3 Aug
2005
3 Aug
'05
3:42 p.m.
There appears to be a serious problem with tap eq and ardour. Given a
session with 8 mono tracks, and 6 of them running thru tap eq, I get
near total gui freeze and/or jack disconnection when I stop rolling or
when moving the playhead.
This is only when the eq plugins are actually modifying the input
signal. If all the levels are set to 0, or the plugins are bypassed,
then all is well.
Upon testing, I added eq to tracks on track at a time. At the 4th
track, I noticed a small GUI hang loading the eq plugin. Sure enough,
after setting it to a preset state and stopping playback, I got the lockup.
Doug
P.S. Any attempts at emailing tom directly keep getting bounced back to me.
--
http://nostar.net/
3 Aug
3 Aug
5:59 p.m.
On Wed, 2005-08-03 at 14:47, Doug McLain wrote:
There appears to be a serious problem with tap eq and
ardour. Given a
session with 8 mono tracks, and 6 of them running thru tap eq, I get
near total gui freeze and/or jack disconnection when I stop rolling or
when moving the playhead.
This is only when the eq plugins are actually modifying the input
signal. If all the levels are set to 0, or the plugins are bypassed,
then all is well.
Upon testing, I added eq to tracks on track at a time. At the 4th
track, I noticed a small GUI hang loading the eq plugin. Sure enough,
after setting it to a preset state and stopping playback, I got the lockup.
Doug
P.S. Any attempts at emailing tom directly keep getting bounced back to me.
--
http://nostar.net/
Just for grins try setting "Stop plugins with transport" in the options
editor.
--
Jan "Evil Twin" Depner
The Fuzzy Dice
http://myweb.cableone.net/eviltwin69/fuzzy.html
"As we enjoy great advantages from the invention of others, we should be
glad of an opportunity to serve others by any invention of ours, and this
we should do freely and generously."
Benjamin Franklin, on declining patents offered by the governor of
Pennsylvania for his "Pennsylvania Fireplace", c. 1744
4 Aug
4 Aug
9:46 p.m.
Jan Depner wrote:
On Wed, 2005-08-03 at 14:47, Doug McLain wrote:
Thats always been disabled here.
--
http://nostar.net/
There appears to be a serious problem with tap eq
and ardour. Given a
session with 8 mono tracks, and 6 of them running thru tap eq, I get
near total gui freeze and/or jack disconnection when I stop rolling or
when moving the playhead.
This is only when the eq plugins are actually modifying the input
signal. If all the levels are set to 0, or the plugins are bypassed,
then all is well.
Upon testing, I added eq to tracks on track at a time. At the 4th
track, I noticed a small GUI hang loading the eq plugin. Sure enough,
after setting it to a preset state and stopping playback, I got the lockup.
Doug
P.S. Any attempts at emailing tom directly keep getting bounced back to me.
--
http://nostar.net/
Just for grins try setting "Stop plugins with transport" in the options
editor.
9:34 a.m.
Maybe it's denormal problems? You can try adding a noise generator set
to a very low level before the plugin, and see if this fixes it.
But I would think the tap plugins were denormal safe...
Andres
On Wed, 2005-08-03 at 14:47, Doug McLain wrote:
There appears to be a serious problem with tap eq and
ardour. Given a
session with 8 mono tracks, and 6 of them running thru tap eq, I get
near total gui freeze and/or jack disconnection when I stop rolling or
when moving the playhead.
This is only when the eq plugins are actually modifying the input
signal. If all the levels are set to 0, or the plugins are bypassed,
then all is well.
Upon testing, I added eq to tracks on track at a time. At the 4th
track, I noticed a small GUI hang loading the eq plugin. Sure enough,
after setting it to a preset state and stopping playback, I got the lockup.
Doug
P.S. Any attempts at emailing tom directly keep getting bounced back to me.
--
http://nostar.net/
10:05 a.m.
On Thu, Aug 04, 2005 at 08:34:59AM -0500, Andres Cabrera wrote:
Maybe it's denormal problems? You can try adding a
noise generator set
to a very low level before the plugin, and see if this fixes it.
But I would think the tap plugins were denormal safe...
That was my first idea as well, since the problems seemed to start
shortly after the input became silence. So I downloaded and checked
the sources. The biquad filter contains explicit denormal-killing code...
--
FA
10:44 a.m.
Alfons Adriaensen wrote:
On Thu, Aug 04, 2005 at 08:34:59AM -0500, Andres
Cabrera wrote:
That FLUSH_TO_ZERO macro doesn't always work though:
http://music.columbia.edu/pipermail/linux-audio-dev/2003-August/004581.html
Unfortunately the "fix" I suggest at the bottom of that mail doesn't always
work either (it turns out that casting to volatile doesn't do what you might
hope that it would) hence this alternative:
http://music.columbia.edu/pipermail/linux-audio-dev/2004-June/008739.html
~ Simon Jenkins
Maybe it's denormal problems? You can try
adding a noise generator set
to a very low level before the plugin, and see if this fixes it.
