[proaudio] Fw: Re: Subscription probs & rt-sources

[ Thread Index | Date Index | More lists.tuxfamily.org/proaudio Archives ]




Daniel Santos wrote:
Hello, either my email server is screwing up again or my attempt to subscribe to proaudio-request@xxxxxxxxxxxxxxxxxxx didn't work, so please reply-all or CC me on responses (and evermind, please subscribe me if you can).

I've subscribed you. So welcome to our list finally;)

I'm expecting to acquire a digiface by the end of next week, so I'm trying to get this stuff up and running. There are some HOWTOs that are in need of update (Gentoo/Jack HOWTO, etc.) and I figure that's something I can do after I'm done "learning the lessons" that need to go into it.

A. We should maybe remove the rt-sources-2.6.25.8-r7 ebuild. It would appear that Ingo Molnar (the rt kernel maintainer) has moved all of the 2.6.25.x rt patches to the "older" directory and has not replaced then with a newer version (http://www.kernel.org/pub/linux/kernel/projects/rt/). This has broken the ebuild its self, because it fails to download the patch. I would say to fix it, but I can't even get the thing to compile and an "alexMK" from the #lad/#ardour irc claims that he did tests and had lower latencies with the unpatched 2.6.25-x kernel than the RT one. This is the errors I got attempting to compile 2.6.25.8-rt7 (after fixing the download URL in the ebuild):

      CC      kernel/sched.o
    kernel/sched.c: In function 'sched_init':
    kernel/sched.c:7702: error: implicit declaration of function
    'global_rt_runtime'
    make[1]: *** [kernel/sched.o] Error 1
    make: *** [kernel] Error 2

hav a look at this thread


quote from http://www.nabble.com/2.6.25.8-rt7-td18081945.html
I can avoid this failure, when I disable the "Group scheduling for
SCHED_RR/FIFO" menu entry. Don't know the impact.
/qoute

B. I've updated the 2.6.24 rt ebuild to 2.6.24.7-r17 (attached). It was almost straight renaming it except that I changed the download URL to grab it from kernel.org (or the new patch can be uploaded to download.tuxfamily.org and the ebuild just renamed). I have thus far compiled it without the fbsplash and also ran just src_unpack with fbsplash to verify that the patch didn't fail or have any fuzzies, but I haven't attempted to compile with fbsplash and I haven't tested the -fbsplash just yet either.

I'll will add an updated ebuild later, thx

Greetz Frieder

C. I'm totally new to this DAW stuff on linux, so forgive me if I'm stating the obvious. I'm guessing you guys are aware of the removal of the preempting the big kernel lock in 2.6.26 (http://kerneltrap.org/Linux/Removing_the_Big_Kernel_Lock). If I understand the implications of this correctly, it will mean that the 2.6.26 kernel and beyond (for some time) will not be suitable for low-latency applications and mingo is proposing a new "kill-BKL" project that will probably be unstable for some time as well.

As a geek, I admit I'm slightly annoyed that the audio hardware (from what I'm getting) seems to lack sufficient hardware buffers to wait a few goddam milliseconds for it's interrupts to get serviced -- at least, I presume that's why such a low latency kernel is needed, or maybe I just lack appreciation of the levels of throughput required to pull off reading and writing 24 tracks of digital audio at the same time.




Mail converted by MHonArc 2.6.19+ http://listengine.tuxfamily.org/