Page 3 of 3
Re: Veritable bug in 1.3.9
Posted: Tue Sep 15, 2009 5:07 pm
by steve
orgelquaeler wrote:In the release for 3.8, it mentions that Nyquist was updated to better use memory.
In some ways, yes this is true.
I've just tried running a Nyquist high-pass filter on a 45 minute 32 bit mono track in both Audacity 1.3.7 and 1.3.9
In 1.3.7 the peak memory usage was a shade over 500MB, whereas in 1.3.9 it went up to a whopping 1GB.
However, what happened next was interesting. In Audacity 1.3.9 the memory usage dropped back down to 23.8 MB, but in 1.3.7 it stayed at 524 MB !
So yes, there has been an improvement - the memory is now released after use (as it should be), but there clearly still seems to be a problem.
What seems particularly disappointing is that Audacity 1.2.6 can perform the exact same task with considerably less memory and a lot faster.
This is a known issue, and it will be corrected eventually, but for now we will just have to avoid processing long tracks in one go.
I'm going to look into this further and see if I can find out anything else. I'll post again if I find out anything.
Re: Veritable bug in 1.3.9
Posted: Wed Sep 16, 2009 1:27 am
by orgelquaeler
Thanks for the info, Steve. I was a little worried that the new memory usage was more the direction they were going permanently, so your explanation was a relief. Please keep us updated.
Just out of curiosity, what is the difference between how 1.2.6 handles nyquist and how 1.3.x does it, to make it less resource friendly?
Re: Veritable bug in 1.3.9
Posted: Wed Sep 16, 2009 6:56 am
by steve
orgelquaeler wrote:Just out of curiosity, what is the difference between how 1.2.6 handles nyquist and how 1.3.x does it, to make it less resource friendly?
Now there's a question. If I knew that I'd be well on my way to fixing it. It's not actually between 1.2.6 and 1.3.x.
There are a few Nyquist effects (notably ClipFix) that have always been memory hungry and slow because of how they work internally. This has not been a problem as these type of effects are specifically designed to work on short sections rather than whole tracks and carry a health warning.
The memory problem with other Nyquist effects did not appear until v.1.3.7 . It was not noticed until recently, probably because people that work on long tracks use mainly the internal (not Nyquist) effects that appear above the line in the effects list.
Re: Veritable bug in 1.3.9
Posted: Thu Sep 17, 2009 7:17 pm
by Gale Andrews
The problem of excessive memory use during processing of Nyquist effects seems to have been caused by changes we made prior to 1.3.7 so that the
progress bars in Nyquist effects worked properly. Prior to that the Nyquist progress bars would often freeze just after start and then suddenly complete right at the end of processing (or do other odd things).
Gale
Re: Veritable bug in 1.3.9
Posted: Fri Sep 18, 2009 11:34 am
by orgelquaeler
Thanks, Gale, for that link.
Re: Veritable bug in 1.3.9
Posted: Sat Sep 19, 2009 12:34 am
by Gale Andrews
We have made a fix and after testing on Windows XP I think processing time and memory usage is back to 1.3.6 levels or somewhat improved (according to the plug-in). For example, Clipfix will still take quite a lot of memory and CPU use and take some time to complete - but at least in 1.3.9 that memory will be released after processing.
If you want to test our fix and let us know of any problems, it's here:
http://www.gaclrecords.org.uk/audacity- ... -alpha.zip
(or in the Mac Nightly Build if anyone is reading on Mac - take the bottom build from
this list on or after 2009-09-19 , 03:15).
Note that the price of this fix (at least for now) is that progress bars in Nyquist generators like Click Track will behave oddly. Elapsed time should be correct but remaining time and progress bars will give false indications.
Thanks
Gale
Re: Veritable bug in 1.3.9
Posted: Sat Sep 19, 2009 1:16 am
by orgelquaeler
Thanks, Gale. I'm not nerdy enough to sit and watch the progress bars anyway lol! Though, seriously, if you are taking a poll, getting the effects processed more quickly (or at all) is the more important thing to me.
Re: Veritable bug in 1.3.9
Posted: Sat Sep 19, 2009 1:33 am
by orgelquaeler
Works Beautifully! It notched a 1:15 stereo file in 3 minutes!!! And the progress bar was right on. Good work!!
Re: Veritable bug in 1.3.9
Posted: Sat Sep 19, 2009 7:53 am
by Gale Andrews
orgelquaeler wrote:Works Beautifully! It notched a 1:15 stereo file in 3 minutes!!! And the progress bar was right on.
For Nyquist process effects like notch, the progress bars should be mostly OK. For Nyquist generators, you can call them broken.

But they may be fixed later if someone with the correct skills can look at it.
Gale
Re: Veritable bug in 1.3.9
Posted: Tue Oct 27, 2009 3:35 pm
by orgelquaeler
Alright, just an update. I have been very naughty, and have been using 1.3.10 (Sept. 16 compilation, I believe) for some production work, and it has been glitch free with what I have been using. I am basically using record, noise removal, compress dynamics, EQ, and all the normal playback stuff. I very much like having a keyboard shortcut available for "Play at speed", and it greatly improves the entire experience of mastering long flies. Thanks!!!