src/intel/compiler/brw_fs_cmod_propagation.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_fs.h"
  25 #include "brw_cfg.h"
  26 #include "brw_eu.h"
  27
  28 /** @file brw_fs_cmod_propagation.cpp
  29  *
  30  * Implements a pass that propagates the conditional modifier from a CMP x 0.0
  31  * instruction into the instruction that generated x. For instance, in this
  32  * sequence
  33  *
  34  *    add(8)          g70<1>F    g69<8,8,1>F    4096F
  35  *    cmp.ge.f0(8)    null       g70<8,8,1>F    0F
  36  *
  37  * we can do the comparison as part of the ADD instruction directly:
  38  *
  39  *    add.ge.f0(8)    g70<1>F    g69<8,8,1>F    4096F
  40  *
  41  * If there had been a use of the flag register and another CMP using g70
  42  *
  43  *    add.ge.f0(8)    g70<1>F    g69<8,8,1>F    4096F
  44  *    (+f0) sel(8)    g71<F>     g72<8,8,1>F    g73<8,8,1>F
  45  *    cmp.ge.f0(8)    null       g70<8,8,1>F    0F
  46  *
  47  * we can recognize that the CMP is generating the flag value that already
  48  * exists and therefore remove the instruction.
  49  */
  50
  51 static bool
  52 opt_cmod_propagation_local(const gen_device_info *devinfo, bblock_t *block)
  53 {
  54    bool progress = false;
  55    int ip = block->end_ip + 1;
  56
  57    foreach_inst_in_block_reverse_safe(fs_inst, inst, block) {
  58       ip--;
  59
  60       if ((inst->opcode != BRW_OPCODE_AND &&
  61            inst->opcode != BRW_OPCODE_CMP &&
  62            inst->opcode != BRW_OPCODE_MOV) ||
  63           inst->predicate != BRW_PREDICATE_NONE ||
  64           !inst->dst.is_null() ||
  65           (inst->src[0].file != VGRF && inst->src[0].file != ATTR &&
  66            inst->src[0].file != UNIFORM) ||
  67           inst->src[0].abs)
  68          continue;
  69
  70       /* Only an AND.NZ can be propagated.  Many AND.Z instructions are
  71        * generated (for ir_unop_not in fs_visitor::emit_bool_to_cond_code).
  72        * Propagating those would require inverting the condition on the CMP.
  73        * This changes both the flag value and the register destination of the
  74        * CMP.  That result may be used elsewhere, so we can't change its value
  75        * on a whim.
  76        */
  77       if (inst->opcode == BRW_OPCODE_AND &&
  78           !(inst->src[1].is_one() &&
  79             inst->conditional_mod == BRW_CONDITIONAL_NZ &&
  80             !inst->src[0].negate))
  81          continue;
  82
  83       if (inst->opcode == BRW_OPCODE_CMP && !inst->src[1].is_zero())
  84          continue;
  85
  86       if (inst->opcode == BRW_OPCODE_MOV &&
  87           inst->conditional_mod != BRW_CONDITIONAL_NZ)
  88          continue;
  89
  90       bool read_flag = false;
  91       foreach_inst_in_block_reverse_starting_from(fs_inst, scan_inst, inst) {
  92          if (regions_overlap(scan_inst->dst, scan_inst->size_written,
  93                              inst->src[0], inst->size_read(0))) {
  94             if (scan_inst->is_partial_write() ||
  95                 scan_inst->dst.offset != inst->src[0].offset ||
  96                 scan_inst->exec_size != inst->exec_size)
  97                break;
  98
  99             /* CMP's result is the same regardless of dest type. */
 100             if (inst->conditional_mod == BRW_CONDITIONAL_NZ &&
 101                 scan_inst->opcode == BRW_OPCODE_CMP &&
 102                 (inst->dst.type == BRW_REGISTER_TYPE_D ||
 103                  inst->dst.type == BRW_REGISTER_TYPE_UD)) {
 104                inst->remove(block);
 105                progress = true;
 106                break;
 107             }
 108
 109             /* If the AND wasn't handled by the previous case, it isn't safe
 110              * to remove it.
