Mp3gain like tool for multicore processors?, Multi-core processors should be able to scan multiple files at once. Options

[Stereodude]

Is there a version of mp3gain, or another tool that does max no clip album gain (like mp3gain) that takes advantage of multi-core processors?

For example, if you have a quad core processor, Foobar's conversion tool converts four files simultaneously. It seems like mp3gain or some other tool should have similar functionality with regard to changing the playback volume of mp3s.

I did some google searching and searching this forum, but didn't find anything.

Does it exist?

Thanks!

[ZinCh]

I believe this programs doesnt need much calculation (check cpu usage), their work speed = hdd speed, so I am not sure you can make it any faster.

[Stereodude]

I believe this programs doesnt need much calculation (check cpu usage), their work speed = hdd speed, so I am not sure you can make it any faster.

Maybe if you're using HDD's from 15 years ago.

I have no problem saturating 1 core worth of my Core 2 Quad Q6600 @ 3gHz while scanning.

CPU usage scanning MP3s:


HDD speed is definitely not a limitation. I can transcode 4 FLAC files to LAME -V 1 mp3s at the same time at ~130x using Foobar and the HDD can keep up no problem allowing the CPU cores to be saturated and that transcoding operation is much more HDD intensive.

CPU usage FLAC to MP3:


So, it should be possible to make it 4x as fast (at least for scanning).

This post has been edited by Stereodude: May 23 2009, 15:31

[Big_Berny]

I have no problem saturating 1 core worth of my Core 2 Quad Q6600 @ 3gHz while scanning.

CPU usage scanning MP3s:

Hmm... Well at least on that screen no core seems to be fully 'saturated'...

EDIT: This seems to be interesting, altough I don't see any sourcefiles or binary: http://www.tatuh.com.br/mcore/en/cases_en.html

This post has been edited by Big_Berny: May 23 2009, 15:58

[Eli]

dBpoweramp will use multiple cores.

[Stereodude]

Hmm... Well at least on that screen no core seems to be fully 'saturated'...

EDIT: This seems to be interesting, altough I don't see any sourcefiles or binary: http://www.tatuh.com.br/mcore/en/cases_en.html

That's because Windows doesn't lock the affinity of a single process to a single core. So the process will move around spreading the load between the cores. The key is that it's using ~25% total CPU which is equal to full utilization of a single core which means it's not HDD limited.

Edit: And, your second link appears to be someone looking to speed up a single scan operation as opposed to scanning multiple files at once.

This post has been edited by Stereodude: May 23 2009, 16:47

[Mike Giacomelli]

Is there a version of mp3gain, or another tool that does max no clip album gain (like mp3gain) that takes advantage of multi-core processors?

foobar2000.

And no its not disk limited until you get out to 8-16 processors, since a typical mid range hard disk can sustain upwards of 40MB/s for a sequential read and a typical 192kbps mp3 is just 24kbps.

[Stereodude]

dBpoweramp will use multiple cores.

Maybe I'm really dense, but I don't see how to make dBpoweramp do the same thing as mp3gain. First it uses replaygain, which as I understand it targets a perceived loudness playback level, which is not what I'm after. I just want to reduce the playback volume of an entire album (folder) just enough to avoid clipping on playback on any of the files by directly altering the files (no tags). Second, assuming replaygain can emulate the mp3gain functionality I'm after I don't see how to scan an existing folder without converting it to something in dBpoweramp.

Foobar can also scan files using replaygain very quickly, but again I don't see how to emulate the mp3gain functionality I'm after with replaygain. I see I can scan an album write tags to the files, and then apply the tags altering the files, but it seems to be targeting a playback volume of ~89dB not the "max no clip gain" I'm after.

[Stereodude]

foobar2000.

Can you elaborate on how to do it? I've played with it and don't see how to get the equivalent result to what I get from mp3gain.

[Mike Giacomelli]

Options > Playback then set the replaygain preamp high and tell it to prevent clipping.

Then right click on your music > replaygain > apply to mp3 data.

[ShowsOn]

Options > Playback then set the replaygain preamp high and tell it to prevent clipping.

Then right click on your music > replaygain > apply to mp3 data.

So does doing this have exactly the same effect as mp3gain? It changes the variable in the MP3 file in a way that effects the way the file plays back on ALL mp3 decoders?

I've wanted an MP3gain front end that would just run multiple instances at the same time.

[Mike Giacomelli]

It changes the variable in the MP3 file in a way that effects the way the file plays back on ALL mp3 decoders?

Yes.

[Stereodude]

Options > Playback then set the replaygain preamp high and tell it to prevent clipping.

