src/glsl/lower_vector_insert.cpp

   1 /*
   2  * Copyright © 2013 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 #include "ir.h"
  24 #include "ir_builder.h"
  25 #include "ir_rvalue_visitor.h"
  26 #include "ir_optimization.h"
  27
  28 using namespace ir_builder;
  29
  30 class vector_insert_visitor : public ir_rvalue_visitor {
  31 public:
  32    vector_insert_visitor(bool lower_nonconstant_index)
  33       : progress(false), lower_nonconstant_index(lower_nonconstant_index)
  34    {
  35       factory.instructions = &factory_instructions;
  36    }
  37
  38    virtual ~vector_insert_visitor()
  39    {
  40       assert(factory_instructions.is_empty());
  41    }
  42
  43    virtual void handle_rvalue(ir_rvalue **rv);
  44
  45    ir_factory factory;
  46    exec_list factory_instructions;
  47    bool progress;
  48    bool lower_nonconstant_index;
  49 };
  50
  51
  52 void
  53 vector_insert_visitor::handle_rvalue(ir_rvalue **rv)
  54 {
  55    if (*rv == NULL || (*rv)->ir_type != ir_type_expression)
  56       return;
  57
  58    ir_expression *const expr = (ir_expression *) *rv;
  59
  60    if (likely(expr->operation != ir_triop_vector_insert))
  61       return;
  62
  63    factory.mem_ctx = ralloc_parent(expr);
  64
  65    ir_constant *const idx = expr->operands[2]->constant_expression_value();
  66    if (idx != NULL) {
  67       /* Replace (vector_insert (vec) (scalar) (index)) with a dereference of
  68        * a new temporary.  The new temporary gets assigned as
  69        *
  70        *     t = vec
  71        *     t.mask = scalar
  72        *
  73        * where mask is the component selected by index.
  74        */
  75       ir_variable *const temp =
  76          factory.make_temp(expr->operands[0]->type, "vec_tmp");
  77
  78       const int mask = 1 << idx->value.i[0];
  79
  80       factory.emit(assign(temp, expr->operands[0]));
  81       factory.emit(assign(temp, expr->operands[1], mask));
  82
  83       this->progress = true;
  84       *rv = new(factory.mem_ctx) ir_dereference_variable(temp);
  85    } else if (this->lower_nonconstant_index) {
  86       /* Replace (vector_insert (vec) (scalar) (index)) with a dereference of
  87        * a new temporary.  The new temporary gets assigned as
  88        *
  89        *     t = vec
  90        *     if (index == 0)
  91        *         t.x = scalar
  92        *     if (index == 1)
  93        *         t.y = scalar
  94        *     if (index == 2)
  95        *         t.z = scalar
  96        *     if (index == 3)
  97        *         t.w = scalar
  98        */
  99       ir_variable *const temp =
 100          factory.make_temp(expr->operands[0]->type, "vec_tmp");
 101
 102       ir_variable *const src_temp =
 103          factory.make_temp(expr->operands[1]->type, "src_temp");
 104
 105       factory.emit(assign(temp, expr->operands[0]));
 106       factory.emit(assign(src_temp, expr->operands[1]));
 107
 108       for (unsigned i = 0; i < expr->type->vector_elements; i++) {
 109          ir_constant *const cmp_index =
 110             new(factory.mem_ctx) ir_constant(int(i));
 111
 112          ir_variable *const cmp_result =
 113             factory.make_temp(glsl_type::bool_type, "index_condition");
 114
 115          factory.emit(assign(cmp_result,
 116                              equal(expr->operands[2]->clone(factory.mem_ctx,
 117                                                             NULL),
 118                                    cmp_index)));
 119
 120          factory.emit(if_tree(cmp_result,
 121                               assign(temp, src_temp, WRITEMASK_X << i)));
 122       }
 123
 124       this->progress = true;
 125       *rv = new(factory.mem_ctx) ir_dereference_variable(temp);
 126    }
 127
 128    base_ir->insert_before(factory.instructions);
 129 }
 130
 131 bool
 132 lower_vector_insert(exec_list *instructions, bool lower_nonconstant_index)
 133 {
 134    vector_insert_visitor v(lower_nonconstant_index);
 135
 136    visit_list_elements(&v, instructions);
 137
 138    return v.progress;
 139 }