6 Oct
2003
6 Oct
'03
9:34 a.m.
two years ago i did a bunch of testing on latency to find the best kernel
and patches for my system, and got some damn good results with several
kernels (the best being 2.4.12-ac3-pe (alan cox's patch and pre-emptive
kernel patch using alsa on suse 7.3 with gcc 2.96 20000731). i now have a
new computer with gentoo linux and alsa 0.9.6, and can only get 16 ms
latency with perfect results. same soundcard - m-audio audiophile 2496
(ice1712). i've used kernel 2.4.20 with pe, ll, and some other options,
compiling with gcc version 3.2.3 20030422, and i get the same result with
all of them (as well as without pe or ll).
does anyone know what the problem could be? alsa version? gcc version?
kernel version? hardware (eth0 and soundcard share irq - never thought to
test without eth0 loaded)? i'd like to get latency under 4 ms.
thanks for your input,
dave
6 Oct
6 Oct
9:52 a.m.
one more note: i'm using pd (version 0.37 TEST 8) for all my realtime
audio, and the best i can get it to do is 16 ms latency. removing eth0
and it's modules does not improve performance.
On Mon, 6 Oct 2003 dwillis(a)stx.rr.com wrote:
two years ago i did a bunch of testing on latency to
find the best kernel
and patches for my system, and got some damn good results with several
kernels (the best being 2.4.12-ac3-pe (alan cox's patch and pre-emptive
kernel patch using alsa on suse 7.3 with gcc 2.96 20000731). i now have a
new computer with gentoo linux and alsa 0.9.6, and can only get 16 ms
latency with perfect results. same soundcard - m-audio audiophile 2496
(ice1712). i've used kernel 2.4.20 with pe, ll, and some other options,
compiling with gcc version 3.2.3 20030422, and i get the same result with
all of them (as well as without pe or ll).
does anyone know what the problem could be? alsa version? gcc version?
kernel version? hardware (eth0 and soundcard share irq - never thought to
test without eth0 loaded)? i'd like to get latency under 4 ms.
thanks for your input,
dave
10:02 a.m.
Monday 06 October 2003 13.20 skrev dwillis(a)stx.rr.com:
two years ago i did a bunch of testing on latency to
find the best kernel
and patches for my system, and got some damn good results with several
kernels (the best being 2.4.12-ac3-pe (alan cox's patch and pre-emptive
kernel patch using alsa on suse 7.3 with gcc 2.96 20000731). i now have a
new computer with gentoo linux and alsa 0.9.6, and can only get 16 ms
latency with perfect results. same soundcard - m-audio audiophile 2496
(ice1712). i've used kernel 2.4.20 with pe, ll, and some other options,
compiling with gcc version 3.2.3 20030422, and i get the same result with
all of them (as well as without pe or ll).
does anyone know what the problem could be? alsa version? gcc version?
kernel version? hardware (eth0 and soundcard share irq - never thought to
test without eth0 loaded)? i'd like to get latency under 4 ms.
thanks for your input,
dave
First, I'd like to know if you by "latency testing" mean that you ran
Benno's
latency test program?
For some info do check (seems to be down at the moment)
http://myweb.cableone.net/eviltwin69/ALSA_JACK_ARDOUR.html
Did you check what IRQ your card was assigned? There's lots of interesting
info in the above document.
Performance optimizing is something of a black art. Both the hardware and the
software needs to be in good condition and go well together...
As a sidenote, my system doesn't do much better than the 16ms scenario
either... I don't know if it's hardware or software problems... Trying the
"tricks" in the document above did help though...
Do let us know if there was something specific that helped.
/Robert
12:14 p.m.
On Mon, 6 Oct 2003, Robert Jonsson wrote:
First, I'd like to know if you by "latency
testing" mean that you ran
Benno's latency test program?
yes, as well as actual pd results. it seems i was using a messed up
custom latencytest, and when i found the proper one (i have many on my
system and it's been 2 years since using them) i got the best results i've
ever seen (1.1ms max for all tests at 0% cpu, 1.2ms at 80%).
so, it looks like the problem must be with pd. i'm going to try some
older versions now.
