24 Oct
2008
24 Oct
'08
6:29 a.m.
After a number of years of making successful music recordings using a
combination of rosegarden and ardour on my Ubuntu Dapper Drake system,
and my home grown real-time kernels, I thought it was time to upgrade to
Hardy Heron and use the proper real time kernel that the experts had
built. What a disappointment!
I have an internal pci sound blaster live card (emu10k1 chip set) and a
M-audio audiophile USB sound module.
On my dapper installation, which happened to have 2.6.22.1 real time
patched kernel (the default ubuntu kernel was 2.6.15), I could get
latencies as low as 1.5ms (32 frames/period) on the sound blaster and
2.7ms ( 64 frames/period) on the USB Audiophile with very few xruns; so
few xruns that I used to wonder why people posted to mailing
lists/forums about the problem.
With the stock Ubuntu 2.6.24-21rt kernel the sound blaster gives xruns
every few seconds with no load (no recording and no playing back) at
23ms latency (512 frames/period) and the Audiophile struggles at 43ms
( 1024 frames/period); I am getting xruns every 10-20 seconds. Infact
the rt kernel gives little, if any, improvement over the generic kernel.
I have the following settings in my limits.conf
@audio - rtprio 95
@audio - memlock 512000
@audio - nice -19
And I have a script to set the real-time priorities for the interrupts
for the interrupts that the sound system use - usually IRQ17 and IRQ19.
My user is also a member of the audio group.
I have tried the nohz=no on the kernel start up command line as
discussed in numerous fora with no change. I did not expect one:
johnt@TOMO001:/boot$ grep NO_HZ config-2.6.24-21-*
config-2.6.24-21-generic:CONFIG_NO_HZ=y
config-2.6.24-21-rt:# CONFIG_NO_HZ is not set
I have also tried the changes described in:
http://alsa.opensrc.org/index.php/Usb-audio
which also made no difference.
Any ideas what to try next? Have I got to go back and start build my own
kernels again (which always involved a battle between me and the nvidia
graphics card)?
John Tomlinson
24 Oct
24 Oct
6:40 a.m.
John Tomlinson wrote:
After a number of years of making successful music
recordings using a
combination of rosegarden and ardour on my Ubuntu Dapper Drake system,
and my home grown real-time kernels, I thought it was time to upgrade to
Hardy Heron and use the proper real time kernel that the experts had
built. What a disappointment!
I have an internal pci sound blaster live card (emu10k1 chip set) and a
M-audio audiophile USB sound module.
On my dapper installation, which happened to have 2.6.22.1 real time
patched kernel (the default ubuntu kernel was 2.6.15), I could get
latencies as low as 1.5ms (32 frames/period) on the sound blaster and
2.7ms ( 64 frames/period) on the USB Audiophile with very few xruns; so
few xruns that I used to wonder why people posted to mailing
lists/forums about the problem.
With the stock Ubuntu 2.6.24-21rt kernel the sound blaster gives xruns
every few seconds with no load (no recording and no playing back) at
23ms latency (512 frames/period) and the Audiophile struggles at 43ms
( 1024 frames/period); I am getting xruns every 10-20 seconds. Infact
the rt kernel gives little, if any, improvement over the generic kernel.
I cannot really help you here, but I can confirm that there are quite
some FFADO users reporting xruns on the more recent ubuntu releases
(Hardy and Intrepid), even with RT kernels. Personally I run Gutsy and
even with the 2.6.22-15-generic kernel it hardly ever gives me xruns.
Good to know that it might be a kernel scheduling issue, and not
something in my code.
Greets,
Pieter
9:14 a.m.
Pieter,
I'm considering adding a firewire audio device at the moment. Given the
current Ubuntustudio kernel challenges (2.6.24.21RT), have you tested FFADO
with a later kernel at all? (any distro)
Alex
I cannot really help you here, but I can confirm that there are quite
some FFADO users reporting xruns on the more recent ubuntu releases
(Hardy and Intrepid), even with RT kernels. Personally I run Gutsy and
even with the 2.6.22-15-generic kernel it hardly ever gives me xruns.
Good to know that it might be a kernel scheduling issue, and not
something in my code.
Greets,
Pieter
_______________________________________________
Linux-audio-user mailing list
Linux-audio-user(a)lists.linuxaudio.org
http://lists.linuxaudio.org/mailman/listinfo/linux-audio-user
9:46 a.m.
alex stone wrote:
Pieter,
I'm considering adding a firewire audio device at the moment. Given the
current Ubuntustudio kernel challenges (2.6.24.21RT), have you tested
FFADO with a later kernel at all? (any distro)
I haven't. But then again, this has nothing to do with firewire devices.
