22 Oct
2002
22 Oct
'02
8:45 p.m.
Benno Senoner wrote:
Hello Joern,
Are you using ALSA right ?
i think so :)
Perhaps an OSS emulation problem of ALSA ?
unlikely. if i use "play" on the very same file i use for latencytest,
it plays ok.
afaik, play is oss, so my oss emulation seems ok.
aplay also works.
I recall that I got grabled sound (even lockups) on
the SBLive
when using 128byte fragments (seemed like a driver or hardware problem).
What kind of audio card do you have ?
sblive. fun thing is, it used to work with kernels before 2.5.44....
ok, maybe the kernel just performs horribly, *BUT*:
i run do_tests, the sound is horrible, but when i aplay the same
soundfile *during the test*, i can hear clear audio over the garbage
(sblive does hardware mixing).
Anyway latencytest is completely outdated (does not
compile cleanly
on newer distros due to wrong includes etc).
Now that I have some spare time again (seems unbelievable but I were
able to finish my university degree in CS just a month before turning
30 ... better late than newer as we use to say :-) )
I'll try to rewrite some of the useful tools including latencytest,
adding alsa support and new stress test methods and like one guy of the
list asked , the possibility to chose the disks (or the path) on which
perform the disk i/o tests.
as i said, what irritates me is it used to work a few days ago...
Anyway low latency came useful in my thesis since a
part of the project
consisted in a real time laser scanner that tracks a laser spot that
you move over the object you like to scan that is filmed by two cameras
that permit a 3D reconstruction of the surface.
(at 25 FPS we have processing cycles of 40msec so it is quite easy for
the low lat patch to keep up since there is basically no disk i/o
present during the scanning activity).
it would be most welcome to have a reworked latencytest program. now
that 2.5. is almost into feature-freeze, we need to jump in, run tests
and bug the kernel guys for latency optimization. i fear most of the
tweaking has been to maximize throughput, which does not buy us much for
audio...
best,
jörn
--
Jörn Nettingsmeier
Kurfürstenstr 49, 45138 Essen, Germany
http://spunk.dnsalias.org (my server)
http://www.linuxdj.com/audio/lad/ (Linux Audio Developers)