Date: Wed, 3 Oct 2018 18:16:11 +1300 From: Thomas Munro <munro@ip9.org> To: allanjude@freebsd.org Cc: freebsd-hackers@freebsd.org Subject: Re: Regression when trying to replace poll() with kqueue() Message-ID: <CADLWmXXTdVa_QuY0ShsVAdde%2BT6QANUznTNDqYDoLhsa704=NQ@mail.gmail.com> In-Reply-To: <09f0dce2-7899-b839-e70e-79be43a0fa6b@freebsd.org> References: <CADLWmXXcdbL6wyLUktGzp=41zmbRjxw30FU=Ait-jfd8NcQSyQ@mail.gmail.com> <09f0dce2-7899-b839-e70e-79be43a0fa6b@freebsd.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, 3 Oct 2018 at 18:01, Allan Jude <allanjude@freebsd.org> wrote: > I have started to look into this a bit. I have not really gotten > anywhere yet, but I have produced a graph comparing the performance of > vanilla postgres vs your patch. > > https://imgur.com/a/gKycGxW > > They scale identically up to the 20 threads of hardware on my test > machine, and then kqueue falls off much more quickly. Great news, thanks for looking at this.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CADLWmXXTdVa_QuY0ShsVAdde%2BT6QANUznTNDqYDoLhsa704=NQ>