Refactor Effect processing #7484

messmerd · 2024-09-02T00:45:57Z

All LMMS effects share some identical boilerplate code at the beginning and end of their process loop, such as an early return if the effect is not enabled and calling the checkGate method, but these tasks should not be the plugin developer's responsibility. So I created a wrapper method that handles these tasks and simplifies the process method for all effect plugins.

And along the way, I also made a few other minor changes.

Changes made:

Introduce processImpl and processBypassedImpl methods, and adapt each effect
plugin to use them
Move plugin state checks, RMS out sum calculations, and checkGate calls to the processAudioBuffer wrapper method
- Compressor and LOMM now use double instead of float for RMS out sum
- checkGate now runs for GranularPitchShifterEffect
Minor changes to LadspaEffect
Remove dynamic allocations and VLAs from VstEffect's process method
Some minor style/formatting fixes

- Introduce `processImpl` and `sleepImpl` methods, and adapt each effect plugin to use them - Use double for RMS out sum in Compressor and LOMM - Run `checkGate` for GranularPitchShifterEffect - Minor changes to LadspaEffect - Remove dynamic allocations and VLAs from VstEffect's process method - Some minor style/formatting fixes

sakertooth

We should probably remove the gate feature since it seems to generally be problematic for many plugins (as @LostRobotMusic mentioned), though I'm not sure what that would mean for backwards compatibility (maybe the risk is low enough since its rarely used I'm assuming?)

include/Effect.h

plugins/DualFilter/DualFilter.cpp

plugins/LadspaEffect/LadspaEffect.cpp

plugins/StereoMatrix/StereoMatrix.cpp

plugins/VstEffect/VstEffect.cpp

plugins/WaveShaper/WaveShaper.cpp

plugins/DynamicsProcessor/DynamicsProcessor.cpp

Rossmaxx · 2024-09-03T13:16:58Z

I am planning to remove gate knob as a seperate PR. Informed it because there's chance of conflicts

Co-authored-by: saker <[email protected]>

This reverts commit 67526f0.

plugins/VstEffect/VstEffect.cpp

JohannesLorenz

Functional review finished.

Outstanding reviews:

Style review (at least, I did not do it)
Testing review

When testing, different things should be tested:

Testing GranularPitchShifterEffect
Testing a LadspaEffect
Testing a Lv2Effect (if the respective comment will be changed)
Trying VstEffect with different (very high?) sampleRates

include/Effect.h

plugins/Lv2Effect/Lv2Effect.cpp

plugins/LadspaEffect/LadspaEffect.cpp

plugins/VstEffect/VstEffect.cpp

include/AudioEngine.h

PhysSong · 2024-09-05T12:15:48Z

I think there was a suggestion to handle W/D mix and check output level in single place, rather than letting each plugin handles them. This might be in the scope of this PR as well.

PhysSong · 2024-09-05T12:17:36Z

Also, for sleepImpl: I think something like processBypassedImpl will match what the operation is commonly called.

messmerd · 2024-09-13T04:14:42Z

I think there was a suggestion to handle W/D mix and check output level in single place, rather than letting each plugin handles them. This might be in the scope of this PR as well.

I like that idea, though I'm not sure if it can be easily done while keeping in-place processing of the buf audio buffer. So that could be left to another PR if it is feasible.

messmerd · 2024-09-13T04:19:19Z

I think all the reviews have been addressed now.

I was able to move the outSum calculations into the wrapper method, and introduced the ProcessStatus enum to allow plugins to specify what they want to happen after the process loop (continue, continue if not quiet, or sleep). I took this idea from the CLAP API. It makes the interface really clean and avoids the -1.0 magic value.

JohannesLorenz

Rework checked, OK from that point of view.
Style checked.
Testing still to be done.

include/Effect.h

plugins/Eq/EqEffect.cpp

plugins/StereoEnhancer/StereoEnhancer.cpp

Co-authored-by: Johannes Lorenz <[email protected]>

bratpeki · 2024-09-16T16:06:57Z

I see this needs testing. I would be interested. How exactly does one test this?

messmerd · 2024-09-16T16:21:48Z

@bratpeki This PR shouldn't change anything from a user's perspective, so testing would just involve using effects plugins (LMMS, VST, Ladspa, Lv2) and verifying that they still work as expected. If there are any regressions, it would probably affect the auto-off/gate

JohannesLorenz · 2024-09-16T20:50:44Z

I see this needs testing. I would be interested. How exactly does one test this?

See also #7484 (review)

bratpeki · 2024-09-17T16:24:51Z

Alright, I'll clone it, build it, and use it for a few days, reporting back on any mishaps, if they occur!

bratpeki · 2024-09-17T22:21:15Z

LB302 seems to not produce sound.

messmerd · 2024-09-18T09:25:22Z

LB302 seems to not produce sound.

