1 perf-bench(1) 1 perf-bench(1)
2 ============= 2 =============
3 3
4 NAME 4 NAME
5 ---- 5 ----
6 perf-bench - General framework for benchmark s 6 perf-bench - General framework for benchmark suites
7 7
8 SYNOPSIS 8 SYNOPSIS
9 -------- 9 --------
10 [verse] 10 [verse]
11 'perf bench' [<common options>] <subsystem> <s 11 'perf bench' [<common options>] <subsystem> <suite> [<options>]
12 12
13 DESCRIPTION 13 DESCRIPTION
14 ----------- 14 -----------
15 This 'perf bench' command is a general framewo 15 This 'perf bench' command is a general framework for benchmark suites.
16 16
17 COMMON OPTIONS 17 COMMON OPTIONS
18 -------------- 18 --------------
19 -r:: 19 -r::
20 --repeat=:: 20 --repeat=::
21 Specify number of times to repeat the run (def !! 21 Specify amount of times to repeat the run (default 10).
22 22
23 -f:: 23 -f::
24 --format=:: 24 --format=::
25 Specify format style. 25 Specify format style.
26 Current available format styles are: 26 Current available format styles are:
27 27
28 'default':: 28 'default'::
29 Default style. This is mainly for human readin 29 Default style. This is mainly for human reading.
30 --------------------- 30 ---------------------
31 % perf bench sched pipe # 31 % perf bench sched pipe # with no style specified
32 (executing 1000000 pipe operations between two 32 (executing 1000000 pipe operations between two tasks)
33 Total time:5.855 sec 33 Total time:5.855 sec
34 5.855061 usecs/op 34 5.855061 usecs/op
35 170792 ops/sec 35 170792 ops/sec
36 --------------------- 36 ---------------------
37 37
38 'simple':: 38 'simple'::
39 This simple style is friendly for automated 39 This simple style is friendly for automated
40 processing by scripts. 40 processing by scripts.
41 --------------------- 41 ---------------------
42 % perf bench --format=simple sched pipe # 42 % perf bench --format=simple sched pipe # specified simple
43 5.988 43 5.988
44 --------------------- 44 ---------------------
45 45
46 SUBSYSTEM 46 SUBSYSTEM
47 --------- 47 ---------
48 48
49 'sched':: 49 'sched'::
50 Scheduler and IPC mechanisms. 50 Scheduler and IPC mechanisms.
51 51
52 'syscall':: <<
53 System call performance (throughput). <<
54 <<
55 'mem':: 52 'mem'::
56 Memory access performance. 53 Memory access performance.
57 54
58 'numa':: 55 'numa'::
59 NUMA scheduling and MM benchmarks. 56 NUMA scheduling and MM benchmarks.
60 57
61 'futex':: 58 'futex'::
62 Futex stressing benchmarks. 59 Futex stressing benchmarks.
63 60
64 'epoll':: <<
65 Eventpoll (epoll) stressing benchmarks <<
66 <<
67 'internals':: <<
68 Benchmark internal perf functionality. <<
69 <<
70 'uprobe':: <<
71 Benchmark overhead of uprobe + BPF. <<
72 <<
73 'all':: 61 'all'::
74 All benchmark subsystems. 62 All benchmark subsystems.
75 63
76 SUITES FOR 'sched' 64 SUITES FOR 'sched'
77 ~~~~~~~~~~~~~~~~~~ 65 ~~~~~~~~~~~~~~~~~~
78 *messaging*:: 66 *messaging*::
