src/glsl/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
  30 /*
  31  * Implements a small peephole optimization that looks for
  32  *
  33  * if (cond) {
  34  *    <empty>
  35  * } else {
  36  *    <empty>
  37  * }
  38  * phi
  39  * ...
  40  * phi
  41  *
  42  * and replaces it with a series of selects.  It can also handle the case
  43  * where, instead of being empty, the if may contain some move operations
  44  * whose only use is one of the following phi nodes.  This happens all the
  45  * time when the SSA form comes from a conditional assignment with a
  46  * swizzle.
  47  */
  48
  49 struct peephole_select_state {
  50    void *mem_ctx;
  51    bool progress;
  52 };
  53
  54 static bool
  55 block_check_for_allowed_instrs(nir_block *block)
  56 {
  57    nir_foreach_instr(block, instr) {
  58       switch (instr->type) {
  59       case nir_instr_type_intrinsic: {
  60          nir_intrinsic_instr *intrin = nir_instr_as_intrinsic(instr);
  61
  62          switch (intrin->intrinsic) {
  63          case nir_intrinsic_load_var:
  64             switch (intrin->variables[0]->var->data.mode) {
  65             case nir_var_shader_in:
  66             case nir_var_uniform:
  67                break;
  68
  69             default:
  70                return false;
  71             }
  72             break;
  73
  74          default:
  75             return false;
  76          }
  77
  78          break;
  79       }
  80
  81       case nir_instr_type_load_const:
  82          break;
  83
  84       case nir_instr_type_alu: {
  85          nir_alu_instr *mov = nir_instr_as_alu(instr);
  86          switch (mov->op) {
  87          case nir_op_fmov:
  88          case nir_op_imov:
  89          case nir_op_fneg:
  90          case nir_op_ineg:
  91          case nir_op_fabs:
  92          case nir_op_iabs:
  93          case nir_op_vec2:
  94          case nir_op_vec3:
  95          case nir_op_vec4:
  96             /* It must be a move-like operation. */
  97             break;
  98          default:
  99             return false;
 100          }
 101
 102          /* Can't handle saturate */
 103          if (mov->dest.saturate)
 104             return false;
 105
 106          /* It must be SSA */
 107          if (!mov->dest.dest.is_ssa)
 108             return false;
 109
 110          /* It cannot have any if-uses */
 111          if (!list_empty(&mov->dest.dest.ssa.if_uses))
 112             return false;
 113
 114          /* The only uses of this definition must be phi's in the successor */
 115          nir_foreach_use(&mov->dest.dest.ssa, use) {
 116             if (use->parent_instr->type != nir_instr_type_phi ||
 117                 use->parent_instr->block != block->successors[0])
 118                return false;
 119          }
 120          break;
 121       }
 122
 123       default:
 124          return false;
 125       }
 126    }
 127
 128    return true;
 129 }
 130
 131 static bool
 132 nir_opt_peephole_select_block(nir_block *block, void *void_state)
 133 {
 134    struct peephole_select_state *state = void_state;
 135
 136    /* If the block is empty, then it certainly doesn't have any phi nodes,
 137     * so we can skip it.  This also ensures that we do an early skip on the
 138     * end block of the function which isn't actually attached to the CFG.
 139     */
 140    if (exec_list_is_empty(&block->instr_list))
 141       return true;
 142
 143    if (nir_cf_node_is_first(&block->cf_node))
 144       return true;
 145
 146    nir_cf_node *prev_node = nir_cf_node_prev(&block->cf_node);
 147    if (prev_node->type != nir_cf_node_if)
 148       return true;
 149
 150    nir_if *if_stmt = nir_cf_node_as_if(prev_node);
 151    nir_cf_node *then_node = nir_if_first_then_node(if_stmt);
 152    nir_cf_node *else_node = nir_if_first_else_node(if_stmt);
 153
 154    /* We can only have one block in each side ... */
 155    if (nir_if_last_then_node(if_stmt) != then_node ||
 156        nir_if_last_else_node(if_stmt) != else_node)
 157       return true;
 158
 159    nir_block *then_block = nir_cf_node_as_block(then_node);
 160    nir_block *else_block = nir_cf_node_as_block(else_node);
 161
 162    /* ... and those blocks must only contain "allowed" instructions. */
 163    if (!block_check_for_allowed_instrs(then_block) ||
 164        !block_check_for_allowed_instrs(else_block))
 165       return true;
 166
 167    /* At this point, we know that the previous CFG node is an if-then
 168     * statement containing only moves to phi nodes in this block.  We can
 169     * just remove that entire CF node and replace all of the phi nodes with
 170     * selects.
