src/compiler/nir/nir_opt_peephole_select.c

   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  * Authors:
  24  *    Jason Ekstrand (jason@jlekstrand.net)
  25  *
  26  */
  27
  28 #include "nir.h"
  29 #include "nir_control_flow.h"
  30
  31 /*
  32  * Implements a small peephole optimization that looks for
  33  *
  34  * if (cond) {
  35  *    <empty>
  36  * } else {
  37  *    <empty>
  38  * }
  39  * phi
  40  * ...
  41  * phi
  42  *
  43  * and replaces it with a series of selects.  It can also handle the case
  44  * where, instead of being empty, the if may contain some move operations
  45  * whose only use is one of the following phi nodes.  This happens all the
  46  * time when the SSA form comes from a conditional assignment with a
  47  * swizzle.
  48  */
  49
  50 static bool
  51 block_check_for_allowed_instrs(nir_block *block)
  52 {
  53    nir_foreach_instr(instr, block) {
  54       switch (instr->type) {
  55       case nir_instr_type_intrinsic: {
  56          nir_intrinsic_instr *intrin = nir_instr_as_intrinsic(instr);
  57
  58          switch (intrin->intrinsic) {
  59          case nir_intrinsic_load_var:
  60             switch (intrin->variables[0]->var->data.mode) {
  61             case nir_var_shader_in:
  62             case nir_var_uniform:
  63                break;
  64
  65             default:
  66                return false;
  67             }
  68             break;
  69
  70          default:
  71             return false;
  72          }
  73
  74          break;
  75       }
  76
  77       case nir_instr_type_load_const:
  78          break;
  79
  80       case nir_instr_type_alu: {
  81          nir_alu_instr *mov = nir_instr_as_alu(instr);
  82          switch (mov->op) {
  83          case nir_op_fmov:
  84          case nir_op_imov:
  85          case nir_op_fneg:
  86          case nir_op_ineg:
  87          case nir_op_fabs:
  88          case nir_op_iabs:
  89          case nir_op_vec2:
  90          case nir_op_vec3:
  91          case nir_op_vec4:
  92             /* It must be a move-like operation. */
  93             break;
  94          default:
  95             return false;
  96          }
  97
  98          /* Can't handle saturate */
  99          if (mov->dest.saturate)
 100             return false;
 101
 102          /* It must be SSA */
 103          if (!mov->dest.dest.is_ssa)
 104             return false;
 105
 106          /* It cannot have any if-uses */
 107          if (!list_empty(&mov->dest.dest.ssa.if_uses))
 108             return false;
 109
 110          /* The only uses of this definition must be phi's in the successor */
 111          nir_foreach_use(&mov->dest.dest.ssa, use) {
 112             if (use->parent_instr->type != nir_instr_type_phi ||
 113                 use->parent_instr->block != block->successors[0])
 114                return false;
 115          }
 116          break;
 117       }
 118
 119       default:
 120          return false;
 121       }
 122    }
 123
 124    return true;
 125 }
 126
 127 static bool
 128 nir_opt_peephole_select_block(nir_block *block, void *mem_ctx)
 129 {
 130    /* If the block is empty, then it certainly doesn't have any phi nodes,
 131     * so we can skip it.  This also ensures that we do an early skip on the
 132     * end block of the function which isn't actually attached to the CFG.
 133     */
 134    if (exec_list_is_empty(&block->instr_list))
 135       return false;
 136
 137    if (nir_cf_node_is_first(&block->cf_node))
 138       return false;
 139
 140    nir_cf_node *prev_node = nir_cf_node_prev(&block->cf_node);
 141    if (prev_node->type != nir_cf_node_if)
 142       return false;
 143
 144    nir_if *if_stmt = nir_cf_node_as_if(prev_node);
 145    nir_cf_node *then_node = nir_if_first_then_node(if_stmt);
 146    nir_cf_node *else_node = nir_if_first_else_node(if_stmt);
 147
 148    /* We can only have one block in each side ... */
 149    if (nir_if_last_then_node(if_stmt) != then_node ||
 150        nir_if_last_else_node(if_stmt) != else_node)
 151       return false;
 152
 153    nir_block *then_block = nir_cf_node_as_block(then_node);
 154    nir_block *else_block = nir_cf_node_as_block(else_node);
 155
 156    /* ... and those blocks must only contain "allowed" instructions. */
 157    if (!block_check_for_allowed_instrs(then_block) ||
 158        !block_check_for_allowed_instrs(else_block))
 159       return false;
 160
 161    /* At this point, we know that the previous CFG node is an if-then
 162     * statement containing only moves to phi nodes in this block.  We can
 163     * just remove that entire CF node and replace all of the phi nodes with
 164     * selects.