79 Suite for evaluating performance of scheduler 67 Suite for evaluating performance of scheduler and IPC mechanisms.
80 Based on hackbench by Rusty Russell. 68 Based on hackbench by Rusty Russell.
81 69
82 Options of *messaging* 70 Options of *messaging*
83 ^^^^^^^^^^^^^^^^^^^^^^ 71 ^^^^^^^^^^^^^^^^^^^^^^
84 -p:: 72 -p::
85 --pipe:: 73 --pipe::
86 Use pipe() instead of socketpair() 74 Use pipe() instead of socketpair()
87 75
88 -t:: 76 -t::
89 --thread:: 77 --thread::
90 Be multi thread instead of multi process 78 Be multi thread instead of multi process
91 79
92 -g:: 80 -g::
93 --group=:: 81 --group=::
94 Specify number of groups 82 Specify number of groups
95 83
96 -l:: 84 -l::
97 --nr_loops=:: 85 --nr_loops=::
98 Specify number of loops 86 Specify number of loops
99 87
100 Example of *messaging* 88 Example of *messaging*
101 ^^^^^^^^^^^^^^^^^^^^^^ 89 ^^^^^^^^^^^^^^^^^^^^^^
102 90
103 --------------------- 91 ---------------------
104 % perf bench sched messaging # 92 % perf bench sched messaging # run with default
105 options (20 sender and receiver processes per 93 options (20 sender and receiver processes per group)
106 (10 groups == 400 processes run) 94 (10 groups == 400 processes run)
107 95
108 Total time:0.308 sec 96 Total time:0.308 sec
109 97
110 % perf bench sched messaging -t -g 20 # 98 % perf bench sched messaging -t -g 20 # be multi-thread, with 20 groups
111 (20 sender and receiver threads per group) 99 (20 sender and receiver threads per group)
112 (20 groups == 800 threads run) 100 (20 groups == 800 threads run)
113 101
114 Total time:0.582 sec 102 Total time:0.582 sec
115 --------------------- 103 ---------------------
116 104
117 *pipe*:: 105 *pipe*::
118 Suite for pipe() system call. 106 Suite for pipe() system call.
119 Based on pipe-test-1m.c by Ingo Molnar. 107 Based on pipe-test-1m.c by Ingo Molnar.
120 108
121 Options of *pipe* 109 Options of *pipe*
122 ^^^^^^^^^^^^^^^^^ 110 ^^^^^^^^^^^^^^^^^
123 -l:: 111 -l::
124 --loop=:: 112 --loop=::
125 Specify number of loops. 113 Specify number of loops.
126 114
127 -G:: <<
128 --cgroups=:: <<
129 Names of cgroups for sender and receiver, sepa <<
130 This is useful to check cgroup context switchi <<
131 Note that perf doesn't create nor delete the c <<
132 make sure that the cgroups exist and are acces <<
133 <<
134 <<
135 Example of *pipe* 115 Example of *pipe*
136 ^^^^^^^^^^^^^^^^^ 116 ^^^^^^^^^^^^^^^^^
137 117
138 --------------------- 118 ---------------------
139 % perf bench sched pipe 119 % perf bench sched pipe
140 (executing 1000000 pipe operations between two 120 (executing 1000000 pipe operations between two tasks)
141 121
142 Total time:8.091 sec 122 Total time:8.091 sec
143 8.091833 usecs/op 123 8.091833 usecs/op
144 123581 ops/sec 124 123581 ops/sec
145 125
146 % perf bench sched pipe -l 1000 # 126 % perf bench sched pipe -l 1000 # loop 1000
147 (executing 1000 pipe operations between two ta 127 (executing 1000 pipe operations between two tasks)
148 128
149 Total time:0.016 sec 129 Total time:0.016 sec
150 16.948000 usecs/op 130 16.948000 usecs/op
151 59004 ops/sec 131 59004 ops/sec
152 <<
153 % perf bench sched pipe -G AAA,BBB <<
154 (executing 1000000 pipe operations between cgr <<
155 # Running 'sched/pipe' benchmark: <<
156 # Executed 1000000 pipe operations between two <<
157 <<
158 Total time: 6.886 [sec] <<
159 <<
160 6.886208 usecs/op <<
161 145217 ops/sec <<
162 <<
163 --------------------- 132 ---------------------
164 133
165 SUITES FOR 'syscall' <<
166 ~~~~~~~~~~~~~~~~~~ <<
167 *basic*:: <<
168 Suite for evaluating performance of core syste <<
169 This uses a single thread simply doing getppid <<
170 cached by glibc. <<
171 <<
172 <<
173 SUITES FOR 'mem' 134 SUITES FOR 'mem'
174 ~~~~~~~~~~~~~~~~ 135 ~~~~~~~~~~~~~~~~
175 *memcpy*:: 136 *memcpy*::
176 Suite for evaluating performance of simple mem 137 Suite for evaluating performance of simple memory copy in various ways.