Any other audio device will most likely suffer too.
Greets,
Pieter
Alex
I cannot really help you here, but I can confirm that there are quite
some FFADO users reporting xruns on the more recent ubuntu releases
(Hardy and Intrepid), even with RT kernels. Personally I run Gutsy and
even with the 2.6.22-15-generic kernel it hardly ever gives me xruns.
Good to know that it might be a kernel scheduling issue, and not
something in my code.
Greets,
Pieter
_______________________________________________
Linux-audio-user mailing list
Linux-audio-user(a)lists.linuxaudio.org
<mailto:Linux-audio-user@lists.linuxaudio.org>
http://lists.linuxaudio.org/mailman/listinfo/linux-audio-user
7:13 a.m.
You should receive answers like "hey, there's 64studio and other audio
related distros out there!!!" in a couple of hours on this list.
By the way, I am running debian lenny and using my own rt kernel.
I succesfully compiled 2.6.24 and 2.6.26 with real time patch and tried
2.6.21 from 64studio and the ubuntu studio one to learn something more using
'make oldconfig'
2.6.26 performances are good (1.5ms 32 frames/period) even if sometimes
jackd freezes and I don't know why. qjackctl load is around 30-50% with
ardour running a 14tracks mix with full fx in playback mode. Duplex mode
loads are around 60-70%
2.6.24 performances are the same but with no freezes and an increased load,
around 1-2% (same mix).
With 64studio and ubuntu studio kernels I didn't noticed the improvements I
expected...
I then started to built my own and benchmarking following (as you probably
know):
http://www.alsa-project.org/main/index.php/Low_latency_howto
For nvidia drivers have a look at this.
http://desiato.tinyplanet.ca/~lsorense/debian/debian-nvidia-dri-howto.html&…
Removing network/server/router... and experimentale features in the
"oldconfig/menuconfig" process helped me a lot.
regards
r
2008/10/24 John Tomlinson <normalblaster-lau(a)yahoo.co.uk>
After a number of years of making successful music
recordings using a
combination of rosegarden and ardour on my Ubuntu Dapper Drake system,
and my home grown real-time kernels, I thought it was time to upgrade to
Hardy Heron and use the proper real time kernel that the experts had
built. What a disappointment!
I have an internal pci sound blaster live card (emu10k1 chip set) and a
M-audio audiophile USB sound module.
On my dapper installation, which happened to have 2.6.22.1 real time
patched kernel (the default ubuntu kernel was 2.6.15), I could get
latencies as low as 1.5ms (32 frames/period) on the sound blaster and
2.7ms ( 64 frames/period) on the USB Audiophile with very few xruns; so
few xruns that I used to wonder why people posted to mailing
lists/forums about the problem.
With the stock Ubuntu 2.6.24-21rt kernel the sound blaster gives xruns
every few seconds with no load (no recording and no playing back) at
23ms latency (512 frames/period) and the Audiophile struggles at 43ms
( 1024 frames/period); I am getting xruns every 10-20 seconds. Infact
the rt kernel gives little, if any, improvement over the generic kernel.
I have the following settings in my limits.conf
@audio - rtprio 95
@audio - memlock 512000
@audio - nice -19
And I have a script to set the real-time priorities for the interrupts
for the interrupts that the sound system use - usually IRQ17 and IRQ19.
My user is also a member of the audio group.
I have tried the nohz=no on the kernel start up command line as
discussed in numerous fora with no change. I did not expect one:
johnt@TOMO001:/boot$ grep NO_HZ config-2.6.24-21-*
config-2.6.24-21-generic:CONFIG_NO_HZ=y
config-2.6.24-21-rt:# CONFIG_NO_HZ is not set
I have also tried the changes described in:
http://alsa.opensrc.org/index.php/Usb-audio
which also made no difference.
Any ideas what to try next? Have I got to go back and start build my own
kernels again (which always involved a battle between me and the nvidia
graphics card)?
John Tomlinson
_______________________________________________
Linux-audio-user mailing list
Linux-audio-user(a)lists.linuxaudio.org
http://lists.linuxaudio.org/mailman/listinfo/linux-audio-user
9:10 a.m.