For some info do check (seems to be down at the
moment)
http://myweb.cableone.net/eviltwin69/ALSA_JACK_ARDOUR.html
thanks. it's up now.
Did you check what IRQ your card was assigned?
There's lots of interesting
info in the above document.
11, and nothing on 8 or 9 (though my eth, usb [unused], agp card, and
onboard sound [unused] are all on 11 as well).
As a sidenote, my system doesn't do much better
than the 16ms scenario
either... I don't know if it's hardware or software problems... Trying the
"tricks" in the document above did help though...
what "tricks?"
thanks,
dave
1:07 p.m.
New subject: [linux-audio-user] latency issues SOLVED
yes, as well as actual pd results. it seems i was
using a messed up
custom latencytest, and when i found the proper one (i have many on my
system and it's been 2 years since using them) i got the best results i've
ever seen (1.1ms max for all tests at 0% cpu, 1.2ms at 80%).
so, it looks like the problem must be with pd. i'm going to try some
older versions now.
pd-0.35.0-r2 and pd-0.36.0 = 12 ms
pd-0.37_pre8 = 16 ms
if i set the delay below 12 ms the sound gets distorted (but no xruns
reported). i guess i'll have to live with that.
thanks for "listening,"
dave
1:08 p.m.
Hallo,
dwillis(a)stx.rr.com hat gesagt: // dwillis(a)stx.rr.com wrote:
so, it looks like the problem must be with pd.
i'm going to try some
older versions now.
It is a somewhat known problem with Pd. If you use Pd in OSS mode ("pd
-oss" or similar) using ALSA's OSS emulation you should get much
better results. I have no idea and much too little understanding of
ALSA programming, why this is so.
ciao
--
Frank Barknecht _ ______footils.org__
10:10 p.m.
On Mon, 6 Oct 2003, Frank Barknecht wrote:
It is a somewhat known problem with Pd. If you use Pd
in OSS mode ("pd
-oss" or similar) using ALSA's OSS emulation you should get much
better results. I have no idea and much too little understanding of
ALSA programming, why this is so.
well, i'll be. according to pd, with oss mode, i can get
audiobuffer set down to 0 msec! now to test with an actual audioloop
(latency.pd).
did that, and i can get latency down to 4ms at 48k with oss emulation,
instead of 14.6ms with plain alsa. woohoo! thanks a bunch!
-dave
11:22 p.m.
It is a
somewhat known problem with Pd. If you use Pd in OSS mode ("pd
-oss" or similar) using ALSA's OSS emulation you should get much
better results. I have no idea and much too little understanding of
ALSA programming, why this is so.
well, i'll be. according to pd, with oss mode, i can get
audiobuffer set down to 0 msec! now to test with an actual audioloop
(latency.pd).
did that, and i can get latency down to 4ms at 48k with oss emulation,
instead of 14.6ms with plain alsa. woohoo! thanks a bunch! 
7 Oct
7 Oct
2:08 a.m.
Aaron Trumm <aaron(a)nquit.com> wrote:
I'm trying to follow this thread here so I can
start testing latency - what IS
PD I've only opened it once and I can't understand it. *laugh* I'll have to
go read up on it I suppose...
A synth/sequencer/development package along the lines of Max and JMax.
You can get {write} all sorts of extensions for it including an animation
module.
http://www.linuxjournal.com/article.php?sid=5994
3:27 a.m.
On Tue, 7 Oct 2003 08:55:25 +0900 (KST), dwillis(a)stx.rr.com wrote / a écrit:
did that, and i can get latency down to 4ms at 48k
with oss emulation,
instead of 14.6ms with plain alsa
am I the only one asking itself about this? oss emulation is faster than alsa? In every
case with any oss capable app?
GuyCLO~
-----------------------------------------------
#219055 http://counter.li.org
-----------------------------------------------
http://perso.wanadoo.fr/guy.clotilde/index.html
-----------------------------------------------
Change your Linux PC in a Rythm Station with holborn's GMORGAN:
http://perso.wanadoo.fr/guy.clotilde/GMORGAN/index.html
-----------------------------------------------
3:45 a.m.
