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
28using namespace ir_builder;
29
30namespace {
31
32class vector_insert_visitor : public ir_rvalue_visitor {
33public:
34   vector_insert_visitor(bool lower_nonconstant_index)
35      : progress(false), lower_nonconstant_index(lower_nonconstant_index),
36        remove_assignment(false)
37   {
38      factory.instructions = &factory_instructions;
39   }
40
41   virtual ~vector_insert_visitor()
42   {
43      assert(factory_instructions.is_empty());
44   }
45
46   virtual void handle_rvalue(ir_rvalue **rv);
47   virtual ir_visitor_status visit_leave(ir_assignment *expr);
48
49   ir_factory factory;
50   exec_list factory_instructions;
51   bool progress;
52   bool lower_nonconstant_index;
53   bool remove_assignment;
54};
55
56} /* anonymous namespace */
57
58void
59vector_insert_visitor::handle_rvalue(ir_rvalue **rv)
60{
61   if (*rv == NULL || (*rv)->ir_type != ir_type_expression)
62      return;
63
64   ir_expression *const expr = (ir_expression *) *rv;
65
66   if (likely(expr->operation != ir_triop_vector_insert))
67      return;
68
69   factory.mem_ctx = ralloc_parent(expr);
70
71   ir_constant *const idx =
72      expr->operands[2]->constant_expression_value(factory.mem_ctx);
73   if (idx != NULL) {
74      unsigned index = idx->value.u[0];
75
76      if (index >= expr->operands[0]->type->vector_elements) {
77         /* Section 5.11 (Out-of-Bounds Accesses) of the GLSL 4.60 spec says:
78          *
79          *  In the subsections described above for array, vector, matrix and
80          *  structure accesses, any out-of-bounds access produced undefined
81          *  behavior.... Out-of-bounds writes may be discarded or overwrite
82          *  other variables of the active program.
83          */
84         this->remove_assignment = true;
85         this->progress = true;
86         return;
87      }
88
89      /* Replace (vector_insert (vec) (scalar) (index)) with a dereference of
90       * a new temporary.  The new temporary gets assigned as
91       *
92       *     t = vec
93       *     t.mask = scalar
94       *
95       * where mask is the component selected by index.
96       */
97      ir_variable *const temp =
98         factory.make_temp(expr->operands[0]->type, "vec_tmp");
99
100      const int mask = 1 << idx->value.i[0];
101
102      factory.emit(assign(temp, expr->operands[0]));
103      factory.emit(assign(temp, expr->operands[1], mask));
104
105      this->progress = true;
106      *rv = new(factory.mem_ctx) ir_dereference_variable(temp);
107   } else if (this->lower_nonconstant_index) {
108      /* Replace (vector_insert (vec) (scalar) (index)) with a dereference of
109       * a new temporary.  The new temporary gets assigned as
110       *
111       *     t = vec
112       *     if (index == 0)
113       *         t.x = scalar
114       *     if (index == 1)
115       *         t.y = scalar
116       *     if (index == 2)
117       *         t.z = scalar
118       *     if (index == 3)
119       *         t.w = scalar
120       */
121      ir_variable *const temp =
122         factory.make_temp(expr->operands[0]->type, "vec_tmp");
123
124      ir_variable *const src_temp =
125         factory.make_temp(expr->operands[1]->type, "src_temp");
126
127      factory.emit(assign(temp, expr->operands[0]));
128      factory.emit(assign(src_temp, expr->operands[1]));
129
130      assert(expr->operands[2]->type == glsl_type::int_type ||
131             expr->operands[2]->type == glsl_type::uint_type);
132
133      for (unsigned i = 0; i < expr->type->vector_elements; i++) {
134         ir_constant *const cmp_index =
135            ir_constant::zero(factory.mem_ctx, expr->operands[2]->type);
136         cmp_index->value.u[0] = i;
137
138         ir_variable *const cmp_result =
139            factory.make_temp(glsl_type::bool_type, "index_condition");
140
141         factory.emit(assign(cmp_result,
142                             equal(expr->operands[2]->clone(factory.mem_ctx,
143                                                            NULL),
144                                   cmp_index)));
145
146         factory.emit(if_tree(cmp_result,
147                              assign(temp, src_temp, WRITEMASK_X << i)));
148      }
149
150      this->progress = true;
151      *rv = new(factory.mem_ctx) ir_dereference_variable(temp);
152   }
153
154   base_ir->insert_before(factory.instructions);
155}
156
157ir_visitor_status
158vector_insert_visitor::visit_leave(ir_assignment *ir)
159{
160   ir_rvalue_visitor::visit_leave(ir);
161
162   if (this->remove_assignment) {
163      ir->remove();
164      this->remove_assignment = false;
165   }
166
167   return visit_continue;
168}
169
170bool
171lower_vector_insert(exec_list *instructions, bool lower_nonconstant_index)
172{
173   vector_insert_visitor v(lower_nonconstant_index);
174
175   visit_list_elements(&v, instructions);
176
177   return v.progress;
178}
179