src/gallium/drivers/lima/ir/lima_nir_duplicate_intrinsic.c

   1 /*
   2  * Copyright (c) 2020 Lima Project
   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 "nir.h"
  25 #include "nir_builder.h"
  26 #include "lima_ir.h"
  27
  28 static bool
  29 lima_nir_duplicate_intrinsic(nir_builder *b, nir_intrinsic_instr *itr,
  30                              nir_intrinsic_op op)
  31 {
  32    nir_intrinsic_instr *last_dupl = NULL;
  33    nir_instr *last_parent_instr = NULL;
  34
  35    nir_foreach_use_safe(use_src, &itr->dest.ssa) {
  36       nir_intrinsic_instr *dupl;
  37
  38       if (last_parent_instr != use_src->parent_instr) {
  39          /* if ssa use, clone for the target block */
  40          b->cursor = nir_before_instr(use_src->parent_instr);
  41          dupl = nir_intrinsic_instr_create(b->shader, op);
  42          dupl->num_components = itr->num_components;
  43          memcpy(dupl->const_index, itr->const_index, sizeof(itr->const_index));
  44          dupl->src[0].is_ssa = itr->src[0].is_ssa;
  45          if (itr->src[0].is_ssa)
  46             dupl->src[0].ssa = itr->src[0].ssa;
  47          else
  48             dupl->src[0].reg = itr->src[0].reg;
  49
  50          nir_ssa_dest_init(&dupl->instr, &dupl->dest,
  51                dupl->num_components, itr->dest.ssa.bit_size,
  52                itr->dest.ssa.name);
  53
  54          dupl->instr.pass_flags = 1;
  55          nir_builder_instr_insert(b, &dupl->instr);
  56       }
  57       else {
  58          dupl = last_dupl;
  59       }
  60
  61       nir_instr_rewrite_src(use_src->parent_instr, use_src, nir_src_for_ssa(&dupl->dest.ssa));
  62       last_parent_instr = use_src->parent_instr;
  63       last_dupl = dupl;
  64    }
  65
  66    last_dupl = NULL;
  67    last_parent_instr = NULL;
  68
  69    nir_foreach_if_use_safe(use_src, &itr->dest.ssa) {
  70       nir_intrinsic_instr *dupl;
  71
  72       if (last_parent_instr != use_src->parent_instr) {
  73          /* if 'if use', clone where it is */
  74          b->cursor = nir_before_instr(&itr->instr);
  75          dupl = nir_intrinsic_instr_create(b->shader, op);
  76          dupl->num_components = itr->num_components;
  77          memcpy(dupl->const_index, itr->const_index, sizeof(itr->const_index));
  78          dupl->src[0].is_ssa = itr->src[0].is_ssa;
  79          if (itr->src[0].is_ssa)
  80             dupl->src[0].ssa = itr->src[0].ssa;
  81          else
  82             dupl->src[0].reg = itr->src[0].reg;
  83
  84          nir_ssa_dest_init(&dupl->instr, &dupl->dest,
  85                dupl->num_components, itr->dest.ssa.bit_size,
  86                itr->dest.ssa.name);
  87
  88          dupl->instr.pass_flags = 1;
  89          nir_builder_instr_insert(b, &dupl->instr);
  90       }
  91       else {
  92          dupl = last_dupl;
  93       }
  94
  95       nir_if_rewrite_condition(use_src->parent_if, nir_src_for_ssa(&dupl->dest.ssa));
  96       last_parent_instr = use_src->parent_instr;
  97       last_dupl = dupl;
  98    }
  99
 100    nir_instr_remove(&itr->instr);
 101    return true;
 102 }
 103
 104 static void
 105 lima_nir_duplicate_intrinsic_impl(nir_shader *shader, nir_function_impl *impl,
 106                                   nir_intrinsic_op op)
 107 {
 108    nir_builder builder;
 109    nir_builder_init(&builder, impl);
 110
 111    nir_foreach_block(block, impl) {
 112       nir_foreach_instr(instr, block) {
 113          instr->pass_flags = 0;
 114       }
 115
 116       nir_foreach_instr_safe(instr, block) {
 117          if (instr->type != nir_instr_type_intrinsic)
 118             continue;
 119
 120          nir_intrinsic_instr *itr = nir_instr_as_intrinsic(instr);
 121
 122          if (itr->intrinsic != op)
 123             continue;
 124
 125          if (itr->instr.pass_flags)
 126             continue;
 127
 128          if (!itr->dest.is_ssa)
 129             continue;
 130
 131          lima_nir_duplicate_intrinsic(&builder, itr, op);
 132       }
 133    }
 134
 135    nir_metadata_preserve(impl, nir_metadata_block_index |
 136                                nir_metadata_dominance);
 137 }
 138
 139 /* Duplicate load uniforms for every user.
 140  * Helps by utilizing the load uniform instruction slots that would
 141  * otherwise stay empty, and reduces register pressure. */
 142 void
 143 lima_nir_duplicate_load_uniforms(nir_shader *shader)
 144 {
 145    nir_foreach_function(function, shader) {
 146       if (function->impl) {
 147          lima_nir_duplicate_intrinsic_impl(shader, function->impl, nir_intrinsic_load_uniform);
 148       }
 149    }
 150 }
 151
 152 /* Duplicate load inputs for every user.
 153  * Helps by utilizing the load input instruction slots that would
 154  * otherwise stay empty, and reduces register pressure. */
 155 void
 156 lima_nir_duplicate_load_inputs(nir_shader *shader)
 157 {
 158    nir_foreach_function(function, shader) {
 159       if (function->impl) {
 160          lima_nir_duplicate_intrinsic_impl(shader, function->impl, nir_intrinsic_load_input);
 161       }
 162    }
 163 }