Re: [Openvortex-dev] alsa 0.9.4 + au8830 = sound is too fast

[Jeff Muizelaar]

On Wed, 18 Jun 2003, Callum Lerwick wrote:

> However I once tried stress testing by running many copies of XMMS at
> once, I think I got to 4 and the machine hosed up really bad. Haven't
> yet dared to repeat this test again, really should sometime. I tried the
> same thing with the OSS drivers and eight XMMS's played for hours with
> no problems.

The alsa driver should be able to take 4 no problem, if you bring things
up too high (probably >8) things will get messy because we start
overwriting previous routes. This is a known bug, however I don't have a
nice fix for it yet.

How many streams can people have playing under Windows? I believe I was
only able to get 8 in NT4. Afaik, the 8830 supports more than this...
If anyone can get more than this it would be interesting to get an adb
route dump. I have some userspace utilities that will do this under linux,
and might work using cygwin in NT/2k/XP with a bit of fiddling. Or if
anyone knows a nice easy way of accessing pci devices under windows
something else can be whipped up.

>
> If you're indeed getting the fastplay bug on a recent CVS, thats either
> maddening, or perhaps a chance to fix the damn thing for good. ;P
>
> > The other thing is, that the oss modules won't load
> > automatically, I have to modprobe them by hand even if my modules.conf
> > is 100 % the same as in the www.alsa-project.org page. Do you know other
> > programs in a gnome environment, that might confuse the card?
>
> One way is to just pull the 'tui' driver from the savannah CVS and use
> that, I used this for quite some time with no problem(*), up until the
> ALSA driver started working on my card. Heh.
>
> (*) Whats this about the OSS driver causing memory corruption? I have
> been noticing corrupt mp3's slowly turning up on my system, I was
> assuming it was from the fact my system has started suffering from an
> odd quirk of locking hard when its bumped too hard, (need to move it
> away from my feet and the falling distance of my heavy IBM model M
> keyboard...) with 512mb RAM and at least 256mb of it being used as file
> cache you lose a lot of data if the system locks hard. ;P Could it have
> been the Aureal driver the whole time? If its known to be capable of
> corrupting RAM, its really should be labled with a big red flashing USER
> BEWARE! ;P

The memory corruption/instability was with the hack driver I wrote
(aureal-2.x in CVS). 'plato' also has issues with adb route deletion, that
will cause crashes.

> Also, who owns the Sourceforge Aureal project? It seems to be quite dead
> for some time now. Can someone get a "This project is dead, everything's
> happening at the Savannah project." pointer posted there? The
> sourceforge page still seems to be the most known and highest listed in
> the search engines last I checked. ;P

Try Warren Chartier or Giao Phan
(http://sourceforge.net/project/memberlist.php?group_id=8109)

>
> Perhaps with the driver becoming mostly useable we should start a bit of
> a PR campaign. Attract more testers and eyeballs on the code. Perhaps
> pretty up the homepage and get it slashdotted. I volunteer for such duty. ;)

Probably the best time for a big PR blitz is when we merge into alsa.
Another nice thing to have would be an update/cleanup to the
aureal-techspec docs that Manuel wrote (aureal-2.x in CVS)...

> Also, what are the plans for merging with mainline ALSA? Will all the
> bleeding edge development still happen in the Savannah CVS and then just
> get pushed to mainline from there?
>

Yeah bleeding edge development will still probably happen here post
mainline-merge.

-Jeff