| Hitoshi Mitake | 9fbc04f | 2009-11-10 20:50:54 +0900 | [diff] [blame] | 1 | perf-bench(1) |
| Arnaldo Carvalho de Melo | 4778e0e | 2010-05-05 11:23:27 -0300 | [diff] [blame] | 2 | ============= |
| Hitoshi Mitake | 9fbc04f | 2009-11-10 20:50:54 +0900 | [diff] [blame] | 3 | |
| 4 | NAME |
| 5 | ---- |
| 6 | perf-bench - General framework for benchmark suites |
| 7 | |
| 8 | SYNOPSIS |
| 9 | -------- |
| 10 | [verse] |
| 11 | 'perf bench' [<common options>] <subsystem> <suite> [<options>] |
| 12 | |
| 13 | DESCRIPTION |
| 14 | ----------- |
| 15 | This 'perf bench' command is general framework for benchmark suites. |
| 16 | |
| 17 | COMMON OPTIONS |
| 18 | -------------- |
| 19 | -f:: |
| 20 | --format=:: |
| 21 | Specify format style. |
| Randy Dunlap | 854c554 | 2010-03-31 11:31:00 -0700 | [diff] [blame] | 22 | Current available format styles are: |
| Hitoshi Mitake | 9fbc04f | 2009-11-10 20:50:54 +0900 | [diff] [blame] | 23 | |
| 24 | 'default':: |
| 25 | Default style. This is mainly for human reading. |
| 26 | --------------------- |
| Randy Dunlap | 854c554 | 2010-03-31 11:31:00 -0700 | [diff] [blame] | 27 | % perf bench sched pipe # with no style specified |
| Hitoshi Mitake | 9fbc04f | 2009-11-10 20:50:54 +0900 | [diff] [blame] | 28 | (executing 1000000 pipe operations between two tasks) |
| 29 | Total time:5.855 sec |
| 30 | 5.855061 usecs/op |
| 31 | 170792 ops/sec |
| 32 | --------------------- |
| 33 | |
| 34 | 'simple':: |
| 35 | This simple style is friendly for automated |
| 36 | processing by scripts. |
| 37 | --------------------- |
| 38 | % perf bench --format=simple sched pipe # specified simple |
| 39 | 5.988 |
| 40 | --------------------- |
| 41 | |
| 42 | SUBSYSTEM |
| 43 | --------- |
| 44 | |
| 45 | 'sched':: |
| 46 | Scheduler and IPC mechanisms. |
| 47 | |
| 48 | SUITES FOR 'sched' |
| 49 | ~~~~~~~~~~~~~~~~~~ |
| 50 | *messaging*:: |
| 51 | Suite for evaluating performance of scheduler and IPC mechanisms. |
| 52 | Based on hackbench by Rusty Russell. |
| 53 | |
| 54 | Options of *pipe* |
| 55 | ^^^^^^^^^^^^^^^^^ |
| 56 | -p:: |
| 57 | --pipe:: |
| 58 | Use pipe() instead of socketpair() |
| 59 | |
| 60 | -t:: |
| 61 | --thread:: |
| 62 | Be multi thread instead of multi process |
| 63 | |
| 64 | -g:: |
| 65 | --group=:: |
| 66 | Specify number of groups |
| 67 | |
| 68 | -l:: |
| 69 | --loop=:: |
| 70 | Specify number of loops |
| 71 | |
| 72 | Example of *messaging* |
| 73 | ^^^^^^^^^^^^^^^^^^^^^^ |
| 74 | |
| 75 | --------------------- |
| 76 | % perf bench sched messaging # run with default |
| 77 | options (20 sender and receiver processes per group) |
| 78 | (10 groups == 400 processes run) |
| 79 | |
| 80 | Total time:0.308 sec |
| 81 | |
| Randy Dunlap | 854c554 | 2010-03-31 11:31:00 -0700 | [diff] [blame] | 82 | % perf bench sched messaging -t -g 20 # be multi-thread, with 20 groups |
| Hitoshi Mitake | 9fbc04f | 2009-11-10 20:50:54 +0900 | [diff] [blame] | 83 | (20 sender and receiver threads per group) |
| 84 | (20 groups == 800 threads run) |
| 85 | |
| 86 | Total time:0.582 sec |
| 87 | --------------------- |
| 88 | |
| 89 | *pipe*:: |
| 90 | Suite for pipe() system call. |
| 91 | Based on pipe-test-1m.c by Ingo Molnar. |
| 92 | |
| 93 | Options of *pipe* |
| 94 | ^^^^^^^^^^^^^^^^^ |
| 95 | -l:: |
| 96 | --loop=:: |
| 97 | Specify number of loops. |
| 98 | |
| 99 | Example of *pipe* |
| 100 | ^^^^^^^^^^^^^^^^^ |
| 101 | |
| 102 | --------------------- |
| 103 | % perf bench sched pipe |
| 104 | (executing 1000000 pipe operations between two tasks) |
| 105 | |
| 106 | Total time:8.091 sec |
| 107 | 8.091833 usecs/op |
| 108 | 123581 ops/sec |
| 109 | |
| 110 | % perf bench sched pipe -l 1000 # loop 1000 |
| 111 | (executing 1000 pipe operations between two tasks) |
| 112 | |
| 113 | Total time:0.016 sec |
| 114 | 16.948000 usecs/op |
| 115 | 59004 ops/sec |
| 116 | --------------------- |
| 117 | |
| 118 | SEE ALSO |
| 119 | -------- |
| 120 | linkperf:perf[1] |