[linux-audio-user] Problems with RP patch + hdsp + jackd at 96000

[Thomas Charbonnel]

Mario Lang wrote :
> Hi.
> 
> I recently patched my kernel with the RP patch by Ingo to get rid
> of the many xruns I was getting with my brand new Multiface (PCI).
> 
> THe result is somewhat strange, so I thought I'd post a summary here and see
> if someone else has had similar problems:
> 
> 1. jackd -d alsa -dhw:1 -p64
>    At 48kHz, jackd ran the whole night with SuperCollider connected
>    to some of its ports without a single xrun!  This is a massive improvement.
> 2. jackd -d alsa -dhw:1 -p64 -r96000
>    This used to work without the RP patch applied, it generated some
>    xruns, but it used to work.  Now, using the RP patch, jackd seem
>    to simply hang.  If I boot up the SuperCollider synthesis server,
>    it results in a "failed to start" message by SC.  Normal jack clients
>    like jack_lsp do still work, but jackd exits after about 10 seconds or
>    so, claiming its watchdog killed it.  Without -R, it runs without
>    dying, but no audio clients work.
> 
>    One indication that the problem is really originating from within
>    jackd is the fact that -v does not produce any of the load-statistic
>    output if used as argument to jackd.
> 
> Linux lexx 2.6.12-rt-v0.7.50-10 #1 Wed Jun 22 14:59:13 CEST 2005 i686 GNU/Linux
> 
> jackd version 0.99.0 tmpdir /dev/shm protocol 13
> 

Hi Mario,

I just did some tests with 2.6.12-RT-V0.7.49-01 and can somehow
reproduce this. In my case jack starts at 96kHz/p64 but stops really
fast being xrun-flooded, but I run a more recent version (jackd version
0.100.1 tmpdir /dev/shm protocol 15).

The culprit seems to be the 'Precise Pointer' at these settings.
I set it to be the default a couple of months ago.
I can run reliably @96kHz/p64 turning it off:
amixer -c X cset numid=22 0 (where X is your alsa card number)

BTW how do you setup irq priorities ?

Thomas