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
|
/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 4 -*-
* vim: set ts=8 sts=4 et sw=4 tw=99:
* This Source Code Form is subject to the terms of the Mozilla Public
* License, v. 2.0. If a copy of the MPL was not distributed with this
* file, You can obtain one at http://mozilla.org/MPL/2.0/. */
#ifndef vm_Time_h
#define vm_Time_h
#include <stddef.h>
#include <stdint.h>
/*
* Broken down form of 64 bit time value.
*/
struct PRMJTime {
int32_t tm_usec; /* microseconds of second (0-999999) */
int8_t tm_sec; /* seconds of minute (0-59) */
int8_t tm_min; /* minutes of hour (0-59) */
int8_t tm_hour; /* hour of day (0-23) */
int8_t tm_mday; /* day of month (1-31) */
int8_t tm_mon; /* month of year (0-11) */
int8_t tm_wday; /* 0=sunday, 1=monday, ... */
int32_t tm_year; /* absolute year, AD */
int16_t tm_yday; /* day of year (0 to 365) */
int8_t tm_isdst; /* non-zero if DST in effect */
};
/* Some handy constants */
#define PRMJ_USEC_PER_SEC 1000000L
#define PRMJ_USEC_PER_MSEC 1000L
/* Return the current local time in micro-seconds */
extern int64_t
PRMJ_Now();
/* Initialize the resources associated with PRMJ_Now. */
#if defined(XP_WIN)
extern void
PRMJ_NowInit();
#else
inline void
PRMJ_NowInit() {}
#endif
/* Release the resources associated with PRMJ_Now; don't call PRMJ_Now again */
#ifdef XP_WIN
extern void
PRMJ_NowShutdown();
#else
inline void
PRMJ_NowShutdown() {}
#endif
/* Format a time value into a buffer. Same semantics as strftime() */
extern size_t
PRMJ_FormatTime(char* buf, int buflen, const char* fmt, PRMJTime* tm);
/**
* Requesting the number of cycles from the CPU.
*
* `rdtsc`, or Read TimeStamp Cycle, is an instruction provided by
* x86-compatible CPUs that lets processes request the number of
* cycles spent by the CPU executing instructions since the CPU was
* started. It may be used for performance monitoring, but you should
* be aware of the following limitations.
*
*
* 1. The value is *not* monotonic.
*
* The value is reset to 0 whenever a CPU is turned off (e.g. computer
* in full hibernation, single CPU going turned off). Moreover, on
* multi-core/multi-CPU architectures, the cycles of each core/CPU are
* generally not synchronized. Therefore, is a process or thread is
* rescheduled to another core/CPU, the result of `rdtsc` may decrease
* arbitrarily.
*
* The only way to prevent this is to pin your thread to a particular
* CPU, which is generally not a good idea.
*
*
*
* 2. The value increases independently.
*
* The value may increase whenever the CPU executes an instruction,
* regardless of the process that has issued this
* instruction. Moreover, if a process or thread is rescheduled to
* another core/CPU, the result of `rdtsc` may increase arbitrarily.
*
* The only way to prevent this is to ensure that your thread is the
* sole owner of the CPU. See [1] for an example. This is also
* generally not a good idea.
*
*
*
* 3. The value does not measure time.
*
* On older architectures (pre-Pentium 4), there was no constant mapping
* between rdtsc and CPU time.
*
*
* 4. Instructions may be reordered.
*
* The CPU can reorder instructions. Also, rdtsc does not necessarily
* wait until all previous instructions have finished executing before
* reading the counter. Similarly, subsequent instructions may begin
* execution before the read operation is performed. If you use rdtsc
* for micro-benchmarking, you may end up measuring something else
* than what you expect. See [1] for a study of countermeasures.
*
*
* ** Performance
*
* According to unchecked sources on the web, the overhead of rdtsc is
* expected to be 150-200 cycles on old architectures, 6-50 on newer
* architectures. Agner's instruction tables [2] seem to confirm the latter
* results.
*
*
* [1]
* http://www.intel.com/content/dam/www/public/us/en/documents/white-papers/ia-32-ia-64-benchmark-code-execution-paper.pdf
* [2] http://www.agner.org/optimize/instruction_tables.pdf
*/
#define MOZ_HAVE_RDTSC 1
#if defined(_WIN32)
#include <intrin.h>
static __inline uint64_t
ReadTimestampCounter(void)
{
return __rdtsc();
}
#elif defined(__i386__)
static __inline__ uint64_t
ReadTimestampCounter(void)
{
uint64_t x;
__asm__ volatile (".byte 0x0f, 0x31" : "=A" (x));
return x;
}
#elif defined(__x86_64__)
static __inline__ uint64_t
ReadTimestampCounter(void)
{
unsigned hi, lo;
__asm__ __volatile__ ("rdtsc" : "=a"(lo), "=d"(hi));
return ( (uint64_t)lo)|( ((uint64_t)hi)<<32 );
}
#else
#undef MOZ_HAVE_RDTSC
#endif
#endif /* vm_Time_h */
|
