On Tue, Apr 5, 2016 at 10:26 AM, Dilip Kumar <dilipbal...@gmail.com> wrote:
> > On Mon, Apr 4, 2016 at 2:28 PM, Andres Freund <and...@anarazel.de> wrote: > >> Hm, interesting. I suspect that's because of the missing backoff in my >> experimental patch. If you apply the attached patch ontop of that >> (requires infrastructure from pinunpin), how does performance develop? >> > > I have applied this patch also, but still results are same, I mean around > 550,000 with 64 threads and 650,000 with 128 client with lot of > fluctuations.. > > *128 client > **(head+**0001-WIP-Avoid-the-use-of-a-separate-spinlock-to-protect > +pinunpin-cas-9+backoff)* > > run1 645769 > run2 643161 > run3 *285546* > run4 *289421* > run5 630772 > run6 *284363* > Could the reason be that we're increasing concurrency for LWLock state atomic variable by placing queue spinlock there? But I wonder why this could happen during "pgbench -S", because it doesn't seem to have high traffic of exclusive LWLocks. ------ Alexander Korotkov Postgres Professional: http://www.postgrespro.com The Russian Postgres Company