 165     */
 166
 167    nir_block *prev_block = nir_cf_node_as_block(nir_cf_node_prev(prev_node));
 168    assert(prev_block->cf_node.type == nir_cf_node_block);
 169
 170    /* First, we move the remaining instructions from the blocks to the
 171     * block before.  We have already guaranteed that this is safe by
 172     * calling block_check_for_allowed_instrs()
 173     */
 174    nir_foreach_instr_safe(instr, then_block) {
 175       exec_node_remove(&instr->node);
 176       instr->block = prev_block;
 177       exec_list_push_tail(&prev_block->instr_list, &instr->node);
 178    }
 179
 180    nir_foreach_instr_safe(instr, else_block) {
 181       exec_node_remove(&instr->node);
 182       instr->block = prev_block;
 183       exec_list_push_tail(&prev_block->instr_list, &instr->node);
 184    }
 185
 186    nir_foreach_instr_safe(instr, block) {
 187       if (instr->type != nir_instr_type_phi)
 188          break;
 189
 190       nir_phi_instr *phi = nir_instr_as_phi(instr);
 191       nir_alu_instr *sel = nir_alu_instr_create(mem_ctx, nir_op_bcsel);
 192       nir_src_copy(&sel->src[0].src, &if_stmt->condition, sel);
 193       /* Splat the condition to all channels */
 194       memset(sel->src[0].swizzle, 0, sizeof sel->src[0].swizzle);
 195
 196       assert(exec_list_length(&phi->srcs) == 2);
 197       nir_foreach_phi_src(phi, src) {
 198          assert(src->pred == then_block || src->pred == else_block);
 199          assert(src->src.is_ssa);
 200
 201          unsigned idx = src->pred == then_block ? 1 : 2;
 202          nir_src_copy(&sel->src[idx].src, &src->src, sel);
 203       }
 204
 205       nir_ssa_dest_init(&sel->instr, &sel->dest.dest,
 206                         phi->dest.ssa.num_components,
 207                         phi->dest.ssa.bit_size, phi->dest.ssa.name);
 208       sel->dest.write_mask = (1 << phi->dest.ssa.num_components) - 1;
 209
 210       nir_ssa_def_rewrite_uses(&phi->dest.ssa,
 211                                nir_src_for_ssa(&sel->dest.dest.ssa));
 212
 213       nir_instr_insert_before(&phi->instr, &sel->instr);
 214       nir_instr_remove(&phi->instr);
 215    }
 216
 217    nir_cf_node_remove(&if_stmt->cf_node);
 218    return true;
 219 }
 220
 221 static bool
 222 nir_opt_peephole_select_impl(nir_function_impl *impl)
 223 {
 224    void *mem_ctx = ralloc_parent(impl);
 225    bool progress = false;
 226
 227    nir_foreach_block_safe(block, impl) {
 228       progress |= nir_opt_peephole_select_block(block, mem_ctx);
 229    }
 230
 231    if (progress)
 232       nir_metadata_preserve(impl, nir_metadata_none);
 233
 234    return progress;
 235 }
 236
 237 bool
 238 nir_opt_peephole_select(nir_shader *shader)
 239 {
 240    bool progress = false;
 241
 242    nir_foreach_function(shader, function) {
 243       if (function->impl)
 244          progress |= nir_opt_peephole_select_impl(function->impl);
 245    }
 246
 247    return progress;
 248 }