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 QOP_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 hash_table
*ht
, struct qinst
*inst
, uint32_t sf_count
,
68 if (inst
->dst
.file
!= QFILE_TEMP
||
69 inst
->op
== QOP_MOV
||
70 qir_get_op_nsrc(inst
->op
) > 4) {
75 memset(&key
, 0, sizeof(key
));
77 memcpy(key
.src
, inst
->src
,
78 qir_get_op_nsrc(inst
->op
) * sizeof(key
.src
[0]));
79 if (qir_depends_on_flags(inst
))
80 key
.implicit_arg_update_count
= sf_count
;
81 if (qir_reads_r4(inst
))
82 key
.implicit_arg_update_count
= r4_count
;
84 uint32_t hash
= _mesa_hash_data(&key
, sizeof(key
));
85 struct hash_entry
*entry
=
86 _mesa_hash_table_search(ht
, hash
, &key
);
90 fprintf(stderr
, "CSE found match:\n");
92 fprintf(stderr
, " Original inst: ");
93 qir_dump_inst(entry
->data
);
94 fprintf(stderr
, "\n");
96 fprintf(stderr
, " Our inst: ");
98 fprintf(stderr
, "\n");
104 struct inst_key
*alloc_key
= ralloc(ht
, struct inst_key
);
107 memcpy(alloc_key
, &key
, sizeof(*alloc_key
));
108 _mesa_hash_table_insert(ht
, hash
, alloc_key
, inst
);
111 fprintf(stderr
, "Added to CSE HT: ");
113 fprintf(stderr
, "\n");
120 qir_opt_cse(struct qcompile
*c
)
122 bool progress
= false;
123 struct simple_node
*node
, *t
;
124 struct qinst
*last_sf
= NULL
;
125 uint32_t sf_count
= 0, r4_count
= 0;
128 struct hash_table
*ht
= _mesa_hash_table_create(NULL
, inst_key_equals
);
132 foreach_s(node
, t
, &c
->instructions
) {
133 struct qinst
*inst
= (struct qinst
*)node
;
135 if (qir_has_side_effects(inst
)) {
136 if (inst
->op
== QOP_TLB_DISCARD_SETUP
)
141 if (inst
->op
== QOP_SF
) {
143 qir_reg_equals(last_sf
->src
[0], inst
->src
[0])) {
146 "Removing redundant SF: ");
148 fprintf(stderr
, "\n");
150 remove_from_list(&inst
->link
);
158 struct qinst
*cse
= vc4_find_cse(ht
, inst
,
161 inst
->src
[0] = cse
->dst
;
162 for (int i
= 1; i
< qir_get_op_nsrc(inst
->op
);
164 inst
->src
[i
] = c
->undef
;
169 fprintf(stderr
, " Turned into: ");
171 fprintf(stderr
, "\n");
176 if (qir_reads_r4(inst
))