src/glsl/lower_vec_index_to_swizzle.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 /**
  25  * \file lower_vec_index_to_swizzle.cpp
  26  *
  27  * Turns constant indexing into vector types to swizzles.  This will
  28  * let other swizzle-aware optimization passes catch these constructs,
  29  * and codegen backends not have to worry about this case.
  30  */
  31
  32 #include "ir.h"
  33 #include "ir_visitor.h"
  34 #include "ir_optimization.h"
  35 #include "glsl_types.h"
  36 #include "main/macros.h"
  37
  38 /**
  39  * Visitor class for replacing expressions with ir_constant values.
  40  */
  41
  42 class ir_vec_index_to_swizzle_visitor : public ir_hierarchical_visitor {
  43 public:
  44    ir_vec_index_to_swizzle_visitor()
  45    {
  46       progress = false;
  47    }
  48
  49    ir_rvalue *convert_vector_extract_to_swizzle(ir_rvalue *val);
  50
  51    virtual ir_visitor_status visit_enter(ir_expression *);
  52    virtual ir_visitor_status visit_enter(ir_swizzle *);
  53    virtual ir_visitor_status visit_enter(ir_assignment *);
  54    virtual ir_visitor_status visit_enter(ir_return *);
  55    virtual ir_visitor_status visit_enter(ir_call *);
  56    virtual ir_visitor_status visit_enter(ir_if *);
  57
  58    bool progress;
  59 };
  60
  61 ir_rvalue *
  62 ir_vec_index_to_swizzle_visitor::convert_vector_extract_to_swizzle(ir_rvalue *ir)
  63 {
  64    ir_expression *const expr = ir->as_expression();
  65    if (expr == NULL || expr->operation != ir_binop_vector_extract)
  66       return ir;
  67
  68    ir_constant *const idx = expr->operands[1]->constant_expression_value();
  69    if (idx == NULL)
  70       return ir;
  71
  72    void *ctx = ralloc_parent(ir);
  73    this->progress = true;
  74
  75    /* Page 40 of the GLSL 1.20 spec says:
  76     *
  77     *     "When indexing with non-constant expressions, behavior is undefined
  78     *     if the index is negative, or greater than or equal to the size of
  79     *     the vector."
  80     *
  81     * The quoted spec text mentions non-constant expressions, but this code
  82     * operates on constants.  These constants are the result of non-constant
  83     * expressions that have been optimized to constants.  The common case here
  84     * is a loop counter from an unrolled loop that is used to index a vector.
  85     *
  86     * The ir_swizzle constructor gets angry if the index is negative or too
  87     * large.  For simplicity sake, just clamp the index to [0, size-1].
  88     */
  89    const int i = CLAMP(idx->value.i[0], 0,
  90                        (int) expr->operands[0]->type->vector_elements - 1);
  91
  92    return new(ctx) ir_swizzle(expr->operands[0], i, 0, 0, 0, 1);
  93 }
  94
  95 ir_visitor_status
  96 ir_vec_index_to_swizzle_visitor::visit_enter(ir_expression *ir)
  97 {
  98    unsigned int i;
  99
 100    for (i = 0; i < ir->get_num_operands(); i++) {
 101       ir->operands[i] = convert_vector_extract_to_swizzle(ir->operands[i]);
 102    }
 103
 104    return visit_continue;
 105 }
 106
 107 ir_visitor_status
 108 ir_vec_index_to_swizzle_visitor::visit_enter(ir_swizzle *ir)
 109 {
 110    /* Can't be hit from normal GLSL, since you can't swizzle a scalar (which
 111     * the result of indexing a vector is.  But maybe at some point we'll end up
 112     * using swizzling of scalars for vector construction.
 113     */
 114    ir->val = convert_vector_extract_to_swizzle(ir->val);
 115
 116    return visit_continue;
 117 }
 118
 119 ir_visitor_status
 120 ir_vec_index_to_swizzle_visitor::visit_enter(ir_assignment *ir)
 121 {
 122    ir->rhs = convert_vector_extract_to_swizzle(ir->rhs);
 123
 124    return visit_continue;
 125 }
 126
 127 ir_visitor_status
 128 ir_vec_index_to_swizzle_visitor::visit_enter(ir_call *ir)
 129 {
 130    foreach_iter(exec_list_iterator, iter, *ir) {
 131       ir_rvalue *param = (ir_rvalue *)iter.get();
 132       ir_rvalue *new_param = convert_vector_extract_to_swizzle(param);
 133
 134       if (new_param != param) {
 135          param->replace_with(new_param);
 136       }
 137    }
 138
 139    return visit_continue;
 140 }
 141
 142 ir_visitor_status
 143 ir_vec_index_to_swizzle_visitor::visit_enter(ir_return *ir)
 144 {
 145    if (ir->value) {
 146       ir->value = convert_vector_extract_to_swizzle(ir->value);
 147    }
 148
 149    return visit_continue;
 150 }
 151
 152 ir_visitor_status
 153 ir_vec_index_to_swizzle_visitor::visit_enter(ir_if *ir)
 154 {
 155    ir->condition = convert_vector_extract_to_swizzle(ir->condition);
 156
 157    return visit_continue;
 158 }
 159
 160 bool
 161 do_vec_index_to_swizzle(exec_list *instructions)
 162 {
 163    ir_vec_index_to_swizzle_visitor v;
 164
 165    v.run(instructions);
 166
 167    return v.progress;
 168 }