Messages in this thread | | | Subject | Re: Low Latency | Date | Sun, 02 Jul 2000 09:45:27 -0400 | From | Paul Barton-Davis <> |
| |
>Robert, Paul: test cases, please. Can you please tell me what system >activity causes the most problems? Is there a particular system >activity or application which will reliably kill scheduling latency? If
There is no specific thing in the experiences that many of us have collected. However, the closest to a "common" source of problems is anything involving disk i/o. Not necessarily disk i/o from the application in question, but anything (e.g. syslog). Also, this is true of both IDE and SCSI (though SCSI is better). For the most part, I can be running a SCHED_FIFO/mlockall process as the only user-driven runnable process on the system, and the combination of xosview, the X server (matrox g{2,4}00) and bits of disk i/o from various system activities will cause scheduling "misses". Turning off xosview can help, but it doesn't get rid of the problems.
Benno will have better clues on this front for you.
However, I imagine that Ingo and/or his previous patches will have some really good clues on where to look. Don't forget that the empirical evidence is that his patches fixed more or less every problem for 2.2-UP. So before you go running off to write a lot of code, it would be worth checking those out to see what specific problem areas he fixed. OK, Larry, "fixed" :)
>I can't reproduce it here will you be able to apply a truckload of >patches (against 2.4.0) so we can get to the bottom of it?
To the extent that I will be around this summer, I will be happy to try any patches if we can agree on a baseline kernel to start from (e.g. 2.4.0-testNN). However, I am away a *lot* between July 4th and September 5th, so it will be unwise to use me as the primary rapid response target.
I really appreciate your offer to help try to identify the trouble spots. Thanks.
I know that the computer music group at the Libre Software Meeting in Bordeaux next week extended an invitation to several linux-kernel members to talk about this stuff. I'd like to second that invitation, both formally (we plan to have a 30-60min session discussing strategies for handling the current low latency situation) and informally (as in: i'll be around).
--p
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |