Still problems with ULE
I gave ULE another try just now, following your recent commits, and I'm seeing even worse problems: At boot time when the X server is loading, disk activity occurs briefly about once every 2 seconds; the mouse is active briefly at the same time, and nothing much else happens for about a minute until the entire system deadlocks. Kris msg52181/pgp0.pgp Description: PGP signature
Re: Still problems with ULE
On Mon, 10 Feb 2003, Kris Kennaway wrote: I gave ULE another try just now, following your recent commits, and I'm seeing even worse problems: At boot time when the X server is loading, disk activity occurs briefly about once every 2 seconds; the mouse is active briefly at the same time, and nothing much else happens for about a minute until the entire system deadlocks. Very weird. Is this on UP or SMP? I'm still working on a couple of issues with ule. I think I am very happy with the dynamic priority selection now but in the process the slice size selection got kinda dirty. Its pretty much bug for bug compatible with the old scheduler's context switching decisions but the errors there are more serious with this design. I also think the max slice size is way too high now. I was experimenting with that and I accidentally checked it in. I think I know what the remaining problems are. I'll make a post on current@ when they're all sorted out. Thanks for testing! Cheers, Jeff To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-current in the body of the message
Re: Still problems with ULE
On 2003-02-10 20:36 +, Jeff Roberson wrote: On Mon, 10 Feb 2003, Kris Kennaway wrote: I gave ULE another try just now, following your recent commits, and I'm seeing even worse problems: At boot time when the X server is loading, disk activity occurs briefly about once every 2 seconds; the mouse is active briefly at the same time, and nothing much else happens for about a minute until the entire system deadlocks. Very weird. Is this on UP or SMP? I'm still working on a couple of issues with ule. I think I am very happy with the dynamic priority selection now but in the process the slice size selection got kinda dirty. Its pretty much bug for bug compatible with the old scheduler's context switching decisions but the errors there are more serious with this design. I also think the max slice size is way too high now. I was experimenting with that and I accidentally checked it in. I'm seeing some of the same with sources from about 90 minutes ago. UP system, loading X is a pain. Interactivity is somewhat alright up until I start a compile, at which point the system goes down on it's knees. It hasn't completely deadlocked (yet?), but it's pretty much unusable if anything intensive is happening. -- Munish Chopra To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-current in the body of the message
Re: Still problems with ULE
On Mon, Feb 10, 2003 at 08:36:50PM -0500, Jeff Roberson wrote: Very weird. Is this on UP or SMP? This is on UP. I can still break into DDB, so let me know if you want me to run console tests (no serial console though). Kris msg52192/pgp0.pgp Description: PGP signature