 111              */
 112             if (inst->opcode == BRW_OPCODE_AND)
 113                break;
 114
 115             /* Comparisons operate differently for ints and floats */
 116             if (scan_inst->dst.type != inst->dst.type &&
 117                 (scan_inst->dst.type == BRW_REGISTER_TYPE_F ||
 118                  inst->dst.type == BRW_REGISTER_TYPE_F))
 119                break;
 120
 121             /* If the instruction generating inst's source also wrote the
 122              * flag, and inst is doing a simple .nz comparison, then inst
 123              * is redundant - the appropriate value is already in the flag
 124              * register.  Delete inst.
 125              */
 126             if (inst->conditional_mod == BRW_CONDITIONAL_NZ &&
 127                 !inst->src[0].negate &&
 128                 scan_inst->flags_written()) {
 129                inst->remove(block);
 130                progress = true;
 131                break;
 132             }
 133
 134             /* The conditional mod of the CMP/CMPN instructions behaves
 135              * specially because the flag output is not calculated from the
 136              * result of the instruction, but the other way around, which
 137              * means that even if the condmod to propagate and the condmod
 138              * from the CMP instruction are the same they will in general give
 139              * different results because they are evaluated based on different
 140              * inputs.
 141              */
 142             if (scan_inst->opcode == BRW_OPCODE_CMP ||
 143                 scan_inst->opcode == BRW_OPCODE_CMPN)
 144                break;
 145
 146             /* From the Sky Lake PRM Vol. 7 "Assigning Conditional Mods":
 147              *
 148              *    * Note that the [post condition signal] bits generated at
 149              *      the output of a compute are before the .sat.
 150              */
 151             if (scan_inst->saturate)
 152                break;
 153
 154             /* From the Sky Lake PRM, Vol 2a, "Multiply":
 155              *
 156              *    "When multiplying integer data types, if one of the sources
 157              *     is a DW, the resulting full precision data is stored in
 158              *     the accumulator. However, if the destination data type is
 159              *     either W or DW, the low bits of the result are written to
 160              *     the destination register and the remaining high bits are
 161              *     discarded. This results in undefined Overflow and Sign
 162              *     flags. Therefore, conditional modifiers and saturation
 163              *     (.sat) cannot be used in this case."
 164              *
 165              * We just disallow cmod propagation on all integer multiplies.
 166              */
 167             if (!brw_reg_type_is_floating_point(scan_inst->dst.type) &&
 168                 scan_inst->opcode == BRW_OPCODE_MUL)
 169                break;
 170
 171             /* Otherwise, try propagating the conditional. */
 172             enum brw_conditional_mod cond =
 173                inst->src[0].negate ? brw_swap_cmod(inst->conditional_mod)
 174                                    : inst->conditional_mod;
 175
 176             if (scan_inst->can_do_cmod() &&
 177                 ((!read_flag && scan_inst->conditional_mod == BRW_CONDITIONAL_NONE) ||
 178                  scan_inst->conditional_mod == cond)) {
 179                scan_inst->conditional_mod = cond;
 180                inst->remove(block);
 181                progress = true;
 182             }
 183             break;
 184          }
 185
 186          if (scan_inst->flags_written())
 187             break;
 188
 189          read_flag = read_flag || scan_inst->flags_read(devinfo);
 190       }
 191    }
 192
 193    return progress;
 194 }
 195
 196 bool
 197 fs_visitor::opt_cmod_propagation()
 198 {
 199    bool progress = false;
 200
 201    foreach_block_reverse(block, cfg) {
 202       progress = opt_cmod_propagation_local(devinfo, block) || progress;
 203    }
 204
 205    if (progress)
 206       invalidate_live_intervals();
 207
 208    return progress;
 209 }