Date: Mon, 16 Aug 2004 01:39:53 -0400 From: "Mark W. Krentel" <krentel@dreamscape.com> To: des@des.no, cvs-all@FreeBSD.org Cc: alc@FreeBSD.org Subject: Re: cvs commit: src/sys/vm vm_map.c vm_map.h Message-ID: <200408160539.i7G5drJE018193@blue.mwk.domain>
next in thread | raw e-mail | index | archive | help
> Great! I've been meaning to do this for a while, but the complexity > was a bit beyond me. This should greatly increase pipe performance, > by the way, because pipe buffers are allocated from a single vm_map, > which accumulates many more entries and much more fragmentation than > (probably) most other maps in the kernel. If you (or anyone else) have a real-world application that runs measurably better with the new vm_map_findspace() algorithm, I'd love to hear about it, especially if you have numbers. Probably on -hackers or something. I can write a one-page C program that mmap()'s many pages with holes between them that is a factor of 50-100 faster with the new algorithm where "many" is around 10,000. But it's a very artificial micro- benchmark. --Mark
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200408160539.i7G5drJE018193>