Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Vintage Ladder CPU Spikes with non-simd-length notes in F2 slot after wave shaper #6905

Closed
baconpaul opened this issue Apr 5, 2023 · 4 comments
Labels
Bug Report Item submitted using the Bug Report template DSP Issues and feature requests related to sound generation in the synth
Milestone

Comments

@baconpaul
Copy link
Collaborator

baconpaul commented Apr 5, 2023

Discord user Trace98 reports April 5 in surge bug reports

Trace98 — Yesterday at 6:55 PM
Hello, 
-using Vintage Ladder (Type 1 not Type 2) in filter 2, in any configuration where the filter comes after an active waveshaper
using one of the following
 (soft medium hard ojd sine add13 add15 addsqr3 full rect halfpos halfneg sft1fold 1fold 2fold wcfold and from sin+x onward)
causes the cpu to spike up to 100 immediately.
Adding to the weirdness the bug doesn't happen when pressing 3 or more notes simultaneously at the same time;
it only happens when pressing 1 or 2 notes, after the bug manifests adding other ntoes won't fix it.
Sometimes it doesn't happen instantly, it takes a moment or some note ons, but after it happens it stays there
-Reaper v6.78, Win 10,  both CLAP and VST3, latest surge nightly 1.2.nightly.7473f5e (bug happened even in b6c8442)
baconpaul — Yesterday at 7:00 PM
Are you in stereo routing? If you put it in mono routing does the problem resolve with 5 notes?
The vintage ladders are deterministic not iterative so something must be either causing a nan or under flow
(but we have software denormals off)
Trace98 — Yesterday at 7:05 PM
it's a bit weird, i'm testing a bit
by mono routing you mean every routing but wide and l-r right?
baconpaul — Yesterday at 7:06 PM
Yeah like s1 is perfect
Trace98 — Yesterday at 7:08 PM
yeah happens there too
with 1 2 3 5 5 7 notes simultaneously
with 4 or 8 notes it resolves
it's hard to track everything, now when switching between modes the bug persisted. it's very erratic
baconpaul — Yesterday at 7:15 PM
Great can you save your crashing patch as an fxp so I can try it?

my theory is that the model is getting an inactive voice in the simd slot (as all models do) but something about the model and un-initialized is making it run away or nan out.

VintageLadderCPUBug.fxp.zip

2023-04-05_13-41-42.mp4
@baconpaul baconpaul added DSP Issues and feature requests related to sound generation in the synth Bug Report Item submitted using the Bug Report template labels Apr 5, 2023
@mkruselj
Copy link
Collaborator

mkruselj commented Jun 20, 2023

Does not seem to happen over here. Hm.

@Trace98J
Copy link

Latest nightly (8ee2567) seems to have fixed this. I can't reproduce it as I could before, even trying various configurations of routings, filters and waveshapers.

@mkruselj
Copy link
Collaborator

@baconpaul Should we then close this? It's apparently a MSVC only thing, does it even make sense investing time in fixing this edge case?

@baconpaul
Copy link
Collaborator Author

yes
i will upgrade the msvc version we still use for 32 bit also but lets close it.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Bug Report Item submitted using the Bug Report template DSP Issues and feature requests related to sound generation in the synth
Projects
None yet
Development

No branches or pull requests

3 participants