But I would think the tap plugins were denormal safe...
That was my first idea as well, since the problems seemed to start
shortly after the input became silence. So I downloaded and checked
the sources. The biquad filter contains explicit denormal-killing code...
11:03 a.m.
On Thu, Aug 04, 2005 at 03:44:57PM +0100, Simon Jenkins wrote:
That FLUSH_TO_ZERO macro doesn't always work
though:
http://music.columbia.edu/pipermail/linux-audio-dev/2003-August/004581.html
Unfortunately the "fix" I suggest at the bottom of that mail doesn't always
work either (it turns out that casting to volatile doesn't do what you
might hope that it would) hence this alternative:
http://music.columbia.edu/pipermail/linux-audio-dev/2004-June/008739.html
Interesting. So again I'm inclined to think that the reported problem is
due to denormals.
I usually do a very blunt ... + 1e-20 at strategic places. So far, nobody
has complained about the DC offset :-)
--
FA
3:21 p.m.
[Alfons Adriaensen]
Interesting. So again I'm inclined to think that
the reported problem is
due to denormals.
I usually do a very blunt ... + 1e-20 at strategic places. So far, nobody
has complained about the DC offset :-)
Similar approach works fine here, too (flipping the sign of the
addition constant at every sample or every block, depending on the
algorithm in question).
Inaudible (in fact barely measurable), code is branchless and simple.
Perfect solution for a stupid little problem.
Cheers, Tim
5 Aug
5 Aug
9:41 a.m.
Interesting. So
again I'm inclined to think that the reported problem
is due to denormals.
I usually do a very blunt ... + 1e-20 at strategic places. So far,
nobody has complained about the DC offset :-)
Similar approach works fine here, too (flipping the sign of the
addition constant at every sample or every block, depending on the
algorithm in question).
Inaudible (in fact barely measurable), code is branchless and simple.
Perfect solution for a stupid little problem.
11:22 a.m.
[Tim Blechmann]
> Similar approach works fine here, too (flipping
the sign of the
> addition constant at every sample or every block, depending on the
> algorithm in question).
>
> Inaudible (in fact barely measurable), code is branchless and simple.
> Perfect solution for a stupid little problem.
the denormal handling on the sse unit is far better
than on the fpu ...
it's also possible to enable hardware FTZ/DAZ ...
better a hardware solution than a software solution ;-)
Hard to disagree :)
However, I find SSE only useful from v2 on, not that much faster than
an AMD FPU, and a pain to code with gcc. It's limited to x86, so
you'll need portable C code to go with it. Likewise, h/w FTZ for the
FPU is nice, but not available anywhere.
In the end, I'll rather write and optimize quick and portable C/C++
than endure the coding, compilation and maintenance inferno that is
SIMD.
Cheers, Tim
2:39 p.m.
Hi all,
just noticed the thread, I hope it's not too late to chime in.
The problem is most likely due to a denormal bug. However, I have not
been able to see it, in fact the machines (two not-so-new Pentium-4
and one fairly old Pentium-3) I can access for testing don't exhibit
any such issue. TAP EQ has been rock solid on all of them (I use it
extensively). I suspect that many others have the same experience,
since I haven't been flooded with bug reports.
This certainly does *not* mean the problem is not real. However, I
have come to suspect that
(1) denormal behavior is very very processor dependant and thus
fairly unpredictable across different processor types (maybe
processors newer than mine are getting even worse?),
OR
(2) there is more to excessive CPU usage than a denormal problem.
Regardless of which is true, it seems I'm unable to provide a fix
since I cannot test whether a particular "denormal fix" macro does the
trick. I just can't make a reliable difference.
So, I would be very happy if someone who does have these problems
would try out the various methods discussed so far, and report which
of them works (if any). It should be possible to provide a replacement
for the FLUSH_TO_ZERO macro defined in tap_utils.h. Replacing that
macro should (positively) affect not just the EQ, but all other
plugins as well.
If you get this far, please please please make a patch of it and send
it to me (or this list). I will merge it and release a new version of
TAP-plugins. I can't think of anything else I could do.
Tom
7090
days inactive
7092
days old
linux-audio-dev@lists.linuxaudio.org
10 comments
8 participants
participants (8)
-
Alfons Adriaensen -
Andres Cabrera -
Doug McLain -
Jan Depner -
Simon Jenkins -
Tim Blechmann -
Tim Goetze -
Tom Szilagyi