Messages in this thread | | | Date | Tue, 23 Jun 2015 16:36:52 +0200 | From | Ingo Molnar <> | Subject | Re: [PATCH v2 2/4] sched:Consider imbalance_pct when comparing loads in numa_has_capacity |
| |
* Srikar Dronamraju <srikar@linux.vnet.ibm.com> wrote:
> * Ingo Molnar <mingo@kernel.org> [2015-06-23 10:10:39]: > > > Please let me know if there are any better ways to observe the > > > spread. [...] > > > > There are. I see you are using prehistoric tooling, but see the various NUMA > > convergence latency measurement utilities in 'perf bench numa': > > > > vega:~> cat numa01-THREAD_ALLOC > > > > perf bench numa mem --no-data_rand_walk -p 2 -t 16 -G 0 -P 0 -T 192 -l 1000 -zZ0c $@ > > > > You can generate very flexible setups of NUMA access patterns, and measure their > > behavior accurately. > > > > It's all so much more capable and more flexible than autonumabench ... > > Okay, thanks for the hint, I will try this out in future. > > > > > Also, when you are trying to report numbers for multiple runs, please use > > something like: > > > > perf stat --null --repeat 3 ... > > > > This will run the workload 3 times (doing only time measurement) and report the > > stddev in a human readable form. > > > > Thanks again for this hint. Wouldnt system time/ user time also matter?
Yeah, would be nice to add stime/utime output to 'perf stat', so that it's an easy replacement for /usr/bin/time.
I've Cc:-ed perf folks who might be able to help out.
Thanks,
Ingo
| |