src/gallium/drivers/freedreno/ir3/ir3_cp.c

   1 /* -*- mode: C; c-file-style: "k&r"; tab-width 4; indent-tabs-mode: t; -*- */
   2
   3 /*
   4  * Copyright (C) 2014 Rob Clark <robclark@freedesktop.org>
   5  *
   6  * Permission is hereby granted, free of charge, to any person obtaining a
   7  * copy of this software and associated documentation files (the "Software"),
   8  * to deal in the Software without restriction, including without limitation
   9  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
  10  * and/or sell copies of the Software, and to permit persons to whom the
  11  * Software is furnished to do so, subject to the following conditions:
  12  *
  13  * The above copyright notice and this permission notice (including the next
  14  * paragraph) shall be included in all copies or substantial portions of the
  15  * Software.
  16  *
  17  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  18  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  19  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
  20  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  21  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
  22  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  23  * SOFTWARE.
  24  *
  25  * Authors:
  26  *    Rob Clark <robclark@freedesktop.org>
  27  */
  28
  29 #include "freedreno_util.h"
  30
  31 #include "ir3.h"
  32
  33 /*
  34  * Copy Propagate:
  35  *
  36  * We could eventually drop this, if the front-end did not insert any
  37  * mov's..  For now keeping it as a separate pass since that is less
  38  * painful than updating the existing frontend.  It is expected that
  39  * with an eventual new NIR based frontend that we won't need this.
  40  */
  41
  42 static void block_cp(struct ir3_block *block);
  43 static struct ir3_instruction * instr_cp(struct ir3_instruction *instr);
  44
  45 static bool is_eligible_mov(struct ir3_instruction *instr)
  46 {
  47         if ((instr->category == 1) &&
  48                         (instr->cat1.src_type == instr->cat1.dst_type)) {
  49                 struct ir3_register *dst = instr->regs[0];
  50                 struct ir3_register *src = instr->regs[1];
  51                 struct ir3_instruction *src_instr = ssa(src);
  52                 if (dst->flags & IR3_REG_ADDR)
  53                         return false;
  54                 /* TODO: propagate abs/neg modifiers if possible */
  55                 if (src->flags & (IR3_REG_ABS | IR3_REG_NEGATE | IR3_REG_RELATIV))
  56                         return false;
  57                 if (!src_instr)
  58                         return false;
  59                 /* TODO: remove this hack: */
  60                 if (is_meta(src_instr) && (src_instr->opc == OPC_META_FO))
  61                         return false;
  62                 return true;
  63         }
  64         return false;
  65 }
  66
  67 static void walk_children(struct ir3_instruction *instr)
  68 {
  69         unsigned i;
  70
  71         /* walk down the graph from each src: */
  72         for (i = 1; i < instr->regs_count; i++) {
  73                 struct ir3_register *src = instr->regs[i];
  74                 if (src->flags & IR3_REG_SSA)
  75                         src->instr = instr_cp(src->instr);
  76         }
  77 }
  78
  79
  80 static struct ir3_instruction *
  81 instr_cp(struct ir3_instruction *instr)
  82 {
  83         /* stay within the block.. don't try to operate across
  84          * basic block boundaries or we'll have problems when
  85          * dealing with multiple basic blocks:
  86          */
  87         if (is_meta(instr) && (instr->opc == OPC_META_INPUT))
  88                 return instr;
  89
  90         if (is_eligible_mov(instr)) {
  91                 struct ir3_instruction *src_instr = ssa(instr->regs[1]);
  92                 return instr_cp(src_instr);
  93         }
  94
  95         /* Check termination condition before walking children (rather
  96          * than before checking eligible-mov).  A mov instruction may
  97          * appear as ssa-src for multiple other instructions, and we
  98          * want to consider it for removal for each, rather than just
  99          * the first one.  (But regardless of how many places it shows
 100          * up as a src, we only need to recursively walk the children
 101          * once.)
 102          */
 103         if (!ir3_instr_check_mark(instr))
 104                 walk_children(instr);
 105
 106         return instr;
 107 }
 108
 109 static void block_cp(struct ir3_block *block)
 110 {
 111         unsigned i;
 112
 113         for (i = 0; i < block->noutputs; i++) {
 114                 if (block->outputs[i]) {
 115                         struct ir3_instruction *out =
 116                                         instr_cp(block->outputs[i]);
 117
 118                         block->outputs[i] = out;
 119                 }
 120         }
 121 }
 122
 123 void ir3_block_cp(struct ir3_block *block)
 124 {
 125         ir3_clear_mark(block->shader);
 126         block_cp(block);
 127 }