Date: Fri, 7 Mar 2014 07:32:36 -0700 From: Warner Losh <imp@bsdimp.com> To: Ian Lepore <ian@FreeBSD.org> Cc: freebsd-arm <freebsd-arm@FreeBSD.org>, Ruslan Bukin <br@FreeBSD.org> Subject: Re: ULE on ARM Message-ID: <3BE23B6A-900C-4104-A398-30D5B2A282DB@bsdimp.com> In-Reply-To: <1394202304.1149.373.camel@revolution.hippie.lan> References: <20140307141406.GA79223@machdep.com> <1394202304.1149.373.camel@revolution.hippie.lan>
next in thread | previous in thread | raw e-mail | index | archive | help
On Mar 7, 2014, at 7:25 AM, Ian Lepore <ian@FreeBSD.org> wrote: > On Fri, 2014-03-07 at 18:14 +0400, Ruslan Bukin wrote: >> I discovered just a couple of ARM kernel configs >> uses SCHED_ULE, but all other uses SCHED_4BSD >> >> any disadvantages to use ULE scheduler on ARM? >> or it is just because of historical reasons? >> >> I enabled ULE on Freescale Vybrid and running >> it for a long time just fine. >> >> according to my subjective impressions ULE >> works better on ARM in sound applications >> >> -Ruslan > > The widespread advice from a few years ago was that ULE was better for > SMP and 4BSD was better for UP. I don't know whether that's still true > (or whether it was ever true). I do know that there are fewer responses > on mailing lists of "try switching the scheduler to 4BSD" as a way of > fixing problems these days. I switched imx6 to ULE when adding SMP > support for it. It all depends on the workload. 4BSD is better for some SMP workloads, while ULE is better for others. But as a general rule, Ian is right: 4BSD tends to be better at UP interactive workloads, while ULE tends to be better at MP work loads that have a larger compute element to them (complex transactions). Warner
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3BE23B6A-900C-4104-A398-30D5B2A282DB>