[snd_hda] Choppiness under high cpu - STABLE

[Ariff Abdullah]

On Mon, 4 Jun 2007 14:06:30 +1000
Norberto Meijome <freebsd at meijome.net> wrote:
> On Tue, 29 May 2007 16:39:49 +1000
> Norberto Meijome <freebsd at meijome.net> wrote:
> 
> > > 
> > > Anyway, you can try adjusting buffer size
> > > with "hint.pcm.0.buffersize" . Start with 4096, 8192, 16384,
> > > etc.. Use either kenv(8) and reload, or slap it into
> > > /boot/device.hints and reboot.  
> > 
> > Thanks, I'll have a play and inform back :)
> 
> ok, so I've had some time to test (apologies for the delay)
> 
> SYSTEM
> updated my kernel to latest -STABLE :
> FreeBSD ayiin.octantis.com.au 6.2-STABLE FreeBSD 6.2-STABLE #4: Mon
> Jun  4 12:45:01 EST 2007    
> root at ayiin.octantis.com.au:/usr/obj/usr/src/sys/AYIIN  i386
> 
> (EST == AU EST, GMT +10)
> 
> I am testing with snd_hda in this tree. World was built from sources
> of May 30th.
> 
> no kernel hints tweaking.
> ( not sure if this is related, but sysctl shows:
> 
> hw.snd.pcm0.buffersize: 16384
> )
> 
> TOOLS:
>  installed sysutils/stress to have a somewhat more reliable method
>  of test, rather than buildworld. It seems the issue is not
>  necessarily high user CPU usage, but system cpu usage , which is
>  triggered by stress' -m switch (more specifically, virtual memory
>  stress testing). I tested with xmms-esound, esd running from my X
>  session, playing a (Smashing Pumkins :D) CD on my CD drive.
> 
> Presents problem.
> -------
> [betom at ayiin] [Mon Jun  4 13:39:59 2007]
> /usr/ports/sysutils/stress
> $ stress -m 1 -c 1
> stress: info: [8731] dispatching hogs: 1 cpu, 0 io, 1 vm, 0 hdd
> ^C
> -------
> 
> No Problems at all:
> ------
> [betom at ayiin] [Mon Jun  4 13:40:15 2007]
> /usr/ports/sysutils/stress
> $ stress -c 1
> stress: info: [8735] dispatching hogs: 1 cpu, 0 io, 0 vm, 0 hdd
> ^C
> --------
> 
> Problems again.
> --------
> [betom at ayiin] [Mon Jun  4 13:40:26 2007]
> /usr/ports/sysutils/stress
> $ stress -m 1
> stress: info: [8738] dispatching hogs: 0 cpu, 0 io, 1 vm, 0 hdd
> ^C
> ----
> 
> Changing buffersize makes no difference at all (tried up to 24K).
> 
> when I loaded the binary blob provided in your site, I can run up to
> stress -m 3 with no problems at all. -m 4 shows some issues if i
> start heavily using the system. -m 5 is somewhat similar to -m 1
> with the non-low-latency driver.
> 
> Please let me know  if there's anything else u'd like me to try.
> 
> Best,
> 

Have you tried with 4096 ? Note: this should go into
hint.pcm.0.buffersize kernel hint, not sysctl variable. After setting
it (using kenv), you must reload the sound module or reboot if it
going through /boot/device.hints. The sysctl buffersize is there
purely as indicator.


--
Ariff Abdullah
FreeBSD

... Recording in stereo is obviously too advanced
    and confusing for us idiot ***** users :P ........
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 187 bytes
Desc: not available
Url : http://lists.freebsd.org/pipermail/freebsd-multimedia/attachments/20070604/b0970344/attachment.pgp