src/intel/compiler/brw_vec4_cmod_propagation.cpp

   1 /*
   2  * Copyright © 2015 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
  25 /** @file brw_vec4_cmod_propagation.cpp
  26  *
  27  * Really similar to brw_fs_cmod_propagation but adapted to vec4 needs. Check
  28  * brw_fs_cmod_propagation for further details on the rationale behind this
  29  * optimization.
  30  */
  31
  32 #include "brw_vec4.h"
  33 #include "brw_cfg.h"
  34 #include "brw_eu.h"
  35
  36 namespace brw {
  37
  38 static bool
  39 opt_cmod_propagation_local(bblock_t *block)
  40 {
  41    bool progress = false;
  42    int ip = block->end_ip + 1;
  43
  44    foreach_inst_in_block_reverse_safe(vec4_instruction, inst, block) {
  45       ip--;
  46
  47       if ((inst->opcode != BRW_OPCODE_AND &&
  48            inst->opcode != BRW_OPCODE_CMP &&
  49            inst->opcode != BRW_OPCODE_MOV) ||
  50           inst->predicate != BRW_PREDICATE_NONE ||
  51           !inst->dst.is_null() ||
  52           inst->src[0].file != VGRF ||
  53           inst->src[0].abs)
  54          continue;
  55
  56       if (inst->opcode == BRW_OPCODE_AND &&
  57           !(inst->src[1].is_one() &&
  58             inst->conditional_mod == BRW_CONDITIONAL_NZ &&
  59             !inst->src[0].negate))
  60          continue;
  61
  62       if (inst->opcode == BRW_OPCODE_CMP && !inst->src[1].is_zero())
  63          continue;
  64
  65       if (inst->opcode == BRW_OPCODE_MOV &&
  66           inst->conditional_mod != BRW_CONDITIONAL_NZ)
  67          continue;
  68
  69       bool read_flag = false;
  70       foreach_inst_in_block_reverse_starting_from(vec4_instruction, scan_inst, inst) {
  71          if (regions_overlap(inst->src[0], inst->size_read(0),
  72                              scan_inst->dst, scan_inst->size_written)) {
  73             if ((scan_inst->predicate && scan_inst->opcode != BRW_OPCODE_SEL) ||
  74                 scan_inst->dst.offset != inst->src[0].offset ||
  75                 (scan_inst->dst.writemask != WRITEMASK_X &&
  76                  scan_inst->dst.writemask != WRITEMASK_XYZW) ||
  77                 (scan_inst->dst.writemask == WRITEMASK_XYZW &&
  78                  inst->src[0].swizzle != BRW_SWIZZLE_XYZW) ||
  79                 (inst->dst.writemask & ~scan_inst->dst.writemask) != 0 ||
  80                 scan_inst->exec_size != inst->exec_size ||
  81                 scan_inst->group != inst->group) {
  82                break;
  83             }
  84
  85             /* CMP's result is the same regardless of dest type. */
  86             if (inst->conditional_mod == BRW_CONDITIONAL_NZ &&
  87                 scan_inst->opcode == BRW_OPCODE_CMP &&
  88                 (inst->dst.type == BRW_REGISTER_TYPE_D ||
  89                  inst->dst.type == BRW_REGISTER_TYPE_UD)) {
  90                inst->remove(block);
  91                progress = true;
  92                break;
  93             }
  94
  95             /* If the AND wasn't handled by the previous case, it isn't safe
  96              * to remove it.
  97              */
  98             if (inst->opcode == BRW_OPCODE_AND)
  99                break;
 100
 101             /* Comparisons operate differently for ints and floats */
 102             if (scan_inst->dst.type != inst->dst.type &&
 103                 (scan_inst->dst.type == BRW_REGISTER_TYPE_F ||
 104                  inst->dst.type == BRW_REGISTER_TYPE_F))
 105                break;
 106
 107             /* If the instruction generating inst's source also wrote the
 108              * flag, and inst is doing a simple .nz comparison, then inst
 109              * is redundant - the appropriate value is already in the flag
 110              * register.  Delete inst.
 111              */
 112             if (inst->conditional_mod == BRW_CONDITIONAL_NZ &&
 113                 !inst->src[0].negate &&
 114                 scan_inst->writes_flag()) {
 115                inst->remove(block);
 116                progress = true;
 117                break;
 118             }
 119
 120             /* The conditional mod of the CMP/CMPN instructions behaves
 121              * specially because the flag output is not calculated from the
 122              * result of the instruction, but the other way around, which
 123              * means that even if the condmod to propagate and the condmod
 124              * from the CMP instruction are the same they will in general give
 125              * different results because they are evaluated based on different
 126              * inputs.
 127              */
 128             if (scan_inst->opcode == BRW_OPCODE_CMP ||
 129                 scan_inst->opcode == BRW_OPCODE_CMPN)
 130                break;
 131
 132             /* From the Sky Lake PRM Vol. 7 "Assigning Conditional Mods":
 133              *
 134              *    * Note that the [post condition signal] bits generated at
 135              *      the output of a compute are before the .sat.
 136              */
 137             if (scan_inst->saturate)
 138                break;
 139
 140             /* From the Sky Lake PRM, Vol 2a, "Multiply":
 141              *
 142              *    "When multiplying integer data types, if one of the sources
 143              *    is a DW, the resulting full precision data is stored in
 144              *    the accumulator. However, if the destination data type is
 145              *    either W or DW, the low bits of the result are written to
 146              *    the destination register and the remaining high bits are
 147              *    discarded. This results in undefined Overflow and Sign
 148              *    flags. Therefore, conditional modifiers and saturation
 149              *    (.sat) cannot be used in this case.
 150              *
 151              * We just disallow cmod propagation on all integer multiplies.
 152              */
 153             if (!brw_reg_type_is_floating_point(scan_inst->dst.type) &&
 154                 scan_inst->opcode == BRW_OPCODE_MUL)
 155                break;
 156
 157             /* Otherwise, try propagating the conditional. */
 158             enum brw_conditional_mod cond =
 159                inst->src[0].negate ? brw_swap_cmod(inst->conditional_mod)
 160                                    : inst->conditional_mod;
 161
 162             if (scan_inst->can_do_cmod() &&
 163                 ((!read_flag && scan_inst->conditional_mod == BRW_CONDITIONAL_NONE) ||
 164                  scan_inst->conditional_mod == cond)) {
 165                scan_inst->conditional_mod = cond;
 166                inst->remove(block);
 167                progress = true;
 168             }
 169             break;
 170          }
 171
 172          if (scan_inst->writes_flag())
 173             break;
 174
 175          read_flag = read_flag || scan_inst->reads_flag();
 176       }
 177    }
 178
 179    return progress;
 180 }
 181
 182 bool
 183 vec4_visitor::opt_cmod_propagation()
 184 {
 185    bool progress = false;
 186
 187    foreach_block_reverse(block, cfg) {
 188       progress = opt_cmod_propagation_local(block) || progress;
 189    }
 190
 191    if (progress)
 192       invalidate_live_intervals();
 193
 194    return progress;
 195 }
 196
 197 } /* namespace brw */