At 03:09 PM 7/10/2003, Guy Clotilde wrote:
On Tue, 7 Oct 2003 08:55:25 +0900 (KST),
dwillis(a)stx.rr.com wrote / a écrit:
I think OSS emulation is faster only for PD, due to the way the OSS code
has bene written for PD, compared to the ALSA code.
Luke
----------------
Luke Yelavich
AudioSlack Founder and main package maintainer
Audio software packaged for the Slackware Linux Distribution
http://www.audioslack.com
luke(a)audioslack.com
did that, and i can get latency down to 4ms at
48k with oss emulation,
instead of 14.6ms with plain alsa
am I the only one asking itself about this? oss emulation is faster than
alsa? In every case with any oss capable app?
4:39 a.m.
Hallo,
Luke Yelavich hat gesagt: // Luke Yelavich wrote:
At 03:09 PM 7/10/2003, Guy Clotilde wrote:
Yes, it's a problem with Pd's ALSA code. What's even worse than the
latency issue, which I could live with, is that it isn't possible to
select ALSA pcms by name. You're stuck with hw:0, plughw:0, hw:1,
plughw:1, and so on. This basically makes it impossible to use Pd with
asoundrc tweaks and thus with some more advanced multichannel setups
in ALSA mode. These are known problems. I lack the ALSA knowledge
to do anything about it and those who have the knowledge lack the
time, but it would be great if those issues could be sorted out some
time. I'm sure, they will, though.
ciao
--
Frank Barknecht _ ______footils.org__
am I the only one asking itself about this? oss
emulation is faster than
alsa? In every case with any oss capable app?
I think OSS emulation is faster only for PD, due to the way the OSS code
has bene written for PD, compared to the ALSA code. 
6 Oct
6 Oct
8:10 p.m.
All of those goodies on IRQ 11 is probably what's killing you. Try
moving your sound card to different slots until it gets an IRQ to
itself. You really want IRQ 9 or 10 if possible otherwise it's going to
be lower priority than your stuff on IRQ 11. I didn't really put any
"tricks" per se in the ALSA_JACK_ARDOUR doc just suggestions from Mark
Knecht and others and my own experience with it. Playing with the
latency setting helps.
Jan
On Mon, 2003-10-06 at 09:00, dwillis(a)stx.rr.com wrote:
On Mon, 6 Oct 2003, Robert Jonsson wrote:
First, I'd like to know if you by
"latency testing" mean that you ran
Benno's latency test program?
yes, as well as actual pd results. it seems i was using a messed up
custom latencytest, and when i found the proper one (i have many on my
system and it's been 2 years since using them) i got the best results i've
ever seen (1.1ms max for all tests at 0% cpu, 1.2ms at 80%).
so, it looks like the problem must be with pd. i'm going to try some
older versions now.
For some info do check (seems to be down at the
moment)
http://myweb.cableone.net/eviltwin69/ALSA_JACK_ARDOUR.html
thanks. it's up now.
Did you check what IRQ your card was assigned?
There's lots of interesting
info in the above document.
11, and nothing on 8 or 9 (though my eth, usb [unused], agp card, and
onboard sound [unused] are all on 11 as well).
As a sidenote, my system doesn't do much
better than the 16ms scenario
either... I don't know if it's hardware or software problems... Trying the
"tricks" in the document above did help though...
what "tricks?"
thanks,
dave
11:42 a.m.
dwillis(a)stx.rr.com writes:
i've used kernel 2.4.20 with pe, ll, and some
other options,
compiling with gcc version 3.2.3 20030422, and i get the same result
with all of them (as well as without pe or ll).
If the preemption and low-latency patches make no difference then
check `sysctl kernel/lowlatency', which should be set to 1. If not,
then (as root) run `sysctl -w kernel/lowlatency=1'. If that works,
you probably need to add it to some /etc/init.d startup script.
--
Jack O'Quin
Austin, Texas, USA
7720
days inactive
7721
days old
linux-audio-user@lists.linuxaudio.org
13 comments
9 participants
participants (9)
-
Aaron Trumm -
dwillis＠stx.rr.com -
Frank Barknecht -
Guy Clotilde -
Jack O'Quin -
Jan Depner -
Luke Yelavich -
Rick Taylor -
Robert Jonsson