I'd have to agree that Hardy with 2.6.24.21 definitely doesn't perform as
well as the gutsy build. More xruns, etc..
Has anyone tested and tried a later kernel in any distro with any success?
2.6.26, 2.6.27? Is there any consequence audio/midi wise with one of these
later kernels?
Alex.
On Fri, Oct 24, 2008 at 1:13 PM, Cassiel <raffaele.morelli(a)gmail.com> wrote:
You should receive answers like "hey, there's
64studio and other audio
related distros out there!!!" in a couple of hours on this list.
By the way, I am running debian lenny and using my own rt kernel.
I succesfully compiled 2.6.24 and 2.6.26 with real time patch and tried
2.6.21 from 64studio and the ubuntu studio one to learn something more using
'make oldconfig'
2.6.26 performances are good (1.5ms 32 frames/period) even if sometimes
jackd freezes and I don't know why. qjackctl load is around 30-50% with
ardour running a 14tracks mix with full fx in playback mode. Duplex mode
loads are around 60-70%
2.6.24 performances are the same but with no freezes and an increased load,
around 1-2% (same mix).
With 64studio and ubuntu studio kernels I didn't noticed the improvements I
expected...
I then started to built my own and benchmarking following (as you probably
know):
http://www.alsa-project.org/main/index.php/Low_latency_howto
For nvidia drivers have a look at this.
http://desiato.tinyplanet.ca/~lsorense/debian/debian-nvidia-dri-howto.html&…
Removing network/server/router... and experimentale features in the
"oldconfig/menuconfig" process helped me a lot.
regards
r
2008/10/24 John Tomlinson <normalblaster-lau(a)yahoo.co.uk>
After a number of years of making successful music recordings using a
combination of rosegarden and ardour on my Ubuntu
Dapper Drake system,
and my home grown real-time kernels, I thought it was time to upgrade to
Hardy Heron and use the proper real time kernel that the experts had
built. What a disappointment!
I have an internal pci sound blaster live card (emu10k1 chip set) and a
M-audio audiophile USB sound module.
On my dapper installation, which happened to have 2.6.22.1 real time
patched kernel (the default ubuntu kernel was 2.6.15), I could get
latencies as low as 1.5ms (32 frames/period) on the sound blaster and
2.7ms ( 64 frames/period) on the USB Audiophile with very few xruns; so
few xruns that I used to wonder why people posted to mailing
lists/forums about the problem.
With the stock Ubuntu 2.6.24-21rt kernel the sound blaster gives xruns
every few seconds with no load (no recording and no playing back) at
23ms latency (512 frames/period) and the Audiophile struggles at 43ms
( 1024 frames/period); I am getting xruns every 10-20 seconds. Infact
the rt kernel gives little, if any, improvement over the generic kernel.
I have the following settings in my limits.conf
@audio - rtprio 95
@audio - memlock 512000
@audio - nice -19
And I have a script to set the real-time priorities for the interrupts
for the interrupts that the sound system use - usually IRQ17 and IRQ19.
My user is also a member of the audio group.
I have tried the nohz=no on the kernel start up command line as
discussed in numerous fora with no change. I did not expect one:
johnt@TOMO001:/boot$ grep NO_HZ config-2.6.24-21-*
config-2.6.24-21-generic:CONFIG_NO_HZ=y
config-2.6.24-21-rt:# CONFIG_NO_HZ is not set
I have also tried the changes described in:
http://alsa.opensrc.org/index.php/Usb-audio
which also made no difference.
Any ideas what to try next? Have I got to go back and start build my own
kernels again (which always involved a battle between me and the nvidia
graphics card)?
John Tomlinson
_______________________________________________
Linux-audio-user mailing list
Linux-audio-user(a)lists.linuxaudio.org
http://lists.linuxaudio.org/mailman/listinfo/linux-audio-user
_______________________________________________
Linux-audio-user mailing list
Linux-audio-user(a)lists.linuxaudio.org
http://lists.linuxaudio.org/mailman/listinfo/linux-audio-user
12:24 p.m.
Cassiel <raffaele.morelli(a)gmail.com> writes:
2.6.26 performances are good
I can not understand how 2.6.26 performance can be good, given the
recent regression I read about:
http://kerneltrap.org/Linux/Removing_the_Big_Kernel_Lock
Can someone shed light on this?.
--
Esben Stien is b0ef@e s a
http://www. s t n m
irc://irc. b - i . e/%23contact
sip:b0ef@ e e
jid:b0ef@ n n
9:53 p.m.
