Date: Tue, 25 Mar 2003 17:54:55 -0800 From: "Sam Leffler" <sam@errno.com> To: "Mike Silbersack" <silby@silby.com>, "Doug Ambrisko" <ambrisko@ambrisko.com> Cc: cvs-all@FreeBSD.org Subject: Re: cvs commit: src/sys/conf options src/sys/netinet ip_output.c Message-ID: <150f01c2f33a$bb131d40$52557f42@errno.com> References: <200303252344.h2PNip18098878@www.ambrisko.com> <20030325194313.F458@odysseus.silby.com>
next in thread | previous in thread | raw e-mail | index | archive | help
> I've been pondering where the tradeoff between avoiding memory copies and
> doing excessive scatter / gather DMA lies. Perhaps we should be
> defragmenting any chain over a certain amount of length, no matter the
> limit imposed by the card. This sounds like a Terry question. :)
I hit this in fast ipsec. I do "agressive coalescing" when creating
writable mbuf chains on output. It's a big win for various things. Full
results will be in the paper I'm writing for bsdcon.
Sam
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?150f01c2f33a$bb131d40$52557f42>