1 /*
2 * Copyright © 2015 Intel
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 DEALINGS
21 * IN THE SOFTWARE.
22 */
23
24 #ifndef _SIMPLE_MTX_H
25 #define _SIMPLE_MTX_H
26
27 #include "util/futex.h"
28
29 #include "c11/threads.h"
30
31 #if defined(__GNUC__) && defined(HAVE_LINUX_FUTEX_H)
32
33 /* mtx_t - Fast, simple mutex
34 *
35 * While modern pthread mutexes are very fast (implemented using futex), they
36 * still incur a call to an external DSO and overhead of the generality and
37 * features of pthread mutexes. Most mutexes in mesa only needs lock/unlock,
38 * and the idea here is that we can inline the atomic operation and make the
39 * fast case just two intructions. Mutexes are subtle and finicky to
40 * implement, so we carefully copy the implementation from Ulrich Dreppers
41 * well-written and well-reviewed paper:
42 *
43 * "Futexes Are Tricky"
44 * http://www.akkadia.org/drepper/futex.pdf
45 *
46 * We implement "mutex3", which gives us a mutex that has no syscalls on
47 * uncontended lock or unlock. Further, the uncontended case boils down to a
48 * locked cmpxchg and an untaken branch, the uncontended unlock is just a
49 * locked decr and an untaken branch. We use __builtin_expect() to indicate
50 * that contention is unlikely so that gcc will put the contention code out of
51 * the main code flow.
52 *
53 * A fast mutex only supports lock/unlock, can't be recursive or used with
54 * condition variables.
55 */
56
57 typedef struct {
58 uint32_t val;
59 } simple_mtx_t;
60
61 #define _SIMPLE_MTX_INITIALIZER_NP { 0 }
62
63 static inline void
simple_mtx_init(simple_mtx_t * mtx,MAYBE_UNUSED int type)64 simple_mtx_init(simple_mtx_t *mtx, MAYBE_UNUSED int type)
65 {
66 assert(type == mtx_plain);
67
68 mtx->val = 0;
69 }
70
71 static inline void
simple_mtx_destroy(UNUSED simple_mtx_t * mtx)72 simple_mtx_destroy(UNUSED simple_mtx_t *mtx)
73 {
74 }
75
76 static inline void
simple_mtx_lock(simple_mtx_t * mtx)77 simple_mtx_lock(simple_mtx_t *mtx)
78 {
79 uint32_t c;
80
81 c = __sync_val_compare_and_swap(&mtx->val, 0, 1);
82 if (__builtin_expect(c != 0, 0)) {
83 if (c != 2)
84 c = __sync_lock_test_and_set(&mtx->val, 2);
85 while (c != 0) {
86 futex_wait(&mtx->val, 2, NULL);
87 c = __sync_lock_test_and_set(&mtx->val, 2);
88 }
89 }
90 }
91
92 static inline void
simple_mtx_unlock(simple_mtx_t * mtx)93 simple_mtx_unlock(simple_mtx_t *mtx)
94 {
95 uint32_t c;
96
97 c = __sync_fetch_and_sub(&mtx->val, 1);
98 if (__builtin_expect(c != 1, 0)) {
99 mtx->val = 0;
100 futex_wake(&mtx->val, 1);
101 }
102 }
103
104 #else
105
106 typedef mtx_t simple_mtx_t;
107
108 #define _SIMPLE_MTX_INITIALIZER_NP _MTX_INITIALIZER_NP
109
110 static inline void
simple_mtx_init(simple_mtx_t * mtx,int type)111 simple_mtx_init(simple_mtx_t *mtx, int type)
112 {
113 mtx_init(mtx, type);
114 }
115
116 static inline void
simple_mtx_destroy(simple_mtx_t * mtx)117 simple_mtx_destroy(simple_mtx_t *mtx)
118 {
119 mtx_destroy(mtx);
120 }
121
122 static inline void
simple_mtx_lock(simple_mtx_t * mtx)123 simple_mtx_lock(simple_mtx_t *mtx)
124 {
125 mtx_lock(mtx);
126 }
127
128 static inline void
simple_mtx_unlock(simple_mtx_t * mtx)129 simple_mtx_unlock(simple_mtx_t *mtx)
130 {
131 mtx_unlock(mtx);
132 }
133
134 #endif
135
136 #endif
137