Date: Wed, 5 Mar 1997 15:43:51 -0500 From: "David S. Miller" <davem@jenolan.rutgers.edu> To: jj@sunsite.ms.mff.cuni.cz Cc: avalon@coombs.anu.edu.au, mingo@pc5829.hil.siemens.at, wong@rogerswave.ca, alan@cymru.net, imb@scgt.oz.au, dg@root.com, netdev@roxanne.nuclecu.unam.mx, hackers@freebsd.org Subject: Re: ok, final sockhash changes, new diff Message-ID: <199703052043.PAA14523@jenolan.caipgeneral> In-Reply-To: <199703052045.VAA00601@sunsite.ms.mff.cuni.cz> (message from Jakub Jelinek on Wed, 5 Mar 1997 21:45:07 %2B0100 (MET))
next in thread | previous in thread | raw e-mail | index | archive | help
From: Jakub Jelinek <jj@sunsite.ms.mff.cuni.cz> Date: Wed, 5 Mar 1997 21:45:07 +0100 (MET) On the other side you can put there three four different routines for different hash table sizes and thus avoid that dead cycle. Or we can use code patching (as we already do on the sparc for other stuff). I'm seriously considering making the lookup routines for TCP be per-architecture so you do it all in assembly for maximum performance. The lookup code is so hot in the profiling logs that only the csum_partial_copy() and memcpy() code dwarf it. ---------------------------------------------//// Yow! 11.26 MB/s remote host TCP bandwidth & //// 199 usec remote TCP latency over 100Mb/s //// ethernet. Beat that! //// -----------------------------------------////__________ o David S. Miller, davem@caip.rutgers.edu /_____________/ / // /_/ ><
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199703052043.PAA14523>