Return to BSD News archive
Path: euryale.cc.adfa.oz.au!newshost.anu.edu.au!harbinger.cc.monash.edu.au!news.rmit.EDU.AU!news.unimelb.EDU.AU!munnari.OZ.AU!spool.mu.edu!howland.reston.ans.net!swrinde!newsfeed.internetmci.com!zdc!zdc-e!szdc-e!news From: "John S. Dyson" <toor@dyson.iquest.net> Newsgroups: comp.os.linux.networking,comp,comp.unix.bsd.freebsd.misc Subject: Re: TCP latency Date: Thu, 18 Jul 1996 10:54:13 -0500 Organization: John S. Dyson's home machine Lines: 23 Message-ID: <31EE5E25.2781E494@dyson.iquest.net> References: <4sbhrq$msh@enomem.lkg.dec.com> <31E995B6.573A3461@lambert.org> <4schqh$ssa@panix2.panix.com> <4sl2c4$jf5@news.swan.ac.uk> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Mailer: Mozilla 3.0b5aGold (X11; I; FreeBSD 2.2-CURRENT i386) Xref: euryale.cc.adfa.oz.au comp.os.linux.networking:45697 comp.unix.bsd.freebsd.misc:23934 Alan Cox wrote: > > In article <4schqh$ssa@panix2.panix.com> tls@panix.com (Thor Lancelot Simon) writes: > >a few months ago. He indicated that at least one version of Mach had done > >exactly that on the VAX, and that it turned out to be a win iff you wanted > >to move more than about 8K. Of course, that number is so utterly hardware > >dependent... > > On the PC its quite painful since you don't have a handy "invalidate one > TLB" type instruction. On a multiprocessor box it can get more involved as > well. If you can find an mmap() semantic then it is a win because you also > remove the cost of the TLB and page table change hits as the page is already > mapped and stays mapped shared with the socket. The problem is much more > what does mmap on a socket mean > On (4,5,6)86 processors there is an invalidate a single page table entry instruction. If used properly, it helps alot. Initially, the addition of the use of that instruction didn't help. Simply required restructuring code to be able to utilize it. Works great!!! John