Better AFP reverse button performance during playback

[Reaper10]

I tried to automate the reversing button on/off on the AFP and LMMS crashed when I to play the song.
step1

step2

step3

when LMMS tries to do, do it froze.

[tresf]

Thanks for the bug report. Please fix your title to reflect the actual bug and provide detailed steps to reproduce. 👍

[badosu]

@Reaper10 Also, if you are able to send a screenshot like you did in the prevous one, that wiould help a lot.

[Reaper10]

I'll work on a list of steps to reproduce.
Is the title better?

[badosu]

We still need to understand what exactly is happening with you, but it is more clear now.

[Reaper10]

step1

step2

step3

when LMMS tries to do, do it froze.

[tresf]

Thanks 👍

[tresf]

Just tried to reproduce this... no crash... Can you provide a sample track for us to reproduce this on?

[Reaper10]

how do I that/

[tresf]

Save as .mmp extension, upload to gist, paste link here.

[Reaper10]

can't feger out how to upload

I'm stupid

[tresf]

Upload to gist. Then link.

[Reaper10]

how do I Upload a jzip file to gist with the mmp and the sample in it..
I'm stupid

[tresf]

Oh... Can you reproduce with a stock wav/ogg file? If not, you may need to host the project on something like Dropbox, google drive, etc.

[Reaper10]

https://drive.google.com/file/d/0B0Ar66xX6aiReGhfaFNja3k2WkE/view?usp=sharing
hope that helps:)

[tresf]

Thanks. This hangs the software for about 1 second each time it switches, but doesn't crash it. Seems like large samples are quite timely to reverse which is likely causing this delay.

[tresf]

I don't consider this a 1.2 milestone unless we can get confirmation of an actual crash. This is really a nice-to-have enhancement in my opinion.

[Sti2nd]

This is really a nice-to-have enhancement in my opinion.

Doing it live is ambitious. Would require a really strong machine. I think I have heard that one usually render out the changes? Maybe Audacity can help us

[Reaper10]

fl can do it live it's slow.

[Sti2nd]

Live and precomputed. https://answers.yahoo.com/question/index?qid=20091029071012AAS7lhe You are probably more right since you actually have FL studio.

It's slow you say? Depends on the machine. Audacity can for me reverse a 1 minute wav file in under half a second, I think it is pretty fast, but for live purposes it is slow of course. Unless you have a monster of a machine, or a short sample, it is going to sound slow when doing it live :/ However one could use tricks to make it seem like it is being done live, like I assumed FL studio did

[tresf]

fl can do it live it's slow.

Please stop comparing LMMS to commercial software.

[Reaper10]

Sorry it's only other music software I know. I learned all I know about music producing on it.
LMMS I like better!!!!:)

[musikBear]

just a tiny input
this could be achived with two wave-tracks, one reversed, and alternated automation of mute-this-track in both tracks..

[diizy]

I think the problem is that we're reloading the sample every time it's reversed, and in cases where the samplerate of the sample != playback samplerate, this means also resampling the sample every time.

A good solution would be going to Samplebuffer.c, and adding the ability to play buffers in reverse, so the reversing can be done on-the-fly. Playing in reverse that way shouldn't be any more cpu-consuming than playing forwards, after all it's just a matter of reversing some indices and counters...

Then, the actual samplebuffer wouldn't have to be reversed and reloaded, it could just be played forwards/backwards.

[Sti2nd]

Wow, I would never though about that. Sounds genious

[badosu]

This can be my opportunity to tinker with some sound processing code, I'll take a look at it.

[sakertooth]

I think the problem is that we're reloading the sample every time it's reversed, and in cases where the samplerate of the sample != playback samplerate, this means also resampling the sample every time.

A good solution would be going to Samplebuffer.c, and adding the ability to play buffers in reverse, so the reversing can be done on-the-fly. Playing in reverse that way shouldn't be any more cpu-consuming than playing forwards, after all it's just a matter of reversing some indices and counters...

Then, the actual samplebuffer wouldn't have to be reversed and reloaded, it could just be played forwards/backwards.

This is exactly what I had done in #6610. Instead of actually reversing the entire buffer in memory with std::reverse, it is now simply played forwards and backwards, which should've been the default behavior from the start.