2 * Copyright © 2014 Broadcom
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 * Implements CSE for QIR without control flow.
29 * For each operation that writes a destination (and isn't just a MOV), put it
30 * in the hash table of all instructions that do so. When faced with another
31 * one, look it up in the hash table by its opcode and operands. If there's
32 * an entry in the table, then just reuse the entry's destination as the
33 * source of a MOV instead of reproducing the computation. That MOV will then
34 * get cleaned up by copy propagation.
39 #include "util/hash_table.h"
40 #include "util/ralloc.h"
48 * If the instruction depends on the flags, how many SFs have been
49 * seen before this instruction, or if it depends on r4, how many r4
50 * writes have been seen.
52 uint32_t implicit_arg_update_count
;
56 inst_key_equals(const void *a
, const void *b
)
58 const struct inst_key
*key_a
= a
;
59 const struct inst_key
*key_b
= b
;
61 return memcmp(key_a
, key_b
, sizeof(*key_a
)) == 0;
65 vc4_find_cse(struct vc4_compile
*c
, struct hash_table
*ht
,
66 struct qinst
*inst
, uint32_t sf_count
,
69 if (inst
->dst
.file
!= QFILE_TEMP
||
70 inst
->op
== QOP_MOV
||
71 qir_get_op_nsrc(inst
->op
) > 4) {
76 memset(&key
, 0, sizeof(key
));
78 memcpy(key
.src
, inst
->src
,
79 qir_get_op_nsrc(inst
->op
) * sizeof(key
.src
[0]));
80 if (qir_depends_on_flags(inst
))
81 key
.implicit_arg_update_count
= sf_count
;
82 if (qir_reads_r4(inst
))
83 key
.implicit_arg_update_count
= r4_count
;
85 uint32_t hash
= _mesa_hash_data(&key
, sizeof(key
));
86 struct hash_entry
*entry
=
87 _mesa_hash_table_search_pre_hashed(ht
, hash
, &key
);
91 fprintf(stderr
, "CSE found match:\n");
93 fprintf(stderr
, " Original inst: ");
94 qir_dump_inst(c
, entry
->data
);
95 fprintf(stderr
, "\n");
97 fprintf(stderr
, " Our inst: ");
98 qir_dump_inst(c
, inst
);
99 fprintf(stderr
, "\n");
105 struct inst_key
*alloc_key
= ralloc(ht
, struct inst_key
);
108 memcpy(alloc_key
, &key
, sizeof(*alloc_key
));
109 _mesa_hash_table_insert_pre_hashed(ht
, hash
, alloc_key
, inst
);
112 fprintf(stderr
, "Added to CSE HT: ");
113 qir_dump_inst(c
, inst
);
114 fprintf(stderr
, "\n");
121 qir_opt_cse(struct vc4_compile
*c
)
123 bool progress
= false;
124 uint32_t sf_count
= 0, r4_count
= 0;
126 struct hash_table
*ht
= _mesa_hash_table_create(NULL
, NULL
,
131 list_for_each_entry(struct qinst
, inst
, &c
->instructions
, link
) {
132 if (qir_has_side_effects(c
, inst
) ||
133 qir_has_side_effect_reads(c
, inst
) ||
134 inst
->op
== QOP_TLB_COLOR_READ
) {
141 struct qinst
*cse
= vc4_find_cse(c
, ht
, inst
,
144 inst
->src
[0] = cse
->dst
;
145 for (int i
= 1; i
< qir_get_op_nsrc(inst
->op
);
147 inst
->src
[i
] = c
->undef
;
152 fprintf(stderr
, " Turned into: ");
153 qir_dump_inst(c
, inst
);
154 fprintf(stderr
, "\n");
159 if (qir_writes_r4(inst
))