New subject: Ubuntu Hard real time kernel (2.6.24-21) poor performance.
On Fri, 2008-10-24 at 18:20 +0200, Esben Stien wrote:
Cassiel <raffaele.morelli(a)gmail.com> writes:
Something must have changed/improved since then.
Pretty good results here running 2.6.26.7 + 2.6.26.6-rt11 (except that
USB midi seems to be broken... arghh...)
So far the most "stable" realtime kernel for me has been is
2.6.24.7-rt21 (if we are talking about recent kernels, of course).
-- Fernando
2.6.26 performances are good
I can not understand how 2.6.26 performance can be good, given the
recent regression I read about:
http://kerneltrap.org/Linux/Removing_the_Big_Kernel_Lock
Can someone shed light on this?. 
10:19 p.m.
New subject: Ubuntu Hard real time kernel (2.6.24-21) poor performance.
Fernando Lopez-Lezcano wrote:
On Fri, 2008-10-24 at 18:20 +0200, Esben Stien wrote:
good to hear that. does the rt patch apply cleanly (since it's for the
previous release), or does it require some massaging?
--
jörn nettingsmeier
home://germany/45128 essen/lortzingstr. 11/
http://spunk.dnsalias.org
phone://+49/201/491621
Kurt is up in Heaven now.
I can not understand how 2.6.26 performance can
be good, given the
recent regression I read about:
http://kerneltrap.org/Linux/Removing_the_Big_Kernel_Lock
Can someone shed light on this?.
Something must have changed/improved since then.
Pretty good results here running 2.6.26.7 + 2.6.26.6-rt11 (except that
USB midi seems to be broken... arghh...)
25 Oct
25 Oct
12:04 a.m.
New subject: Ubuntu Hard real time kernel (2.6.24-21) poor performance.
On Sat, 2008-10-25 at 02:24 +0200, Joern Nettingsmeier wrote:
Fernando Lopez-Lezcano wrote:
It applies cleanly as far as I remember (it is in the testing repos for
Planet CCRMA fc9/8). Previous kernels of the same lineage have frozen on
me, I guess I have not tested this one enough...
But the midi thing is a big big problem if you want to use midi - I
do :-)
-- Fernando
On Fri, 2008-10-24 at 18:20 +0200, Esben Stien
wrote:
good to hear that. does the rt patch apply cleanly (since it's for the
previous release), or does it require some massaging? I can not understand how 2.6.26 performance can
be good, given the
recent regression I read about:
http://kerneltrap.org/Linux/Removing_the_Big_Kernel_Lock
Can someone shed light on this?.
Something must have changed/improved since then.
Pretty good results here running 2.6.26.7 + 2.6.26.6-rt11 (except that
USB midi seems to be broken... arghh...)
3 Dec
3 Dec
7:22 p.m.
Fernando Lopez-Lezcano <nando(a)ccrma.Stanford.EDU> writes:
Something must have changed/improved since then.
Well, it surely doesn't seem like that, since no rt kernel has been
released after the detonation.
--
Esben Stien is b0ef@e s a
http://www. s t n m
irc://irc. b - i . e/%23contact
sip:b0ef@ e e
jid:b0ef@ n n
8:44 p.m.
New subject: Ubuntu Hard real time kernel (2.6.24-21) poor performance.
Fernando Lopez-Lezcano wrote:
So far the most "stable" realtime kernel for me has been is
2.6.24.7-rt21 (if we are talking about recent kernels, of course).
Well, it's "good to know" that RT and its typical usage domains such as
music and audio processing are still a pretty low priority for the
kernel developers, even after all these years. :-(
--
Florin Andrei
http://florin.myip.org/
4 Dec
4 Dec
12:30 a.m.
New subject: Ubuntu Hard real time kernel (2.6.24-21) poor performance.
On Wed, 2008-12-03 at 16:44 -0800, Florin Andrei wrote:
Fernando Lopez-Lezcano wrote:
I don't know if there's development I don't know about. I suspect that a
lot of the development was with the goal of realtime transaction
oriented applications, not audio... we got our wishes as a side
effect :-) If that is so (and I may be completely wrong) we may get our
wishes again when the companies marketing these products have to upgrade
their kernels (it is a slower cycle)...
-- Fernando
So far the most "stable" realtime kernel for me has been is
2.6.24.7-rt21 (if we are talking about recent kernels, of course).
