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
|
/* Measure memmove function combined throughput for different alignments.
Copyright (C) 2017-2023 Free Software Foundation, Inc.
This file is part of the GNU C Library.
The GNU C Library is free software; you can redistribute it and/or
modify it under the terms of the GNU Lesser General Public
License as published by the Free Software Foundation; either
version 2.1 of the License, or (at your option) any later version.
The GNU C Library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public
License along with the GNU C Library; if not, see
<https://www.gnu.org/licenses/>. */
/* This microbenchmark measures the throughput of memmove for various sizes from
1 byte to 32MiB, doubling every iteration and then misaligning by 0-15
bytes. The copies are done from source to destination and then back and the
source walks forward across the array and the destination walks backward by
one byte each, thus measuring misaligned accesses as well. The idea is to
avoid caching effects by copying a different string and far enough from each
other, walking in different directions so that we can measure prefetcher
efficiency (software or hardware) more closely than with a loop copying the
same data over and over, which eventually only gives us L1 cache
performance. */
#ifndef MEMMOVE_RESULT
# define MEMMOVE_RESULT(dst, len) dst
# define START_SIZE 128
# define MIN_PAGE_SIZE (getpagesize () + 32 * 1024 * 1024)
# define TEST_MAIN
# define TEST_NAME "memmove"
# define TIMEOUT (20 * 60)
# include "bench-string.h"
#define NO_OVERLAP 0
#define PARTIAL_OVERLAP 1
#define COMPLETE_OVERLAP 2
IMPL (memmove, 1)
#endif
#include "json-lib.h"
typedef char *(*proto_t) (char *, const char *, size_t);
static void
do_one_test (json_ctx_t *json_ctx, impl_t *impl, char *dst, char *src,
size_t len)
{
size_t i = 0;
timing_t start, stop, cur;
char *dst_end = dst + MIN_PAGE_SIZE - len;
char *src_end = src + MIN_PAGE_SIZE - len;
TIMING_NOW (start);
/* Copy the entire buffer backwards, LEN at a time. */
for (; src_end >= src && dst <= dst_end; dst += len, src_end -= len, i++)
CALL (impl, dst, src_end, len);
TIMING_NOW (stop);
TIMING_DIFF (cur, start, stop);
/* Get time taken per function call. */
json_element_double (json_ctx, (double) cur / i);
}
static void
do_test (json_ctx_t *json_ctx, size_t len, int overlap, int both_ways)
{
char *s1, *s2, *tmp;
size_t repeats;
s1 = (char *) (buf1);
s2 = (char *) (buf2);
if (overlap != NO_OVERLAP)
s2 = s1;
if (overlap == PARTIAL_OVERLAP)
s2 += len / 2;
for (repeats = both_ways ? 2 : 1; repeats; --repeats)
{
json_element_object_begin (json_ctx);
json_attr_uint (json_ctx, "length", (double) len);
json_attr_string(json_ctx, "overlap",
overlap == NO_OVERLAP ? "none"
: overlap == PARTIAL_OVERLAP ? "partial"
: "complete");
json_attr_uint (json_ctx, "dst > src", (double) (s2 > s1));
json_array_begin (json_ctx, "timings");
FOR_EACH_IMPL (impl, 0)
do_one_test (json_ctx, impl, (char *) buf2, (char *) buf1, len);
json_array_end (json_ctx);
json_element_object_end (json_ctx);
tmp = s1;
s1 = s2;
s2 = tmp;
}
}
int
test_main (void)
{
json_ctx_t json_ctx;
test_init ();
json_init (&json_ctx, 0, stdout);
json_document_begin (&json_ctx);
json_attr_string (&json_ctx, "timing_type", TIMING_TYPE);
json_attr_object_begin (&json_ctx, "functions");
json_attr_object_begin (&json_ctx, "memmove");
json_attr_string (&json_ctx, "bench-variant", "walk");
json_array_begin (&json_ctx, "ifuncs");
FOR_EACH_IMPL (impl, 0)
json_element_string (&json_ctx, impl->name);
json_array_end (&json_ctx);
json_array_begin (&json_ctx, "results");
/* Non-overlapping buffers. */
for (size_t i = START_SIZE; i <= MIN_PAGE_SIZE; i <<= 1)
{
do_test (&json_ctx, i, NO_OVERLAP, 1);
do_test (&json_ctx, i + 1, NO_OVERLAP, 1);
}
/* Partially-overlapping buffers. */
for (size_t i = START_SIZE; i <= MIN_PAGE_SIZE / 2; i <<= 1)
{
do_test (&json_ctx, i, PARTIAL_OVERLAP, 1);
do_test (&json_ctx, i + 1, PARTIAL_OVERLAP, 1);
}
/* Complete-overlapping buffers. */
for (size_t i = START_SIZE; i <= MIN_PAGE_SIZE; i <<= 1)
{
do_test (&json_ctx, i, COMPLETE_OVERLAP, 0);
do_test (&json_ctx, i + 1, COMPLETE_OVERLAP, 0);
}
json_array_end (&json_ctx);
json_attr_object_end (&json_ctx);
json_attr_object_end (&json_ctx);
json_document_end (&json_ctx);
return ret;
}
#include <support/test-driver.c>
|
