Date: Tue, 19 Apr 2005 14:29:36 +0200 (CEST) From: Sten Spans <sten@blinkenlights.nl> To: Pyun YongHyeon <yongari@rndsoft.co.kr> Cc: freebsd-sparc64@freebsd.org Subject: Re: em(4) patch Message-ID: <Pine.SOC.4.61.0504191346380.11438@tea.blinkenlights.nl> In-Reply-To: <20050419043359.GA4941@michelle.rndsoft.co.kr> References: <20050414092608.GB2855@michelle.rndsoft.co.kr> <20050419043359.GA4941@michelle.rndsoft.co.kr>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tue, 19 Apr 2005, Pyun YongHyeon wrote: > On Thu, Apr 14, 2005 at 06:26:08PM +0900, To freebsd-sparc64@freebsd.org wrote: > > > > I have a initial patch that make em(4) work on sparc64. > > The patch seems to work but it may have uncovered bugs in it. > > Here is updated patch. > http://www.kr.freebsd.org/~yongari/em.patch.0419 > > The new patch includes jmg's suggestion and supports JUMBO frames. > Though OpenBSD's aligment code works well I noticed it suffers > from poor performance. So I used simple mbuf prepend method which > is slightly faster than that of OpenBSD's code. Of course, best > performance would be obtained when JUMBO frame is not used at all. :-( Great, thanks for looking into this. I've tested the patch on an alpha (ds10) running 5.4-RC3, with dev/em from current. Some netperf results: Recv Send Send Socket Socket Message Elapsed Size Size Size Time Throughput bytes bytes bytes secs. 10^6bits/sec openbsd code in: 87380 32768 32768 10.01 415.38 openbsd code out: 65536 16384 16384 10.00 403.87 fbsd patch in: 87380 32768 32768 10.01 409.02 fbsd patch out: 65536 16384 16384 10.00 464.78 Aka roughly comparable performance, and no crash :) -- Sten Spans "There is a crack in everything, that's how the light gets in." Leonard Cohen - Anthem
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.SOC.4.61.0504191346380.11438>