Messages in this thread | | | Date | Wed, 26 Jun 2013 11:11:40 -0500 | From | Nathan Zimmer <> | Subject | Re: [RFC 2/2] x86_64, mm: Reinsert the absent memory |
| |
On 06/26/2013 10:58 AM, Ingo Molnar wrote: > * Nathan Zimmer <nzimmer@sgi.com> wrote: > >> On 06/26/2013 10:12 AM, Dave Hansen wrote: >>> On 06/26/2013 07:49 AM, Nathan Zimmer wrote: >>>> My guess it is the NMIs overwhelming the system but I have not found a good way >>>> to profile perf gone wild so it is only a guess. >>> I've got an 80-core system and the symptoms sound similar to perf issues >>> I'm seeing. Dropping the sample rate helped for me, and there's a patch >>> in -tip at the moment to do it automatically: >>> >>>> http://git.kernel.org/cgit/linux/kernel/git/tip/tip.git/commit/?id=14c63f17b1fde5a575a28e96547a22b451c71fb5 >>> Looks like you've also found that debugging NMIs is fun. :) >> More fun then one person can handle. >> I'll have to take this patch for a spin. > Please try latest -tip:master, there's a series of patches from Dave that > address this problem category. > > Does it only happen with -g, or with regular perf record and/or with perf > top as well? > > Thanks, > > Ingo
It happens in all those cases, but I notice it most when doing a perf record -g -a I'll try the tip series and see how it compares.
Thanks Nate
| |