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_vec_index_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_vec_index_to_swizzle(ir_rvalue *ir)
  63 {
  64    ir_dereference_array *deref = ir->as_dereference_array();
  65    ir_constant *ir_constant;
  66
  67    if (!deref)
  68       return ir;
  69
  70    if (deref->array->type->is_matrix() || deref->array->type->is_array())
  71       return ir;
  72
  73    assert(deref->array_index->type->base_type == GLSL_TYPE_INT);
  74    ir_constant = deref->array_index->constant_expression_value();
  75    if (!ir_constant)
  76       return ir;
  77
  78    void *ctx = ralloc_parent(ir);
  79    this->progress = true;
  80
  81    /* Page 40 of the GLSL 1.20 spec says:
  82     *
  83     *     "When indexing with non-constant expressions, behavior is undefined
  84     *     if the index is negative, or greater than or equal to the size of
  85     *     the vector."
  86     *
  87     * The quoted spec text mentions non-constant expressions, but this code
  88     * operates on constants.  These constants are the result of non-constant
  89     * expressions that have been optimized to constants.  The common case here
  90     * is a loop counter from an unrolled loop that is used to index a vector.
  91     *
  92     * The ir_swizzle constructor gets angry if the index is negative or too
  93     * large.  For simplicity sake, just clamp the index to [0, size-1].
  94     */
  95    const int i = MIN2(MAX2(ir_constant->value.i[0], 0),
  96                       (deref->array->type->vector_elements - 1));
  97
  98    return new(ctx) ir_swizzle(deref->array, i, 0, 0, 0, 1);
  99 }
 100
 101 ir_visitor_status
 102 ir_vec_index_to_swizzle_visitor::visit_enter(ir_expression *ir)
 103 {
 104    unsigned int i;
 105
 106    for (i = 0; i < ir->get_num_operands(); i++) {
 107       ir->operands[i] = convert_vec_index_to_swizzle(ir->operands[i]);
 108    }
 109
 110    return visit_continue;
 111 }
 112
 113 ir_visitor_status
 114 ir_vec_index_to_swizzle_visitor::visit_enter(ir_swizzle *ir)
 115 {
 116    /* Can't be hit from normal GLSL, since you can't swizzle a scalar (which
 117     * the result of indexing a vector is.  But maybe at some point we'll end up
 118     * using swizzling of scalars for vector construction.
 119     */
 120    ir->val = convert_vec_index_to_swizzle(ir->val);
 121
 122    return visit_continue;
 123 }
 124
 125 ir_visitor_status
 126 ir_vec_index_to_swizzle_visitor::visit_enter(ir_assignment *ir)
 127 {
 128    ir->set_lhs(convert_vec_index_to_swizzle(ir->lhs));
 129    ir->rhs = convert_vec_index_to_swizzle(ir->rhs);
 130
 131    return visit_continue;
 132 }
 133
 134 ir_visitor_status
 135 ir_vec_index_to_swizzle_visitor::visit_enter(ir_call *ir)
 136 {
 137    foreach_iter(exec_list_iterator, iter, *ir) {
 138       ir_rvalue *param = (ir_rvalue *)iter.get();
 139       ir_rvalue *new_param = convert_vec_index_to_swizzle(param);
 140
 141       if (new_param != param) {
 142          param->replace_with(new_param);
 143       }
 144    }
 145
 146    return visit_continue;
 147 }
 148
 149 ir_visitor_status
 150 ir_vec_index_to_swizzle_visitor::visit_enter(ir_return *ir)
 151 {
 152    if (ir->value) {
 153       ir->value = convert_vec_index_to_swizzle(ir->value);
 154    }
 155
 156    return visit_continue;
 157 }
 158
 159 ir_visitor_status
 160 ir_vec_index_to_swizzle_visitor::visit_enter(ir_if *ir)
 161 {
 162    ir->condition = convert_vec_index_to_swizzle(ir->condition);
 163
 164    return visit_continue;
 165 }
 166
 167 bool
 168 do_vec_index_to_swizzle(exec_list *instructions)
 169 {
 170    ir_vec_index_to_swizzle_visitor v;
 171
 172    v.run(instructions);
 173
 174    return v.progress;
 175 }