ralloc_context(this->mem_ctx);
 134
 135    fs_cfg cfg(this);
 136
 137    for (int b = 0; b < cfg.num_blocks; b++) {
 138       fs_bblock *block = cfg.blocks[b];
 139       exec_list acp;
 140
 141       progress = opt_copy_propagate_local(mem_ctx, block, &acp) || progress;
 142    }
 143
 144    ralloc_free(mem_ctx);
 145
 146    return progress;
 147 }