1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
|
/*
* Copyright © 2014 Intel Corporation
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice (including the next
* paragraph) shall be included in all copies or substantial portions of the
* Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
* IN THE SOFTWARE.
*
* Authors:
* Jason Ekstrand (jason@jlekstrand.net)
*
*/
#ifndef _NIR_WORKLIST_
#define _NIR_WORKLIST_
#include "nir.h"
#include "util/set.h"
#include "util/u_vector.h"
#ifdef __cplusplus
extern "C" {
#endif
/** Represents a double-ended queue of unique blocks
*
* The worklist datastructure guarantees that eacy block is in the queue at
* most once. Pushing a block onto either end of the queue is a no-op if
* the block is already in the queue. In order for this to work, the
* caller must ensure that the blocks are properly indexed.
*/
typedef struct {
/* The total size of the worklist */
unsigned size;
/* The number of blocks currently in the worklist */
unsigned count;
/* The offset in the array of blocks at which the list starts */
unsigned start;
/* A bitset of all of the blocks currently present in the worklist */
BITSET_WORD *blocks_present;
/* The actual worklist */
nir_block **blocks;
} nir_block_worklist;
void nir_block_worklist_init(nir_block_worklist *w, unsigned num_blocks,
void *mem_ctx);
void nir_block_worklist_fini(nir_block_worklist *w);
void nir_block_worklist_add_all(nir_block_worklist *w, nir_function_impl *impl);
static inline bool
nir_block_worklist_is_empty(const nir_block_worklist *w)
{
return w->count == 0;
}
void nir_block_worklist_push_head(nir_block_worklist *w, nir_block *block);
nir_block *nir_block_worklist_peek_head(const nir_block_worklist *w);
nir_block *nir_block_worklist_pop_head(nir_block_worklist *w);
void nir_block_worklist_push_tail(nir_block_worklist *w, nir_block *block);
nir_block *nir_block_worklist_peek_tail(const nir_block_worklist *w);
nir_block *nir_block_worklist_pop_tail(nir_block_worklist *w);
/*
* This worklist implementation, in contrast to the block worklist, does not
* have unique entries, meaning a nir_instr can be inserted more than once
* into the worklist. It uses u_vector to keep the overhead and memory
* footprint at a minimum.
*
* Making it unique by using a set was tested, but for the single usecase
* (nir_opt_dce) it did not improve speed. There we check the pass_flag bit
* and abort immediately if there's nothing to do, so the added overhead of
* the set was higher than just processing the few extra entries.
*/
typedef struct {
struct u_vector instr_vec;
} nir_instr_worklist;
static inline nir_instr_worklist *
nir_instr_worklist_create() {
nir_instr_worklist *wl = malloc(sizeof(nir_instr_worklist));
if (!wl)
return NULL;
if (!u_vector_init(&wl->instr_vec, sizeof(struct nir_instr *),
sizeof(struct nir_instr *) * 8)) {
free(wl);
return NULL;
}
return wl;
}
static inline uint32_t
nir_instr_worklist_length(nir_instr_worklist *wl)
{
return u_vector_length(&wl->instr_vec);
}
static inline bool
nir_instr_worklist_is_empty(nir_instr_worklist *wl)
{
return nir_instr_worklist_length(wl) == 0;
}
static inline void
nir_instr_worklist_destroy(nir_instr_worklist *wl)
{
u_vector_finish(&wl->instr_vec);
free(wl);
}
static inline void
nir_instr_worklist_push_tail(nir_instr_worklist *wl, nir_instr *instr)
{
struct nir_instr **vec_instr = u_vector_add(&wl->instr_vec);
*vec_instr = instr;
}
static inline nir_instr *
nir_instr_worklist_pop_head(nir_instr_worklist *wl)
{
struct nir_instr **vec_instr = u_vector_remove(&wl->instr_vec);
if (vec_instr == NULL)
return NULL;
return *vec_instr;
}
#define nir_foreach_instr_in_worklist(instr, wl) \
for (nir_instr *instr; (instr = nir_instr_worklist_pop_head(wl));)
#ifdef __cplusplus
} /* extern "C" */
#endif
#endif /* _NIR_WORKLIST_ */
|