 171     */
 172
 173    nir_block *prev_block = nir_cf_node_as_block(nir_cf_node_prev(prev_node));
 174    assert(prev_block->cf_node.type == nir_cf_node_block);
 175
 176    /* First, we move the remaining instructions from the blocks to the
 177     * block before.  We have already guaranteed that this is safe by
 178     * calling block_check_for_allowed_instrs()
 179     */
 180    nir_foreach_instr_safe(then_block, instr) {
 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(else_block, instr) {
 187       exec_node_remove(&instr->node);
 188       instr->block = prev_block;
 189       exec_list_push_tail(&prev_block->instr_list, &instr->node);
 190    }
 191
 192    nir_foreach_instr_safe(block, instr) {
 193       if (instr->type != nir_instr_type_phi)
 194          break;
 195
 196       nir_phi_instr *phi = nir_instr_as_phi(instr);
 197       nir_alu_instr *sel = nir_alu_instr_create(state->mem_ctx, nir_op_bcsel);
 198       nir_src_copy(&sel->src[0].src, &if_stmt->condition, state->mem_ctx);
 199       /* Splat the condition to all channels */
 200       memset(sel->src[0].swizzle, 0, sizeof sel->src[0].swizzle);
 201
 202       assert(exec_list_length(&phi->srcs) == 2);
 203       nir_foreach_phi_src(phi, src) {
 204          assert(src->pred == then_block || src->pred == else_block);
 205          assert(src->src.is_ssa);
 206
 207          unsigned idx = src->pred == then_block ? 1 : 2;
 208          nir_src_copy(&sel->src[idx].src, &src->src, state->mem_ctx);
 209       }
 210
 211       nir_ssa_dest_init(&sel->instr, &sel->dest.dest,
 212                         phi->dest.ssa.num_components, phi->dest.ssa.name);
 213       sel->dest.write_mask = (1 << phi->dest.ssa.num_components) - 1;
 214
 215       nir_ssa_def_rewrite_uses(&phi->dest.ssa,
 216                                nir_src_for_ssa(&sel->dest.dest.ssa),
 217                                state->mem_ctx);
 218
 219       nir_instr_insert_before(&phi->instr, &sel->instr);
 220       nir_instr_remove(&phi->instr);
 221    }
 222
 223    nir_cf_node_remove(&if_stmt->cf_node);
 224    state->progress = true;
 225
 226    return true;
 227 }
 228
 229 static bool
 230 nir_opt_peephole_select_impl(nir_function_impl *impl)
 231 {
 232    struct peephole_select_state state;
 233
 234    state.mem_ctx = ralloc_parent(impl);
 235    state.progress = false;
 236
 237    nir_foreach_block(impl, nir_opt_peephole_select_block, &state);
 238
 239    if (state.progress)
 240       nir_metadata_preserve(impl, nir_metadata_none);
 241
 242    return state.progress;
 243 }
 244
 245 bool
 246 nir_opt_peephole_select(nir_shader *shader)
 247 {
 248    bool progress = false;
 249
 250    nir_foreach_overload(shader, overload) {
 251       if (overload->impl)
 252          progress |= nir_opt_peephole_select_impl(overload->impl);
 253    }
 254
 255    return progress;
 256 }