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    /* See resolve_ud_negate() and comment in brw_fs_emit.cpp. */
  44    if (inst->conditional_mod &&
  45        inst->src[arg].type == BRW_REGISTER_TYPE_UD &&
  46        entry->src.negate)
  47       return false;
  48
  49    bool has_source_modifiers = entry->src.abs || entry->src.negate;
  50
  51    if (intel->gen == 6 && inst->is_math() &&
  52        (has_source_modifiers || entry->src.file == UNIFORM))
  53       return false;
  54
  55    inst->src[arg].file = entry->src.file;
  56    inst->src[arg].reg = entry->src.reg;
  57    inst->src[arg].reg_offset = entry->src.reg_offset;
  58
  59    if (!inst->src[arg].abs) {
  60       inst->src[arg].abs = entry->src.abs;
  61       inst->src[arg].negate ^= entry->src.negate;
  62    }
  63
  64    return true;
  65 }
  66
  67 /** @file brw_fs_copy_propagation.cpp
  68  *
  69  * Support for local copy propagation by walking the list of instructions
  70  * and maintaining the ACP table of available copies for propagation.
  71  *
  72  * See Muchnik's Advanced Compiler Design and Implementation, section
  73  * 12.5 (p356).
  74  */
  75
  76 /* Walks a basic block and does copy propagation on it using the acp
  77  * list.
  78  */
  79 bool
  80 fs_visitor::opt_copy_propagate_local(void *mem_ctx,
  81                                      fs_bblock *block, exec_list *acp)
  82 {
  83    bool progress = false;
  84
  85    for (fs_inst *inst = block->start;
  86         inst != block->end->next;
  87         inst = (fs_inst *)inst->next) {
  88
  89       /* Try propagating into this instruction. */
  90       foreach_list(entry_node, acp) {
  91          acp_entry *entry = (acp_entry *)entry_node;
  92
  93          for (int i = 0; i < 3; i++) {
  94             if (try_copy_propagate(inst, i, entry))
  95                progress = true;
  96          }
  97       }
  98
  99       /* kill the destination from the ACP */
 100       if (inst->dst.file == GRF) {
 101          int start_offset = inst->dst.reg_offset;
 102          int end_offset = start_offset + inst->regs_written();
 103
 104          foreach_list_safe(entry_node, acp) {
 105             acp_entry *entry = (acp_entry *)entry_node;
 106
 107             if (entry->dst.file == GRF &&
 108                 entry->dst.reg == inst->dst.reg &&
 109                 entry->dst.reg_offset >= start_offset &&
 110                 entry->dst.reg_offset < end_offset) {
 111                entry->remove();
 112                continue;
 113             }
 114             if (entry->src.file == GRF &&
 115                 entry->src.reg == inst->dst.reg &&
 116                 entry->src.reg_offset >= start_offset &&
 117                 entry->src.reg_offset < end_offset) {
 118                entry->remove();
 119             }
 120          }
 121       }
 122
 123       /* If this instruction is a raw copy, add it to the ACP. */
 124       if (inst->opcode == BRW_OPCODE_MOV &&
 125           inst->dst.file == GRF &&
 126           ((inst->src[0].file == GRF &&
 127             (inst->src[0].reg != inst->dst.reg ||
 128              inst->src[0].reg_offset != inst->dst.reg_offset)) ||
 129            inst->src[0].file == UNIFORM) &&
 130           inst->src[0].type == inst->dst.type &&
 131           !inst->saturate &&
 132           !inst->predicated &&
 133           !inst->force_uncompressed &&
 134           !inst->force_sechalf &&
 135           inst->src[0].smear == -1) {
 136          acp_entry *entry = ralloc(mem_ctx, acp_entry);
 137          entry->dst = inst->dst;
 138          entry->src = inst->src[0];
 139          acp->push_tail(entry);
 140       }
 141    }
 142
 143    return progress;
 144 }
 145
 146 bool
 147 fs_visitor::opt_copy_propagate()
 148 {
 149    bool progress = false;
 150    void *mem_ctx = ralloc_context(this->mem_ctx);
 151
 152    fs_cfg cfg(this);
 153
 154    for (int b = 0; b < cfg.num_blocks; b++) {
 155       fs_bblock *block = cfg.blocks[b];
 156       exec_list acp;
 157
 158       progress = opt_copy_propagate_local(mem_ctx, block, &acp) || progress;
 159    }
 160
 161    ralloc_free(mem_ctx);
 162
 163    if (progress)
 164       live_intervals_valid = false;
 165
 166    return progress;
 167 }