[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Xorg 6.8.1 and SCHED_ULE vs. SCHED_4BSD



Godwin Stewart napisa?(a):
> On Sat, 26 Feb 2005 21:49:00 +0100, Michael Nottebrock
> <michaelnottebrock_(_at_)_gmx_(_dot_)_net> wrote:
> 
> 
>>>I've been using that for a long time now, since Xorg 6.8.1 breaks vt-
>>>switching for me.
> 
> One of the things I've been doing is to record some of my old cassettes
> (you know, those old plastic things with 2 holes and a tape inside :) onto
> CD. Applying a FFT filter to 50 minutes of audio takes between 10 and 15
> minutes on this machine (P-III/550, 384MB) depending on the complexity of
> the filter. During this time, with SCHED_ULE and PREEMTION, the machine is
> unusable. It freezes hard for periods of 10-12 seconds and then when it
> unfreezes (while doing disk i/o apparently) the keys you typed turn up in
> the wrong order.

I currently run 5.4-PRE with ULE and PREEMPTION, on a similar machine 
(pIII-733 192RAM), also 6.8.1, and i do have to say that ULE has 
improved responsiveness /alongside kern.hz=800/ incredibly, with 
none-whatsoever speed degradation (actually my compilations seem to run 
faster, although that is only a mere hunch not yet backed up by any 
benchmarking).

The only time i might encounter problems, is with the lack of ram, and a 
lot of disk swap usage, or during untarring of big distfiles (yet it is 
still giving me better response than with 4BSD, which was utterly 
terrible :)

I honestly would have to say, great job on the ULE, if mere fixing of 
the possible disk i/o lock ups were to be commited its much better than 
4BSD.

> 
> However, now that I've reverted to SCHED_4BSD, the machine remains
> perfectly snappy while performing the FFT filter, which doesn't happen
> perceptibly slower.

Perhaphs you can try with different kern.hz settings?

--
Mateusz J?drasik < imachine_(_at_)_toya_(_dot_)_net_(_dot_)_pl >