On Wed, Jun 08, 2005 at 11:24:32AM -0400, Jeff Roberson wrote: > > On Wed, 8 Jun 2005, Jeff Roberson wrote: > > > On Wed, 8 Jun 2005, Jean-Yves Lefort wrote: > > > > > On Wed, 8 Jun 2005 09:57:15 -0400 (EDT) > > > Jeff Roberson <jroberson_at_chesapeake.net> wrote: > > > > > > > On Wed, 8 Jun 2005, Jean-Yves Lefort wrote: > > > > > > > > > On Wed, 8 Jun 2005 12:48:44 +0200 > > > > > Simon Barner <barner_at_FreeBSD.org> wrote: > > > > > > > > > > > Jeff Roberson wrote: > > > > > > > I have a patch that should greatly improve the sound skipping problems > > > > > > > people have under heavy io load. Several people sent me traces that > > > > > > > showed the buf daemon running for hundreds of milliseconds with Giant > > > > > > > held, which can hold up the pcm code. The patch is available at: > > > > > > > > > > > > > > http://www.chesapeake.net/~jroberson/flushbuf.diff > > > > > > > > > > > > [...] > > > > > > > > > > > > Can I use the patch on 5.4-STABLE? It applies cleanly (with offset, > > > > > > though), but I prefer to ask before doing useless tests. > > > > > > > > > > I've just tested it on 5.4-RELEASE, and it causes the system to hang > > > > > after a few seconds of heavy disk I/O. > > > > > > > > I have updated the patch to partially revert some earlier behavior. The > > > > hang didn't happen on my machine because I have more memory, but I'm > > > > confident of the cause. I updated it in place, feel free to try it on > > > > current or 5.x. > > > > > > This one works, and completely removes the sound jitter that used to > > > appear under heavy disk I/O (eg. untarring a 50 megs tarball). > > > > > > Congratulations! > > > > Thanks. I'd like to get some more experience with it under extreme load > > conditions, but I do eventually plan to MFC. > > I made one more slight change to the patch which should reduce the number > of times we needlessly scan the list. Can you make sure it does not cause > any problems for you? > Wow! It helps a lot. Even if maestro3(4) does not support channel_setblocksize I couldn't notice audio stuttering while extracting mozilla source. I noticed a few sound skipping when a command "rm -rf mozilla" was in progress. But I think systems that have audio hardwares with channel_setblocksize would have worked better than maestro(3). Thanks a lot! -- Regards, Pyun YongHyeon http://www.kr.freebsd.org/~yongari | yongari_at_freebsd.orgReceived on Thu Jun 09 2005 - 07:06:29 UTC
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:38:36 UTC