From mboxrd@z Thu Jan 1 00:00:00 1970 From: Con Kolivas Date: Sun, 31 Jul 2005 04:35:06 +0000 Subject: Re: [sched, patch] better wake-balancing, #2 Message-Id: <200507311435.09225.kernel@kolivas.org> MIME-Version: 1 Content-Type: multipart/mixed; boundary="nextPart6421577.OHCWCaKClm" List-Id: References: <200507301929_MC3-1-A601-D4C2@compuserve.com> In-Reply-To: <200507301929_MC3-1-A601-D4C2@compuserve.com> To: linux-kernel@vger.kernel.org Cc: Chuck Ebbert <76306.1226@compuserve.com>, Ingo Molnar , "Chen, Kenneth W" , Andrew Morton , Nick Piggin , linux-ia64 --nextPart6421577.OHCWCaKClm Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Content-Disposition: inline On Sun, 31 Jul 2005 09:26, Chuck Ebbert wrote: > On Fri, 29 Jul 2005 at 17:02:07 +0200, Ingo Molnar wrote: > > do wakeup-balancing only if the wakeup-CPU is idle. > > > > this prevents excessive wakeup-balancing while the system is highly > > loaded, but helps spread out the workload on partly idle systems. > > I tested this with Volanomark on dual-processor PII Xeon -- the > results were very bad: > > Before: 5863 messages per second > After: 5569 messages per second Can you check schedstats or otherwise to find if volanomark uses=20 sched_yield() ? When last this benchmark came up, it appeared that no jvm=20 used futexes and left locking to yielding. We really should find out if tha= t=20 is the case before trying to optimise for this benchmark. Cheers, Con --nextPart6421577.OHCWCaKClm Content-Type: application/pgp-signature -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.1 (GNU/Linux) iD8DBQBC7FT9ZUg7+tp6mRURAk7zAJ4vwv1n0picU3J4ODTzTg6IUBT6ogCfaby4 FN2C76fkrGxucZeWjfMY1FY= =sKKn -----END PGP SIGNATURE----- --nextPart6421577.OHCWCaKClm--