From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jack Steiner Date: Tue, 05 Apr 2005 03:02:12 +0000 Subject: Re: per_cpu_pagesets degrades MPI performance Message-Id: <20050405030212.GA10812@sgi.com> List-Id: References: <20050404192827.GA15142@sgi.com> <4251DE87.10002@yahoo.com.au> In-Reply-To: <4251DE87.10002@yahoo.com.au> MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: Nick Piggin Cc: akpm@osdl.org, hugh@veritas.com, linux-mm@kvack.org, linux-ia64@vger.kernel.org On Tue, Apr 05, 2005 at 10:40:39AM +1000, Nick Piggin wrote: > Jack Steiner wrote: > > [snip nice detective work] > > >Has anyone else seen this problem? I am considering adding > >a config option to allow a site to control the batch size > >used for per_cpu_pagesets. Are there other ideas that should > >be pursued? > > > > What about using a non power of 2 for the batch? Like 5. > If that helps, then we can make a patch to clamp it to a > good value. At a guess I'd say a power of 2 +/- 1 might be > the way to go. > > Nick > > -- > SUSE Labs, Novell Inc. Sounds reasonable. I'll give it a try & see what happens. -- Thanks Jack Steiner (steiner@sgi.com) 651-683-5302 Principal Engineer SGI - Silicon Graphics, Inc.