Re: Linux-next Alsa hda_intel events/0 high CPU usage, bisected

[Takashi Iwai <tiwai@suse.de>]

At Tue, 10 Jun 2008 05:20:49 -0700 (PDT),
Benjamin Kidwell wrote:
> 
> > --- Takashi Iwai <> wrote:
> >Try to run
> >
> >	# echo 32 > /sys/modules/snd_hda_intel/parameters/bdl_pos_adj
> >and test again.
> 
> Thanks, I think that value must be just about perfect! Tested as follows:
> 
> (kill gdm)
> #modprobe -r snd_hda_intel
> #modprobe snd_hda_intel
> #echo 32 > /sys/module/snd_hda_intel/bdl_pos_adj
> (restart gdm, start playing sound)
> 
> That caused events/0 CPU% to return to almost nothing and the XXX hda delay
> output did not trigger. I assume this means the latency has disappeared. For
> comparison, I repeated the above processand returned bdl_pos_adj to 1, and the
> delay printks reappeared and events/0 started showing significant cpu usage
> again. Another reset of the driver and I reset bdl_pos_adj to 32 -- no delay
> printks, almost no events/0 activity apparent.
> 
> So setting bdl_pos_adj to 32 with all new patches applied makes everything
> perfect so far as I can tell. Thanks for the fix-ups for my buggy hardware,
> and for reminding me to use reply and Cc: correctly. Let me know if there's
> anything else I should test.

Thanks for quick testing!

The above value 32 is likely the size of FIFO.  I guess ATI chip
reports the position with FIFO offset while ICH reports smaller delay.
I'll create one more patch to assign this value for ATI chipset.

BTW, my latest test patches can be found on the following git tree:
    git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound-unstable-2.6.git

These are not for 2.6.27, but some of them should go to the linux-next
tree now.


Takashi