Re: [Jack-Devel] L_ProAudio: A Linux Proaudio FrameWork with improved Wine/kernel(rt) support
Hi, i replied to jeremy off list, so i thought i would address this on
list, as well. (for those interested)
> * how do the above things influence Linux audio performance?
2 words: OS jitter.
> * what is delayed acpi?
> * what is indirect fan control?
> * what is tame conntrack?
> * what is stable_page_write?
delaying acpi, using an indirect patch to fan control and making sure
conntrack (netfilter) doesn't cause interference, ensures that we
don't have these kernel-level components causing inconsistent and
undesiered behavior (that can produce things like xruns /
interference.
Stable_page_write is fairly incompatible with wine and causes huse
latencies with [poll schedule timeouts], these can be very higher
numbers like 80 to 200+ms - which is TERRIBLE. Google, Oracle,
samsung, Bao Mao and others have chosen to revert this commit -
because they (in all of their glory) couldn't work around the issue ~
and neither can I.. That being said, this is a somewhat temparary
patch (although it has been is use by me, for quite some time now). I
know that a solution is just down the pipe, but may not surface for a
while yet. revert_stable_page_write.patch in my patchset is CRITICAL
for Wine, for now.
> * why 3.6? Aren't there still some issues with it (softirq related stuff)?
I was plagued by that, on one system ~ which appears to be fixed. ~
this led to me releasing L_ProAudio, after cleaning up and rebasing my
patchsets against upstream linux-rt and upstream wine.
so i guess the softirq problems, might be a blessing in disguise. (ie:
i have been using _most_ of L_Proaudio for months).
cheerz
1357079646.29858_0.ltw:2,a <CAOcfFMzx0FD+wKd=vykPAO2U3kcaH+8R_+wuD77A+3YTodixzw at mail dot gmail dot com>