perf-bench.txt (4951B)
1perf-bench(1) 2============= 3 4NAME 5---- 6perf-bench - General framework for benchmark suites 7 8SYNOPSIS 9-------- 10[verse] 11'perf bench' [<common options>] <subsystem> <suite> [<options>] 12 13DESCRIPTION 14----------- 15This 'perf bench' command is a general framework for benchmark suites. 16 17COMMON OPTIONS 18-------------- 19-r:: 20--repeat=:: 21Specify amount of times to repeat the run (default 10). 22 23-f:: 24--format=:: 25Specify format style. 26Current available format styles are: 27 28'default':: 29Default style. This is mainly for human reading. 30--------------------- 31% perf bench sched pipe # with no style specified 32(executing 1000000 pipe operations between two tasks) 33 Total time:5.855 sec 34 5.855061 usecs/op 35 170792 ops/sec 36--------------------- 37 38'simple':: 39This simple style is friendly for automated 40processing by scripts. 41--------------------- 42% perf bench --format=simple sched pipe # specified simple 435.988 44--------------------- 45 46SUBSYSTEM 47--------- 48 49'sched':: 50 Scheduler and IPC mechanisms. 51 52'syscall':: 53 System call performance (throughput). 54 55'mem':: 56 Memory access performance. 57 58'numa':: 59 NUMA scheduling and MM benchmarks. 60 61'futex':: 62 Futex stressing benchmarks. 63 64'epoll':: 65 Eventpoll (epoll) stressing benchmarks. 66 67'internals':: 68 Benchmark internal perf functionality. 69 70'all':: 71 All benchmark subsystems. 72 73SUITES FOR 'sched' 74~~~~~~~~~~~~~~~~~~ 75*messaging*:: 76Suite for evaluating performance of scheduler and IPC mechanisms. 77Based on hackbench by Rusty Russell. 78 79Options of *messaging* 80^^^^^^^^^^^^^^^^^^^^^^ 81-p:: 82--pipe:: 83Use pipe() instead of socketpair() 84 85-t:: 86--thread:: 87Be multi thread instead of multi process 88 89-g:: 90--group=:: 91Specify number of groups 92 93-l:: 94--nr_loops=:: 95Specify number of loops 96 97Example of *messaging* 98^^^^^^^^^^^^^^^^^^^^^^ 99 100--------------------- 101% perf bench sched messaging # run with default 102options (20 sender and receiver processes per group) 103(10 groups == 400 processes run) 104 105 Total time:0.308 sec 106 107% perf bench sched messaging -t -g 20 # be multi-thread, with 20 groups 108(20 sender and receiver threads per group) 109(20 groups == 800 threads run) 110 111 Total time:0.582 sec 112--------------------- 113 114*pipe*:: 115Suite for pipe() system call. 116Based on pipe-test-1m.c by Ingo Molnar. 117 118Options of *pipe* 119^^^^^^^^^^^^^^^^^ 120-l:: 121--loop=:: 122Specify number of loops. 123 124Example of *pipe* 125^^^^^^^^^^^^^^^^^ 126 127--------------------- 128% perf bench sched pipe 129(executing 1000000 pipe operations between two tasks) 130 131 Total time:8.091 sec 132 8.091833 usecs/op 133 123581 ops/sec 134 135% perf bench sched pipe -l 1000 # loop 1000 136(executing 1000 pipe operations between two tasks) 137 138 Total time:0.016 sec 139 16.948000 usecs/op 140 59004 ops/sec 141--------------------- 142 143SUITES FOR 'syscall' 144~~~~~~~~~~~~~~~~~~ 145*basic*:: 146Suite for evaluating performance of core system call throughput (both usecs/op and ops/sec metrics). 147This uses a single thread simply doing getppid(2), which is a simple syscall where the result is not 148cached by glibc. 149 150 151SUITES FOR 'mem' 152~~~~~~~~~~~~~~~~ 153*memcpy*:: 154Suite for evaluating performance of simple memory copy in various ways. 155 156Options of *memcpy* 157^^^^^^^^^^^^^^^^^^^ 158-l:: 159--size:: 160Specify size of memory to copy (default: 1MB). 161Available units are B, KB, MB, GB and TB (case insensitive). 162 163-f:: 164--function:: 165Specify function to copy (default: default). 166Available functions are depend on the architecture. 167On x86-64, x86-64-unrolled, x86-64-movsq and x86-64-movsb are supported. 168 169-l:: 170--nr_loops:: 171Repeat memcpy invocation this number of times. 172 173-c:: 174--cycles:: 175Use perf's cpu-cycles event instead of gettimeofday syscall. 176 177*memset*:: 178Suite for evaluating performance of simple memory set in various ways. 179 180Options of *memset* 181^^^^^^^^^^^^^^^^^^^ 182-l:: 183--size:: 184Specify size of memory to set (default: 1MB). 185Available units are B, KB, MB, GB and TB (case insensitive). 186 187-f:: 188--function:: 189Specify function to set (default: default). 190Available functions are depend on the architecture. 191On x86-64, x86-64-unrolled, x86-64-stosq and x86-64-stosb are supported. 192 193-l:: 194--nr_loops:: 195Repeat memset invocation this number of times. 196 197-c:: 198--cycles:: 199Use perf's cpu-cycles event instead of gettimeofday syscall. 200 201SUITES FOR 'numa' 202~~~~~~~~~~~~~~~~~ 203*mem*:: 204Suite for evaluating NUMA workloads. 205 206SUITES FOR 'futex' 207~~~~~~~~~~~~~~~~~~ 208*hash*:: 209Suite for evaluating hash tables. 210 211*wake*:: 212Suite for evaluating wake calls. 213 214*wake-parallel*:: 215Suite for evaluating parallel wake calls. 216 217*requeue*:: 218Suite for evaluating requeue calls. 219 220*lock-pi*:: 221Suite for evaluating futex lock_pi calls. 222 223SUITES FOR 'epoll' 224~~~~~~~~~~~~~~~~~~ 225*wait*:: 226Suite for evaluating concurrent epoll_wait calls. 227 228*ctl*:: 229Suite for evaluating multiple epoll_ctl calls. 230 231SUITES FOR 'internals' 232~~~~~~~~~~~~~~~~~~~~~~ 233*synthesize*:: 234Suite for evaluating perf's event synthesis performance. 235 236SEE ALSO 237-------- 238linkperf:perf[1]