src/glsl/loop_unroll.cpp

   1 /*
   2  * Copyright © 2010 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
  21  * DEALINGS IN THE SOFTWARE.
  22  */
  23
  24 #include "glsl_types.h"
  25 #include "loop_analysis.h"
  26 #include "ir_hierarchical_visitor.h"
  27
  28 class loop_unroll_visitor : public ir_hierarchical_visitor {
  29 public:
  30    loop_unroll_visitor(loop_state *state, unsigned max_iterations)
  31    {
  32       this->state = state;
  33       this->progress = false;
  34       this->max_iterations = max_iterations;
  35    }
  36
  37    virtual ir_visitor_status visit_leave(ir_loop *ir);
  38
  39    loop_state *state;
  40
  41    bool progress;
  42    unsigned max_iterations;
  43 };
  44
  45
  46 ir_visitor_status
  47 loop_unroll_visitor::visit_leave(ir_loop *ir)
  48 {
  49    loop_variable_state *const ls = this->state->get(ir);
  50    int iterations;
  51
  52    /* If we've entered a loop that hasn't been analyzed, something really,
  53     * really bad has happened.
  54     */
  55    if (ls == NULL) {
  56       assert(ls != NULL);
  57       return visit_continue;
  58    }
  59
  60    iterations = ls->max_iterations;
  61
  62    /* Don't try to unroll loops where the number of iterations is not known
  63     * at compile-time.
  64     */
  65    if (iterations < 0)
  66       return visit_continue;
  67
  68    /* Don't try to unroll loops that have zillions of iterations either.
  69     */
  70    if (iterations > max_iterations)
  71       return visit_continue;
  72
  73    if (ls->num_loop_jumps > 1)
  74       return visit_continue;
  75    else if (ls->num_loop_jumps) {
  76       /* recognize loops in the form produced by ir_lower_jumps */
  77       ir_instruction *last_ir =
  78          ((ir_instruction*)ir->body_instructions.get_tail());
  79
  80       assert(last_ir != NULL);
  81
  82       ir_if *last_if = last_ir->as_if();
  83       if (last_if) {
  84          bool continue_from_then_branch;
  85
  86          /* Determine which if-statement branch, if any, ends with a break.
  87           * The branch that did *not* have the break will get a temporary
  88           * continue inserted in each iteration of the loop unroll.
  89           *
  90           * Note that since ls->num_loop_jumps is <= 1, it is impossible for
  91           * both branches to end with a break.
  92           */
  93          ir_instruction *last =
  94             (ir_instruction *) last_if->then_instructions.get_tail();
  95
  96          if (last && last->ir_type == ir_type_loop_jump
  97              && ((ir_loop_jump*) last)->is_break()) {
  98             continue_from_then_branch = false;
  99          } else {
 100             last = (ir_instruction *) last_if->then_instructions.get_tail();
 101
 102             if (last && last->ir_type == ir_type_loop_jump
 103                 && ((ir_loop_jump*) last)->is_break())
 104                continue_from_then_branch = true;
 105             else
 106                /* Bail out if neither if-statement branch ends with a break.
 107                 */
 108                return visit_continue;
 109          }
 110
 111          /* Remove the break from the if-statement.
 112           */
 113          last->remove();
 114
 115          void *const mem_ctx = talloc_parent(ir);
 116          ir_instruction *ir_to_replace = ir;
 117
 118          for (int i = 0; i < iterations; i++) {
 119             exec_list copy_list;
 120
 121             copy_list.make_empty();
 122             clone_ir_list(mem_ctx, &copy_list, &ir->body_instructions);
 123
 124             last_if = ((ir_instruction*)copy_list.get_tail())->as_if();
 125             assert(last_if);
 126
 127             ir_to_replace->insert_before(&copy_list);
 128             ir_to_replace->remove();
 129
 130             /* placeholder that will be removed in the next iteration */
 131             ir_to_replace =
 132                new(mem_ctx) ir_loop_jump(ir_loop_jump::jump_continue);
 133
 134             exec_list *const list = (continue_from_then_branch)
 135                ? &last_if->then_instructions : &last_if->else_instructions;
 136
 137             list->push_tail(ir_to_replace);
 138          }
 139
 140          ir_to_replace->remove();
 141
 142          this->progress = true;
 143          return visit_continue;
 144       } else if (last_ir->ir_type == ir_type_loop_jump
 145                  && ((ir_loop_jump *)last_ir)->is_break()) {
 146          /* If the only loop-jump is a break at the end of the loop, the loop
 147           * will execute exactly once.  Remove the break, set the iteration
 148           * count, and fall through to the normal unroller.
 149           */
 150          last_ir->remove();
 151          iterations = 1;
 152
 153          this->progress = true;
 154       } else
 155          return visit_continue;
 156    }
 157
 158    void *const mem_ctx = talloc_parent(ir);
 159
 160    for (int i = 0; i < iterations; i++) {
 161       exec_list copy_list;
 162
 163       copy_list.make_empty();
 164       clone_ir_list(mem_ctx, &copy_list, &ir->body_instructions);
 165
 166       ir->insert_before(&copy_list);
 167    }
 168
 169    /* The loop has been replaced by the unrolled copies.  Remove the original
 170     * loop from the IR sequence.
 171     */
 172    ir->remove();
 173
 174    this->progress = true;
 175    return visit_continue;
 176 }
 177
 178
 179 bool
 180 unroll_loops(exec_list *instructions, loop_state *ls, unsigned max_iterations)
 181 {
 182    loop_unroll_visitor v(ls, max_iterations);
 183
 184    v.run(instructions);
 185
 186    return v.progress;
 187 }