Beat detection slow for 48 k and 96 k tracks

Bug #1879588 reported by Daniel Schürmann
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
Mixxx
Fix Released
Medium
Daniel Schürmann

Bug Description

The beat detection in 2.3 for 48 k tracks and is significant slower that before.
I may case it has raised from 4 to 14 sec.

I did a bisect and the first failing commit is.
https://github.com/mixxxdj/mixxx/pull/2683/commits/4ed2e132693c3c3d286488daf400fb8559b45f27

I did not investigate this further to check the results.
At least the BPM values are different.
It looks like 44.2 k tracks are not affected.

We need to investigate this more.

Revision history for this message
Daniel Schürmann (daschuer) wrote :

It looks like this is a symptom of a deeper issue.
With 2.2.4 the analysis of 48 k Tracks is also significant longer than for 44.2k tracks.
However the result is looking reasonable.

In 2.3 the result looks random, before and after the commit of question.

We have also fixed a buffer overflow later in the code.
So maybe we see here a QM bug?

Revision history for this message
Cristiano Lacerda (crislacerda) wrote :

If this was working with VAMP the bug is unlikely to be in QM. The plugins call the exact same qm-lib.

In VAMP stepsize for 48khz is also 557 IIRC so that's why it's slower than 44khz which lands nicely in 512. But it should not be slower than VAMP using qm-lib directly.

Maybe it's something related to the downmix helper where the buffer overflow was happening..

I'm going to investigate this

Revision history for this message
Cristiano Lacerda (crislacerda) wrote :

Are you sure master is significantly slower than 2.2?

I just tried a couple of 48khz tracks here and they were equally slow on 2.2 and master. Taking a bit mote than 45 seconds as opposed to a bit less than 15 in 44khz for the same tracks.

The results a random before and after that commit because of the buffer overflow that the odd stepsize was causing probably.

But after that was fixed, it appears to me that it's working equally slow than it was before.
Tested current master twice with two 48khz tracks each time, and there is no difference at all from 2.2

summary: - 2.3 Regression: Beat detection slow for 48 k tracks
+ Beat detection slow for 48 k and 96 k tracks
Changed in mixxx:
importance: Undecided → Medium
status: New → In Progress
assignee: nobody → Daniel Schürmann (daschuer)
milestone: none → 2.3.0
Revision history for this message
Daniel Schürmann (daschuer) wrote :

Ok, this is no regression, but room for improvements.
Here it is:
https://github.com/mixxxdj/mixxx/pull/2817

Revision history for this message
Foss-4 (foss-4) wrote :

Should this be fix committed?

Revision history for this message
Daniel Schürmann (daschuer) wrote :

Yes.

Changed in mixxx:
status: In Progress → Fix Committed
Changed in mixxx:
status: Fix Committed → Fix Released
Revision history for this message
Swiftb0y (swiftb0y) wrote :

Mixxx now uses GitHub for bug tracking. This bug has been migrated to:
https://github.com/mixxxdj/mixxx/issues/9983

lock status: Metadata changes locked and limited to project staff
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.