src/freedreno/ir3/ir3_group.c

   1 /*
   2  * Copyright (C) 2014 Rob Clark <robclark@freedesktop.org>
   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 FROM,
  20  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  21  * SOFTWARE.
  22  *
  23  * Authors:
  24  *    Rob Clark <robclark@freedesktop.org>
  25  */
  26
  27 #include "ir3.h"
  28
  29 /*
  30  * Find/group instruction neighbors:
  31  */
  32
  33 static void
  34 insert_mov(struct ir3_instruction *collect, int idx)
  35 {
  36         struct ir3_instruction *src = ssa(collect->regs[idx+1]);
  37         struct ir3_instruction *mov = ir3_MOV(src->block, src, TYPE_F32);
  38         collect->regs[idx+1]->instr = mov;
  39
  40         /* if collect and src are in the same block, move the inserted mov
  41          * to just before the collect to avoid a use-before-def.  Otherwise
  42          * it should be safe to leave at the end of the block it is in:
  43          */
  44         if (src->block == collect->block) {
  45                 list_delinit(&mov->node);
  46                 list_addtail(&mov->node, &collect->node);
  47         }
  48 }
  49
  50 /* verify that cur != instr, but cur is also not in instr's neighbor-list: */
  51 static bool
  52 in_neighbor_list(struct ir3_instruction *instr, struct ir3_instruction *cur, int pos)
  53 {
  54         int idx = 0;
  55
  56         if (!instr)
  57                 return false;
  58
  59         if (instr == cur)
  60                 return true;
  61
  62         for (instr = ir3_neighbor_first(instr); instr; instr = instr->cp.right)
  63                 if ((idx++ != pos) && (instr == cur))
  64                         return true;
  65
  66         return false;
  67 }
  68
  69 static void
  70 group_collect(struct ir3_instruction *collect)
  71 {
  72         struct ir3_register **regs = &collect->regs[1];
  73         unsigned n = collect->regs_count - 1;
  74
  75         /* first pass, figure out what has conflicts and needs a mov
  76          * inserted.  Do this up front, before starting to setup
  77          * left/right neighbor pointers.  Trying to do it in a single
  78          * pass could result in a situation where we can't even setup
  79          * the mov's right neighbor ptr if the next instr also needs
  80          * a mov.
  81          */
  82 restart:
  83         for (unsigned i = 0; i < n; i++) {
  84                 struct ir3_instruction *instr = ssa(regs[i]);
  85                 if (instr) {
  86                         struct ir3_instruction *left = (i > 0) ? ssa(regs[i - 1]) : NULL;
  87                         struct ir3_instruction *right = (i < (n-1)) ? ssa(regs[i + 1]) : NULL;
  88                         bool conflict;
  89
  90                         /* check for left/right neighbor conflicts: */
  91                         conflict = conflicts(instr->cp.left, left) ||
  92                                 conflicts(instr->cp.right, right);
  93
  94                         /* Mixing array elements and higher register classes
  95                          * (ie. groups) doesn't really work out in RA.  See:
  96                          *
  97                          * https://trello.com/c/DqeDkeVf/156-bug-with-stk-70frag
  98                          */
  99                         if (instr->regs[0]->flags & IR3_REG_ARRAY)
 100                                 conflict = true;
 101
 102                         /* we also can't have an instr twice in the group: */
 103                         for (unsigned j = i + 1; (j < n) && !conflict; j++)
 104                                 if (in_neighbor_list(ssa(regs[j]), instr, i))
 105                                         conflict = true;
 106
 107                         if (conflict) {
 108                                 insert_mov(collect, i);
 109                                 /* inserting the mov may have caused a conflict
 110                                  * against the previous:
 111                                  */
 112                                 goto restart;
 113                         }
 114                 }
 115         }
 116
 117         /* second pass, now that we've inserted mov's, fixup left/right
 118          * neighbors.  This is guaranteed to succeed, since by definition
 119          * the newly inserted mov's cannot conflict with anything.
 120          */
 121         for (unsigned i = 0; i < n; i++) {
 122                 struct ir3_instruction *instr = ssa(regs[i]);
 123                 if (instr) {
 124                         struct ir3_instruction *left = (i > 0) ? ssa(regs[i - 1]) : NULL;
 125                         struct ir3_instruction *right = (i < (n-1)) ? ssa(regs[i + 1]) : NULL;
 126
 127                         debug_assert(!conflicts(instr->cp.left, left));
 128                         if (left) {
 129                                 instr->cp.left_cnt++;
 130                                 instr->cp.left = left;
 131                         }
 132
 133                         debug_assert(!conflicts(instr->cp.right, right));
 134                         if (right) {
 135                                 instr->cp.right_cnt++;
 136                                 instr->cp.right = right;
 137                         }
 138                 }
 139         }
 140 }
 141
 142 static bool
 143 instr_find_neighbors(struct ir3_instruction *instr)
 144 {
 145         bool progress = false;
 146
 147         if (ir3_instr_check_mark(instr))
 148                 return false;
 149
 150         if (instr->opc == OPC_META_COLLECT) {
 151                 group_collect(instr);
 152                 progress = true;
 153         }
 154
 155         foreach_ssa_src (src, instr)
 156                 progress |= instr_find_neighbors(src);
 157
 158         return progress;
 159 }
 160
 161 static bool
 162 find_neighbors(struct ir3 *ir)
 163 {
 164         bool progress = false;
 165         unsigned i;
 166
 167         foreach_output (out, ir)
 168                 progress |= instr_find_neighbors(out);
 169
 170         foreach_block (block, &ir->block_list) {
 171                 for (i = 0; i < block->keeps_count; i++) {
 172                         struct ir3_instruction *instr = block->keeps[i];
 173                         progress |= instr_find_neighbors(instr);
 174                 }
 175
 176                 /* We also need to account for if-condition: */
 177                 if (block->condition)
 178                         progress |= instr_find_neighbors(block->condition);
 179         }
 180
 181         return progress;
 182 }
 183
 184 bool
 185 ir3_group(struct ir3 *ir)
 186 {
 187         ir3_clear_mark(ir);
 188         return find_neighbors(ir);
 189 }