mirror of https://github.com/torvalds/linux.git
99476fa085
The function worker_thread() is programmed in a way that roughly
doubles the number of expectable context switches, because it enforces
blocking reads:
Performance counter stats for 'perf bench sched pipe':
2,000,004 context-switches
11.859548321 seconds time elapsed
0.674871000 seconds user
8.076890000 seconds sys
The result of this behavior is that the blocking reads by far dominate
the performance analysis of 'perf bench sched pipe':
Samples: 78K of event 'cycles:P', Event count (approx.): 27964965844
Overhead Command Shared Object Symbol
25.28% sched-pipe [kernel.kallsyms] [k] read_hpet
8.11% sched-pipe [kernel.kallsyms] [k] retbleed_untrain_ret
2.82% sched-pipe [kernel.kallsyms] [k] pipe_write
From the code, it is unclear if that behavior is wanted but the log
says that at least Ingo Molnar aims to mimic lmbench's lat_ctx, that
doesn't handle the pipe ends that way
(https://sourceforge.net/p/lmbench/code/HEAD/tree/trunk/lmbench2/src/lat_ctx.c)
Fix worker_thread() by always first feeding the write ends of the pipes
and then trying to read.
This roughly halves the context switches and runtime of pure
'perf bench sched pipe':
Performance counter stats for 'perf bench sched pipe':
1,005,770 context-switches
6.033448041 seconds time elapsed
0.423142000 seconds user
4.519829000 seconds sys
And the blocking reads do no longer dominate the analysis at the above
extreme:
Samples: 40K of event 'cycles:P', Event count (approx.): 14309364879
Overhead Command Shared Object Symbol
12.20% sched-pipe [kernel.kallsyms] [k] read_hpet
9.23% sched-pipe [kernel.kallsyms] [k] retbleed_untrain_ret
3.68% sched-pipe [kernel.kallsyms] [k] pipe_write
Signed-off-by: Dirk Gouders <dirk@gouders.net>
Acked-by: Ingo Molnar <mingo@kernel.org>
Link: https://lore.kernel.org/r/20250323140316.19027-2-dirk@gouders.net
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
|
||
|---|---|---|
| .. | ||
| Build | ||
| bench.h | ||
| breakpoint.c | ||
| epoll-ctl.c | ||
| epoll-wait.c | ||
| evlist-open-close.c | ||
| find-bit-bench.c | ||
| futex-hash.c | ||
| futex-lock-pi.c | ||
| futex-requeue.c | ||
| futex-wake-parallel.c | ||
| futex-wake.c | ||
| futex.h | ||
| inject-buildid.c | ||
| kallsyms-parse.c | ||
| mem-functions.c | ||
| mem-memcpy-arch.h | ||
| mem-memcpy-x86-64-asm-def.h | ||
| mem-memcpy-x86-64-asm.S | ||
| mem-memset-arch.h | ||
| mem-memset-x86-64-asm-def.h | ||
| mem-memset-x86-64-asm.S | ||
| numa.c | ||
| pmu-scan.c | ||
| sched-messaging.c | ||
| sched-pipe.c | ||
| sched-seccomp-notify.c | ||
| synthesize.c | ||
| syscall.c | ||
| uprobe.c | ||