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

   1 /*
   2  * Copyright © 2013 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_live_variables.h"
  26 #include "brw_cfg.h"
  27
  28 /** @file brw_fs_saturate_propagation.cpp
  29  *
  30  * Implements a pass that propagates the SAT modifier from a MOV.SAT into the
  31  * instruction that produced the source of the MOV.SAT, thereby allowing the
  32  * MOV's src and dst to be coalesced and the MOV removed.
  33  *
  34  * For instance,
  35  *
  36  *    ADD     tmp, src0, src1
  37  *    MOV.SAT dst, tmp
  38  *
  39  * would be transformed into
  40  *
  41  *    ADD.SAT tmp, src0, src1
  42  *    MOV     dst, tmp
  43  */
  44
  45 static bool
  46 opt_saturate_propagation_local(fs_visitor *v, bblock_t *block)
  47 {
  48    bool progress = false;
  49    int ip = block->end_ip + 1;
  50
  51    foreach_inst_in_block_reverse(fs_inst, inst, block) {
  52       ip--;
  53
  54       if (inst->opcode != BRW_OPCODE_MOV ||
  55           !inst->saturate ||
  56           inst->dst.file != VGRF ||
  57           inst->dst.type != inst->src[0].type ||
  58           inst->src[0].file != VGRF ||
  59           inst->src[0].abs)
  60          continue;
  61
  62       int src_var = v->live_intervals->var_from_reg(inst->src[0]);
  63       int src_end_ip = v->live_intervals->end[src_var];
  64
  65       bool interfered = false;
  66       foreach_inst_in_block_reverse_starting_from(fs_inst, scan_inst, inst) {
  67          if (scan_inst->overwrites_reg(inst->src[0])) {
  68             if (scan_inst->is_partial_write() ||
  69                 (scan_inst->dst.type != inst->dst.type &&
  70                  !scan_inst->can_change_types()))
  71                break;
  72
  73             if (scan_inst->saturate) {
  74                inst->saturate = false;
  75                progress = true;
  76             } else if (src_end_ip == ip || inst->dst.equals(inst->src[0])) {
  77                if (scan_inst->can_do_saturate()) {
  78                   if (scan_inst->dst.type != inst->dst.type) {
  79                      scan_inst->dst.type = inst->dst.type;
  80                      for (int i = 0; i < scan_inst->sources; i++) {
  81                         scan_inst->src[i].type = inst->dst.type;
  82                      }
  83                   }
  84
  85                   if (inst->src[0].negate) {
  86                      if (scan_inst->opcode == BRW_OPCODE_MUL) {
  87                         scan_inst->src[0].negate = !scan_inst->src[0].negate;
  88                         inst->src[0].negate = false;
  89                      } else if (scan_inst->opcode == BRW_OPCODE_MAD) {
  90                         scan_inst->src[0].negate = !scan_inst->src[0].negate;
  91                         scan_inst->src[1].negate = !scan_inst->src[1].negate;
  92                         inst->src[0].negate = false;
  93                      } else if (scan_inst->opcode == BRW_OPCODE_ADD) {
  94                         if (scan_inst->src[1].file == IMM) {
  95                            if (!brw_negate_immediate(scan_inst->src[1].type,
  96                                                      &scan_inst->src[1].as_brw_reg())) {
  97                               break;
  98                            }
  99                         } else {
 100                            scan_inst->src[1].negate = !scan_inst->src[1].negate;
 101                         }
 102                         scan_inst->src[0].negate = !scan_inst->src[0].negate;
 103                         inst->src[0].negate = false;
 104                      } else {
 105                         break;
 106                      }
 107                   }
 108
 109                   scan_inst->saturate = true;
 110                   inst->saturate = false;
 111                   progress = true;
 112                }
 113             }
 114             break;
 115          }
 116          for (int i = 0; i < scan_inst->sources; i++) {
 117             if (scan_inst->src[i].file == VGRF &&
 118                 scan_inst->src[i].nr == inst->src[0].nr &&
 119                 scan_inst->src[i].reg_offset == inst->src[0].reg_offset) {
 120                if (scan_inst->opcode != BRW_OPCODE_MOV ||
 121                    !scan_inst->saturate ||
 122                    scan_inst->src[0].abs ||
 123                    scan_inst->src[0].negate ||
 124                    scan_inst->src[0].abs != inst->src[0].abs ||
 125                    scan_inst->src[0].negate != inst->src[0].negate) {
 126                   interfered = true;
 127                   break;
 128                }
 129             }
 130          }
 131
 132          if (interfered)
 133             break;
 134       }
 135    }
 136
 137    return progress;
 138 }
 139
 140 bool
 141 fs_visitor::opt_saturate_propagation()
 142 {
 143    bool progress = false;
 144
 145    calculate_live_intervals();
 146
 147    foreach_block (block, cfg) {
 148       progress = opt_saturate_propagation_local(this, block) || progress;
 149    }
 150
 151    /* Live intervals are still valid. */
 152
 153    return progress;
 154 }