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,
  38                 (collect->regs[idx+1]->flags & IR3_REG_HALF) ? TYPE_U16 : TYPE_U32);
  39
  40         collect->regs[idx+1]->instr = mov;
  41
  42         /* if collect and src are in the same block, move the inserted mov
  43          * to just before the collect to avoid a use-before-def.  Otherwise
  44          * it should be safe to leave at the end of the block it is in:
  45          */
  46         if (src->block == collect->block) {
  47                 ir3_instr_move_before(mov, collect);
  48         }
  49 }
  50
  51 /* verify that cur != instr, but cur is also not in instr's neighbor-list: */
  52 static bool
  53 in_neighbor_list(struct ir3_instruction *instr, struct ir3_instruction *cur, int pos)
  54 {
  55         int idx = 0;
  56
  57         if (!instr)
  58                 return false;
  59
  60         if (instr == cur)
  61                 return true;
  62
  63         for (instr = ir3_neighbor_first(instr); instr; instr = instr->cp.right)
  64                 if ((idx++ != pos) && (instr == cur))
  65                         return true;
  66
  67         return false;
  68 }
  69
  70 static void
  71 group_collect(struct ir3_instruction *collect)
  72 {
  73         struct ir3_register **regs = &collect->regs[1];
  74         unsigned n = collect->regs_count - 1;
  75
  76         /* first pass, figure out what has conflicts and needs a mov
  77          * inserted.  Do this up front, before starting to setup
  78          * left/right neighbor pointers.  Trying to do it in a single
  79          * pass could result in a situation where we can't even setup
  80          * the mov's right neighbor ptr if the next instr also needs
  81          * a mov.
  82          */
  83 restart:
  84         for (unsigned i = 0; i < n; i++) {
  85                 struct ir3_instruction *instr = ssa(regs[i]);
  86                 if (instr) {
  87                         struct ir3_instruction *left = (i > 0) ? ssa(regs[i - 1]) : NULL;
  88                         struct ir3_instruction *right = (i < (n-1)) ? ssa(regs[i + 1]) : NULL;
  89                         bool conflict;
  90
  91                         /* check for left/right neighbor conflicts: */
  92                         conflict = conflicts(instr->cp.left, left) ||
  93                                 conflicts(instr->cp.right, right);
  94
  95                         /* Mixing array elements and higher register classes
  96                          * (ie. groups) doesn't really work out in RA.  See:
  97                          *
  98                          * https://trello.com/c/DqeDkeVf/156-bug-with-stk-70frag
  99                          */
 100                         if (instr->regs[0]->flags & IR3_REG_ARRAY)
 101                                 conflict = true;
 102
 103                         /* we also can't have an instr twice in the group: */
 104                         for (unsigned j = i + 1; (j < n) && !conflict; j++)
 105                                 if (in_neighbor_list(ssa(regs[j]), instr, i))
 106                                         conflict = true;
 107
 108                         if (conflict) {
 109                                 insert_mov(collect, i);
 110                                 /* inserting the mov may have caused a conflict
 111                                  * against the previous:
 112                                  */
 113                                 goto restart;
 114                         }
 115                 }
 116         }
 117
 118         /* second pass, now that we've inserted mov's, fixup left/right
 119          * neighbors.  This is guaranteed to succeed, since by definition
 120          * the newly inserted mov's cannot conflict with anything.
 121          */
 122         for (unsigned i = 0; i < n; i++) {
 123                 struct ir3_instruction *instr = ssa(regs[i]);
 124                 if (instr) {
 125                         struct ir3_instruction *left = (i > 0) ? ssa(regs[i - 1]) : NULL;
 126                         struct ir3_instruction *right = (i < (n-1)) ? ssa(regs[i + 1]) : NULL;
 127
 128                         debug_assert(!conflicts(instr->cp.left, left));
 129                         if (left) {
 130                                 instr->cp.left_cnt++;
 131                                 instr->cp.left = left;
 132                         }
 133
 134                         debug_assert(!conflicts(instr->cp.right, right));
 135                         if (right) {
 136                                 instr->cp.right_cnt++;
 137                                 instr->cp.right = right;
 138                         }
 139                 }
 140         }
 141 }
 142
 143 static bool
 144 instr_find_neighbors(struct ir3_instruction *instr)
 145 {
 146         bool progress = false;
 147
 148         if (ir3_instr_check_mark(instr))
 149                 return false;
 150
 151         if (instr->opc == OPC_META_COLLECT) {
 152                 group_collect(instr);
 153                 progress = true;
 154         }
 155
 156         foreach_ssa_src (src, instr)
 157                 progress |= instr_find_neighbors(src);
 158
 159         return progress;
 160 }
 161
 162 static bool
 163 find_neighbors(struct ir3 *ir)
 164 {
 165         bool progress = false;
 166         unsigned i;
 167
 168         foreach_output (out, ir)
 169                 progress |= instr_find_neighbors(out);
 170
 171         foreach_block (block, &ir->block_list) {
 172                 for (i = 0; i < block->keeps_count; i++) {
 173                         struct ir3_instruction *instr = block->keeps[i];
 174                         progress |= instr_find_neighbors(instr);
 175                 }
 176
 177                 /* We also need to account for if-condition: */
 178                 if (block->condition)
 179                         progress |= instr_find_neighbors(block->condition);
 180         }
 181
 182         return progress;
 183 }
 184
 185 bool
 186 ir3_group(struct ir3 *ir)
 187 {
 188         ir3_clear_mark(ir);
 189         return find_neighbors(ir);
 190 }