[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <1354898530.2707.21.camel@bwh-desktop.uk.solarflarecom.com>
Date: Fri, 7 Dec 2012 16:42:10 +0000
From: Ben Hutchings <bhutchings@...arflare.com>
To: Joe Perches <joe@...ches.com>
CC: Eric Dumazet <eric.dumazet@...il.com>,
Tom Herbert <therbert@...gle.com>,
David Miller <davem@...emloft.net>,
netdev <netdev@...r.kernel.org>
Subject: Re: [RFC PATCH] dynamic_queue_limit.h: Make the struct
___cacheline_aligned_on_smp
On Fri, 2012-12-07 at 08:29 -0800, Joe Perches wrote:
> On Fri, 2012-12-07 at 08:19 -0800, Eric Dumazet wrote:
> > On Fri, 2012-12-07 at 08:05 -0800, Joe Perches wrote:
> >
> > > So it seemed somewhat sensible to make the
> > > entire struct in a single cacheline.
> >
> > Any layout change in an object used in network fast path need a complete
> > performance study.
> >
> > Even if you provide such a study, we'll need to reproduce your numbers
> > here.
> >
> > BQL/DQL is not on our radars, spending two cache lines on a critical
> > object is fine.
>
> Well Maybe Tom can provide some information as to why
> the limit variable was cacheline_aligned_in_smp and not
> the struct.
>
> I didn't find any discussion about it.
Structure alignment has to be at least the maximum of each member's
alignment, so the struct *is* effectively cacheline_aligned_in_smp.
Ben.
--
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Powered by blists - more mailing lists