2 * Copyright © 2012 Intel Corporation
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:
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
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 DEALINGS
27 /** @file brw_fs_cse.cpp
29 * Support for local common subexpression elimination.
31 * See Muchnik's Advanced Compiler Design and Implementation, section
36 struct aeb_entry
: public exec_node
{
37 /** The instruction that generates the expression value. */
40 /** The temporary where the value is stored. */
46 is_expression(const fs_inst
*const inst
)
48 switch (inst
->opcode
) {
69 case FS_OPCODE_CINTERP
:
70 case FS_OPCODE_LINTERP
:
78 operands_match(fs_reg
*xs
, fs_reg
*ys
)
80 return xs
[0].equals(ys
[0]) && xs
[1].equals(ys
[1]) && xs
[2].equals(ys
[2]);
84 fs_visitor::opt_cse_local(bblock_t
*block
, exec_list
*aeb
)
86 bool progress
= false;
88 void *mem_ctx
= ralloc_context(this->mem_ctx
);
90 for (fs_inst
*inst
= (fs_inst
*)block
->start
;
91 inst
!= block
->end
->next
;
92 inst
= (fs_inst
*) inst
->next
) {
94 /* Skip some cases. */
95 if (is_expression(inst
) && !inst
->predicate
&& inst
->mlen
== 0 &&
96 !inst
->force_uncompressed
&& !inst
->force_sechalf
&&
97 !inst
->conditional_mod
)
102 foreach_list(entry_node
, aeb
) {
103 entry
= (aeb_entry
*) entry_node
;
105 /* Match current instruction's expression against those in AEB. */
106 if (inst
->opcode
== entry
->generator
->opcode
&&
107 inst
->saturate
== entry
->generator
->saturate
&&
108 operands_match(entry
->generator
->src
, inst
->src
)) {
117 /* Our first sighting of this expression. Create an entry. */
118 aeb_entry
*entry
= ralloc(mem_ctx
, aeb_entry
);
119 entry
->tmp
= reg_undef
;
120 entry
->generator
= inst
;
121 aeb
->push_tail(entry
);
123 /* This is at least our second sighting of this expression.
124 * If we don't have a temporary already, make one.
126 bool no_existing_temp
= entry
->tmp
.file
== BAD_FILE
;
127 if (no_existing_temp
) {
128 entry
->tmp
= fs_reg(this, glsl_type::float_type
);
129 entry
->tmp
.type
= inst
->dst
.type
;
131 fs_inst
*copy
= new(ralloc_parent(inst
))
132 fs_inst(BRW_OPCODE_MOV
, entry
->generator
->dst
, entry
->tmp
);
133 entry
->generator
->insert_after(copy
);
134 entry
->generator
->dst
= entry
->tmp
;
138 fs_inst
*copy
= new(ralloc_parent(inst
))
139 fs_inst(BRW_OPCODE_MOV
, inst
->dst
, entry
->tmp
);
140 inst
->replace_with(copy
);
142 /* Appending an instruction may have changed our bblock end. */
143 if (inst
== block
->end
) {
147 /* Continue iteration with copy->next */
152 /* Kill all AEB entries that use the destination. */
153 foreach_list_safe(entry_node
, aeb
) {
154 aeb_entry
*entry
= (aeb_entry
*)entry_node
;
156 for (int i
= 0; i
< 3; i
++) {
157 if (inst
->overwrites_reg(entry
->generator
->src
[i
])) {
166 ralloc_free(mem_ctx
);
169 this->live_intervals_valid
= false;
175 fs_visitor::opt_cse()
177 bool progress
= false;
181 for (int b
= 0; b
< cfg
.num_blocks
; b
++) {
182 bblock_t
*block
= cfg
.blocks
[b
];
185 progress
= opt_cse_local(block
, &aeb
) || progress
;