Re: Grr, audio trouble.

[Marc Espie <Marc_(_dot_)_Espie_(_at_)_liafa_(_dot_)_jussieu_(_dot_)_fr>]

On Sat, May 01, 1999 at 08:30:06AM -0700, Hugh Graham wrote:
> Erich Zigler wrote:
> > I'm still having problems with mp3s in OpenBSD. I have a Soundblaster 16
> > ISA PnP card. That does work. It plays Quicktime movies and the like
> > beatifully under OpenBSD. But I just cannot get the mp3s to decode right.
> > I have an AMD K6-2 350, and when using mpg123 
> > 
> > 15783 needo    105    0  292K  544K run   -        0:50 106.98% mpg123
> > 
> > I know there is no reason mpg123 should be using that much cpu. It also
> > takes 20 seconds for it to get through 4 seconds of the mp3. mpg123 runs
> > fine under Linux, and I do not see any reason why it shouldn't under
> > OpenBSD. I am playing the mp3s off of a ext2 partition, but I dont think
> > that would cause taht much trouble. Included is my dmesg, but like I said
> > xanim is working just fine.
> > 
> 
> Is your mpg123 a linux bin, built from current ports, or built yourself?
> All will likely have somewhat different startup behaviour.. For instance
> the one in ports has had its startup ioctls changed to make it start
> playing sooner on certain sound cards, but the fix may be experimental...
> Try using whichever of those you're not running.

> I haven't used the one in ports, but fwiw mpg123 works fine with my
> sound blaster here.

The patch in ports is only for audio startup. It doesn't affect the rest
of the playing, where mpg123 is a quite normal application.

The fact that xanim works perfectly would tend to indicate that the
kernel audio driver is fine... 

On many soundcards, ioctl are cheap, so the audio ports' patch won't yield
any improvement.  This makes a very noticeable difference on CS4231, I
can assure you.

There is not enough information to know what's going on.

- run mpg123 in verbose mode, so that we can get an idea what kind of audio
sampling it figures out from your card. There might be a bug which might make
it run in a weird emulated mode.
- there might be a large performance hit in some audio mode, as compared to
what xanim does.
- AMD K6 don't show the same performance as genuine intel. This is especially
true in floating point computations.  There is some hand-tuned code in mpg123,
and this varies from version to version... Knowing from which version the 
linux binary was built, which hand-tuning version of the code it used, and
comparing with OpenBSD might be important.  Even the compiler version may
have a large influence...  there have been some very noticeable changes in
egcs or pgcc, specifically for some variants of intel-like procs.

-- 
	Marc Espie		
|anime, sf, juggling, unicycle, acrobatics, comics...
|AmigaOS, OpenBSD, C++, perl, Icon, PostScript...
| `real programmers don't die, they just get out of beta'