Well, it's "good to know" that RT and its typical usage domains such as
music and audio processing are still a pretty low priority for the
kernel developers, even after all these years. :-( 
6 Dec
6 Dec
2:41 a.m.
On Wed, Dec 3, 2008 at 7:44 PM, Florin Andrei <florin(a)andrei.myip.org> wrote:
Fernando Lopez-Lezcano wrote:
Judging from LKML activity it's actually a higher priority than ever.
I think the issue is that the Linux audio community has not been very
engaged with kernel development lately (except for Fernando) and
naturally the kernel devs will assume that if no one complains to them
then it's all good. The RT kernel work lately has been driven by
traditional hard RT users like simulation, machine control, HPC, etc
who all seem to be happy. Also the kernel RT people are focused on
getting the RT kernel patches into mainline.
Does the Ubuntu RT kernel enable the latency tracing stuff?
Lee
So far the most "stable" realtime kernel for me has been is
2.6.24.7-rt21 (if we are talking about recent kernels, of course).
Well, it's "good to know" that RT and its typical usage domains such as
music and audio processing are still a pretty low priority for the
kernel developers, even after all these years. :-(
5:05 p.m.
New subject: Ubuntu Hard real time kernel (2.6.24-21) poor performance.
On Sat, 2008-12-06 at 01:41 -0500, Lee Revell wrote:
On Wed, Dec 3, 2008 at 7:44 PM, Florin Andrei
<florin(a)andrei.myip.org> wrote:
Sigh, I'm not that good at all. I spent two weeks on and off on that for
nothing, at least on my builds MIDI is still not working for 2.6.26.x rt
(your suggestion - thanks! - enabled midi input to work but there was
not followup and something is still drastically wrong with the alsa
timer - apps that use it don't actually output anything). I meant to
post it but this last couple of weeks I had other priorities and posts
like "midi does not work" or "it freezes" without hard data will
probably not help either.
... and the last few times I have posted with problems I have not
received any feedback from lkml except for you. And no forthcoming
patches to test or anything like that (with one exception for some BUG
reports but that lead nowhere). At least from the point of view of my
email account I get a lot of silence. Maybe I'm not being cc'ed? I just
can't at this point try to follow lkml directly (I did till the
beginning of this year).
Fernando Lopez-Lezcano wrote:
Judging from LKML activity it's actually a higher priority than ever.
I think the issue is that the Linux audio community has not been very
engaged with kernel development lately (except for Fernando) and
naturally the kernel devs will assume that if no one complains to them
then it's all good.
So far the most "stable" realtime kernel for me has been is
2.6.24.7-rt21 (if we are talking about recent kernels, of course).
Well, it's "good to know" that RT and its typical usage domains such as
music and audio processing are still a pretty low priority for the
kernel developers, even after all these years. :-( The RT kernel work lately has been driven by
traditional hard RT users like simulation, machine control, HPC, etc
who all seem to be happy.
Strange. Maybe different configurations or hardware? When I try to run
2.6.26.x rt patched kernels I eventually end with a completely dead
machine, not very often (once every two days?) but it happens (I don't
currently have a way to trace what happens). And there's nothing
official AFAIK for 2.6.27.x.
Also the kernel RT people are focused on
getting the RT kernel patches into mainline.
That would be sooooo good.... at least then it would _have_ to keep up
with new releases.
At this point I'm also ready to give up and just try to find latency
problems in the vanilla kernel, find a way to trace their origins and
report that to lkml.
-- Fernando
10 Dec
10 Dec
6:43 a.m.
"Lee Revell" <rlrevell(a)joe-job.com> writes:
I think the issue is that the Linux audio community
has not been
very engaged with kernel development lately (except for Fernando)
and naturally the kernel devs will assume that if no one complains
to them then it's all good.
Isn't the problem about the kernel lock thing. According to that
article on kerneltrap, as I understand it, it would take quite some
time to remedy this regression and that's probably why there will be
no rt release for a while?.
--
Esben Stien is b0ef@e s a
http://www. s t n m
irc://irc. b - i . e/%23contact
sip:b0ef@ e e
jid:b0ef@ n n
5828
days inactive
5875
days old
linux-audio-user@lists.linuxaudio.org
15 comments
9 participants
participants (9)
-
alex stone -
Cassiel -
Esben Stien -
Fernando Lopez-Lezcano -
Florin Andrei -
Joern Nettingsmeier -
John Tomlinson -
Lee Revell -
Pieter Palmers