Then right click on your music > replaygain > apply to mp3 data.

This definitely didn't work for me.

I used these settings:


I scanned the files as an album, then applied the album gain data to the files.

Here's what I got:


The first set of files are from Foobar2000 0.9.6.5 and the second set it from MP3Gain. The files do not have the same gain. Foobar's replaygain'd files still used 89dB as the target album gain level.

[greynol]

I use foobar2000 for scanning (because of the speed) and then make sure the data is stored in an APE tag so it can be read by mp3gain (because of the interface, and I use 92 dB as the target volume).

EDIT: Grammar.

This post has been edited by greynol: May 23 2009, 18:54

[ShowsOn]

If you set up the ReplayGain dsp in Dbpoweramp ripper, can you get that to apply the value when using dbPoweramp converter to convert between FLAC and MP3 files?

[Mike Giacomelli]

I just experimented with foobar and you're right it seems to ignore those settings when applying to albums and converting to WAV. I'm not sure why.

[Stereodude]

I guess someone just needs to rewrite the GUI front end for MP3Gain to add this capability.

[frozenspeed]

Is there a version of mp3gain, or another tool that does max no clip album gain (like mp3gain) that takes advantage of multi-core processors?

For example, if you have a quad core processor, Foobar's conversion tool converts four files simultaneously. It seems like mp3gain or some other tool should have similar functionality with regard to changing the playback volume of mp3s.

I did some google searching and searching this forum, but didn't find anything.

Does it exist?

Thanks!

You can just open it (mp3gain) so ou have as many instances as cores & set the affinity for each process to a different core & viola! multiprocessing mp3gain.

[greynol]

I've found that method to be cumbersome and still nowhere near as efficient.

I could be wrong, but I think foobar2000 decodes any given single mp3 file for analysis much faster than mp3gain. Furthermore, if you're interested in calculating album gain (which is more useful to me, even when shuffling because a quiet song should still be quiet) you'll only see a benefit if you're doing more than one album at a time.

This post has been edited by greynol: May 23 2009, 22:10

[Stereodude]

You can just open it (mp3gain) so ou have as many instances as cores & set the affinity for each process to a different core & viola! multiprocessing mp3gain.

Uh, that would be awful...

If I have 20 albums to scan, I would need to open up 4 copies, load 5 in each, make sure I didn't overlap any, and then start them all, etc...

I'll pass.

[frozenspeed]

You can just open it (mp3gain) so ou have as many instances as cores & set the affinity for each process to a different core & viola! multiprocessing mp3gain.

Uh, that would be awful...

If I have 20 albums to scan, I would need to open up 4 copies, load 5 in each, make sure I didn't overlap any, and then start them all, etc...

I'll pass.

If you have 20 albums to scan and you would rather wait for someone to write a multi-threaded gain app than just wait the few mins it will take then your priorities are pretty uh... interesting...

tell you what, write an mp3 header writing class & I'll write some multithreading around it for you. FB2k may all ready do the trick.

[greynol]

You know, looking at your desired settings it's clear that all you want to do is album-based peak normalization, which sort of defeats the purpose of Replaygain.

Answer me something, just how many of your albums aren't already normalized to within 1.5 dB of 0 dBFS?

For the situation where you're reducing the gain of an album that is already normalized because the compressed version goes over full-scale, have you demonstrated to yourself that you can actually hear the clipping caused by decompression at playback? I bet you haven't.

This post has been edited by greynol: May 24 2009, 01:36

[Stereodude]

Answer me something, just how many of your albums aren't already normalized to within 1.5 dB of 0 dBFS?

Nearly all tracks will clip by 1.5dB or 3.0dB after being converted to mp3 regardless of the music type. This is what I want to avoid by lowering their playback level by a few dB.

For the situation where you're reducing the gain of an album that is already normalized because the compressed version goes over full-scale, have you demonstrated to yourself that you can actually hear the clipping caused by decompression at playback? I bet you haven't.

I haven't specifically tried an ABX test to listen for clipping since I'm not sure how to formulate a proper test for it. A file the clips will always be louder than one that doesn't and that invalidates the ABX test. The clipping affects the RMS volume calculation, so it becomes a fine matter of adjusting the volume of the decoded clipped file so it matches the one that doesn't clip. Since it's pretty easy to fix the files so they don't clip on playback I always do it.

[greynol]

You simply take the file that clips, decode it and then apply the same attenuation that mp3gain would have applied to prevent it from cliping. ABX this with the decoded version of the file that has had mp3gain applied to prevent it from clipping.

http://www.hydrogenaudio.org/forums/index....st&p=558514