Cleanup: move generic caa_get_cycles to arch/generic.h
[urcu.git] / urcu / arch / ppc.h
... / ...
CommitLineData
1#ifndef _URCU_ARCH_PPC_H
2#define _URCU_ARCH_PPC_H
3
4/*
5 * arch_ppc.h: trivial definitions for the powerpc architecture.
6 *
7 * Copyright (c) 2009 Paul E. McKenney, IBM Corporation.
8 * Copyright (c) 2009 Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
9 *
10 * This library is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU Lesser General Public
12 * License as published by the Free Software Foundation; either
13 * version 2.1 of the License, or (at your option) any later version.
14 *
15 * This library is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
18 * Lesser General Public License for more details.
19 *
20 * You should have received a copy of the GNU Lesser General Public
21 * License along with this library; if not, write to the Free Software
22 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
23 */
24
25#include <urcu/compiler.h>
26#include <urcu/config.h>
27#include <urcu/syscall-compat.h>
28
29#ifdef __cplusplus
30extern "C" {
31#endif
32
33/* Include size of POWER5+ L3 cache lines: 256 bytes */
34#define CAA_CACHE_LINE_SIZE 256
35
36#ifdef __NO_LWSYNC__
37#define LWSYNC_OPCODE "sync\n"
38#else
39#define LWSYNC_OPCODE "lwsync\n"
40#endif
41
42/*
43 * Use sync for all cmm_mb/rmb/wmb barriers because lwsync does not
44 * preserve ordering of cacheable vs. non-cacheable accesses, so it
45 * should not be used to order with respect to MMIO operations. An
46 * eieio+lwsync pair is also not enough for cmm_rmb, because it will
47 * order cacheable and non-cacheable memory operations separately---i.e.
48 * not the latter against the former.
49 */
50#define cmm_mb() __asm__ __volatile__ ("sync":::"memory")
51
52/*
53 * lwsync orders loads in cacheable memory with respect to other loads,
54 * and stores in cacheable memory with respect to other stores.
55 * Therefore, use it for barriers ordering accesses to cacheable memory
56 * only.
57 */
58#define cmm_smp_rmb() __asm__ __volatile__ (LWSYNC_OPCODE:::"memory")
59#define cmm_smp_wmb() __asm__ __volatile__ (LWSYNC_OPCODE:::"memory")
60
61#define mftbl() \
62 __extension__ \
63 ({ \
64 unsigned long rval; \
65 __asm__ __volatile__ ("mftbl %0" : "=r" (rval)); \
66 rval; \
67 })
68
69#define mftbu() \
70 __extension__ \
71 ({ \
72 unsigned long rval; \
73 __asm__ __volatile__ ("mftbu %0" : "=r" (rval)); \
74 rval; \
75 })
76
77#define mftb() \
78 __extension__ \
79 ({ \
80 unsigned long long rval; \
81 __asm__ __volatile__ ("mftb %0" : "=r" (rval)); \
82 rval; \
83 })
84
85#define HAS_CAA_GET_CYCLES
86
87typedef unsigned long long cycles_t;
88
89#ifdef __powerpc64__
90static inline cycles_t caa_get_cycles(void)
91{
92 return (cycles_t) mftb();
93}
94#else
95static inline cycles_t caa_get_cycles(void)
96{
97 unsigned long h, l;
98
99 for (;;) {
100 h = mftbu();
101 cmm_barrier();
102 l = mftbl();
103 cmm_barrier();
104 if (mftbu() == h)
105 return (((cycles_t) h) << 32) + l;
106 }
107}
108#endif
109
110/*
111 * Define the membarrier system call number if not yet available in the
112 * system headers.
113 */
114#ifndef __NR_membarrier
115#define __NR_membarrier 365
116#endif
117
118#ifdef __cplusplus
119}
120#endif
121
122#include <urcu/arch/generic.h>
123
124#endif /* _URCU_ARCH_PPC_H */
This page took 0.022424 seconds and 4 git commands to generate.