177 138
178 Options of *memcpy* 139 Options of *memcpy*
179 ^^^^^^^^^^^^^^^^^^^ 140 ^^^^^^^^^^^^^^^^^^^
180 -l:: 141 -l::
181 --size:: 142 --size::
182 Specify size of memory to copy (default: 1MB). 143 Specify size of memory to copy (default: 1MB).
183 Available units are B, KB, MB, GB and TB (case 144 Available units are B, KB, MB, GB and TB (case insensitive).
184 145
185 -f:: 146 -f::
186 --function:: 147 --function::
187 Specify function to copy (default: default). 148 Specify function to copy (default: default).
188 Available functions are depend on the architec 149 Available functions are depend on the architecture.
189 On x86-64, x86-64-unrolled, x86-64-movsq and x 150 On x86-64, x86-64-unrolled, x86-64-movsq and x86-64-movsb are supported.
190 151
191 -l:: 152 -l::
192 --nr_loops:: 153 --nr_loops::
193 Repeat memcpy invocation this number of times. 154 Repeat memcpy invocation this number of times.
194 155
195 -c:: 156 -c::
196 --cycles:: 157 --cycles::
197 Use perf's cpu-cycles event instead of gettime 158 Use perf's cpu-cycles event instead of gettimeofday syscall.
198 159
199 *memset*:: 160 *memset*::
200 Suite for evaluating performance of simple mem 161 Suite for evaluating performance of simple memory set in various ways.
201 162
202 Options of *memset* 163 Options of *memset*
203 ^^^^^^^^^^^^^^^^^^^ 164 ^^^^^^^^^^^^^^^^^^^
204 -l:: 165 -l::
205 --size:: 166 --size::
206 Specify size of memory to set (default: 1MB). 167 Specify size of memory to set (default: 1MB).
207 Available units are B, KB, MB, GB and TB (case 168 Available units are B, KB, MB, GB and TB (case insensitive).
208 169
209 -f:: 170 -f::
210 --function:: 171 --function::
211 Specify function to set (default: default). 172 Specify function to set (default: default).
212 Available functions are depend on the architec 173 Available functions are depend on the architecture.
213 On x86-64, x86-64-unrolled, x86-64-stosq and x 174 On x86-64, x86-64-unrolled, x86-64-stosq and x86-64-stosb are supported.
214 175
215 -l:: 176 -l::
216 --nr_loops:: 177 --nr_loops::
217 Repeat memset invocation this number of times. 178 Repeat memset invocation this number of times.
218 179
219 -c:: 180 -c::
220 --cycles:: 181 --cycles::
221 Use perf's cpu-cycles event instead of gettime 182 Use perf's cpu-cycles event instead of gettimeofday syscall.
222 183
223 SUITES FOR 'numa' 184 SUITES FOR 'numa'
224 ~~~~~~~~~~~~~~~~~ 185 ~~~~~~~~~~~~~~~~~
225 *mem*:: 186 *mem*::
226 Suite for evaluating NUMA workloads. 187 Suite for evaluating NUMA workloads.
227 188
228 SUITES FOR 'futex' 189 SUITES FOR 'futex'
229 ~~~~~~~~~~~~~~~~~~ 190 ~~~~~~~~~~~~~~~~~~
230 *hash*:: 191 *hash*::
231 Suite for evaluating hash tables. 192 Suite for evaluating hash tables.
232 193
233 *wake*:: 194 *wake*::
234 Suite for evaluating wake calls. 195 Suite for evaluating wake calls.
235 196
236 *wake-parallel*:: 197 *wake-parallel*::
237 Suite for evaluating parallel wake calls. 198 Suite for evaluating parallel wake calls.
238 199
239 *requeue*:: 200 *requeue*::
240 Suite for evaluating requeue calls. 201 Suite for evaluating requeue calls.
241 202
242 *lock-pi*:: 203 *lock-pi*::
243 Suite for evaluating futex lock_pi calls. 204 Suite for evaluating futex lock_pi calls.
244 205
245 SUITES FOR 'epoll' <<
246 ~~~~~~~~~~~~~~~~~~ <<
247 *wait*:: <<
248 Suite for evaluating concurrent epoll_wait cal <<
249 <<
250 *ctl*:: <<
251 Suite for evaluating multiple epoll_ctl calls. <<
252 <<
253 SUITES FOR 'internals' <<
254 ~~~~~~~~~~~~~~~~~~~~~~ <<
255 *synthesize*:: <<
256 Suite for evaluating perf's event synthesis pe <<
257 206
258 SEE ALSO 207 SEE ALSO
259 -------- 208 --------
260 linkperf:perf[1] 209 linkperf:perf[1]
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.