By the way, I'm seeing more and more bug reports on nvnews about the 256 series causing the cpu spikes I reported there.
Back to top

Yes, my Gentoo,Funtoo~ system runs with the old nvidia v195 without problems, but for the newer nvidia driver I cannot say yet.
Back to top

eh, at least the nV devs are pretty responsive and the releases are pretty often
Back to top

damentz, just one question:

Since you use Zen-patches why do you leave cfq scheduler as default instead of bfq ? For stability sake or what?

And is it possible to switch to bfq in real time without rebooting? I mean by addressing 'echo cfq > /sys/block/sda/queue/scheduler' ?

Thank you.
Back to top

I got a report that KDE4 loaded slowly for a new kernel I released using BFQ as default. That must have been the issue, so I changed the default elevator to CFQ.

If you can debunk that myth, then I may just switch it back since the reporter didn't confirm that my kernels afterward fixed the problem.

And yes, you can update the I/O scheduler on-the-fly by echoing the scheduler you want to use into /sys/block/sda/queue/scheduler.

You can also cat the scheduler file to see what you are using and what's available:


Back to top

Thank you very much.
Back to top

It's great that you keep updating the CK BFS, but why bother if it's never usable? :(

Back to top

I'm updating BFS in the zen-kernel sources for other people to use, this doesn't mean I will use it.

There are still problems with BFS that do not make it appropriate for a distro kernel configuration. The kernel maintainer on Yoper also decided to switch back to CFS after his community reported that some things stopped working.

I hope that the changes I made recently to the CFS scheduler make it perform better for your workloads; latencies should be reduced and your desktop should perform better in general.
Back to top

That's ok, don't use by default, but give the users the choice to use if they want without them needing to recompile and install another custom kernel.
Back to top

There's probably a good reason why we can't choose CFS and BFS at runtime, most likely runtime overhead or needless complexity. Whatever the reason, Con chose not to go down this path when developing BFS to be configureable alongside CFS.

Until someone finds a good way to do this, you'll need to recompile the kernel.
Back to top