It looks like it doesn't work on master either... strange. Not a bug from this PR fortunately

sakertooth

I'm currently looking into the regression reported by @bratpeki that seems to not be from this PR. This PR however looks good to me.

plugins/PeakControllerEffect/PeakControllerEffect.cpp

plugins/StereoEnhancer/StereoEnhancer.cpp

bratpeki · 2024-09-18T18:10:33Z

Also, please check if TripleOsc is louder than before.
This could just be my mind playing games with me!

It could also be unrelated to this PR, of course!

messmerd · 2024-09-18T19:36:35Z

@bratpeki This PR doesn't affect any of the instrument plugins so you don't need to test them

bratpeki · 2024-09-18T19:55:58Z

LADSPA works great, LV2 has also been tested, I'll look at some LinuxVSTs as well!

bratpeki · 2024-09-20T15:20:26Z

I have done tested a few AirWindows LinuxVST plugins, and they sound perfectly fine! I've tested the audio with Air, Chorus, StereoChorus, ADClip, and a few other plugins from the suite.

I've left a somewhat detailed report and question on the Discord server.

In any case, the audio sounds just fine, so this LGTM, but I'll leave it to you. Please let me know if you would like me to conduct any more testing on this!

messmerd · 2024-09-20T16:15:21Z

@bratpeki Thank you very much for testing! I'll look into what you mentioned on Discord, though I imagine it won't be an issue.

Co-authored-by: saker <[email protected]>

…-process-wrapper

messmerd · 2024-09-21T00:00:20Z

I couldn't find anything in the VST changes made by this PR that could explain the minor phase difference between this PR and master in the renders Peki shared on Discord.

With everything else in the testing looking fine, and with 2 approvals, I think this is ready to merge.

* Move common effect processing code to wrapper method - Introduce `processImpl` and `sleepImpl` methods, and adapt each effect plugin to use them - Use double for RMS out sum in Compressor and LOMM - Run `checkGate` for GranularPitchShifterEffect - Minor changes to LadspaEffect - Remove dynamic allocations and VLAs from VstEffect's process method - Some minor style/formatting fixes * Fix VstEffect regression * GranularPitchShifterEffect should not call `checkGate` * Apply suggestions from code review Co-authored-by: saker <[email protected]> * Follow naming convention for local variables * Add `MAXIMUM_BUFFER_SIZE` and use it in VstEffect * Revert "GranularPitchShifterEffect should not call `checkGate`" This reverts commit 67526f0. * VstEffect: Simplify setting "Don't Run" state * Rename `sleepImpl` to `processBypassedImpl` * Use `MAXIMUM_BUFFER_SIZE` in SetupDialog * Pass `outSum` as out parameter; Fix LadspaEffect mutex * Move outSum calculations to wrapper method * Fix Linux build * Oops * Apply suggestions from code review Co-authored-by: Johannes Lorenz <[email protected]> * Apply suggestions from code review Co-authored-by: saker <[email protected]> --------- Co-authored-by: saker <[email protected]> Co-authored-by: Johannes Lorenz <[email protected]>

This is slightly more thread-safe since the lock now covers the calls to `Effect::isSleeping()` and `AudioPorts::active()`, which matches what we did prior to PR LMMS#7484.

messmerd added 3 commits September 1, 2024 20:01

Fix VstEffect regression

51647ff

GranularPitchShifterEffect should not call checkGate

67526f0

sakertooth reviewed Sep 2, 2024

View reviewed changes

Apply suggestions from code review

cf99a6d

Co-authored-by: saker <[email protected]>

messmerd mentioned this pull request Sep 3, 2024

Remove "gate" option completely #7486

Closed

messmerd added 3 commits September 3, 2024 20:15

Follow naming convention for local variables

5993030

Add MAXIMUM_BUFFER_SIZE and use it in VstEffect

8f77cfa

Revert "GranularPitchShifterEffect should not call checkGate"

f552f7c

This reverts commit 67526f0.

sakertooth reviewed Sep 4, 2024

View reviewed changes

plugins/VstEffect/VstEffect.cpp Outdated Show resolved Hide resolved

JohannesLorenz self-requested a review September 4, 2024 18:05

JohannesLorenz reviewed Sep 4, 2024

View reviewed changes

messmerd added 6 commits September 8, 2024 21:42

VstEffect: Simplify setting "Don't Run" state

df58a76

Rename sleepImpl to processBypassedImpl

a262d17

Use MAXIMUM_BUFFER_SIZE in SetupDialog

0f070e4

Pass outSum as out parameter; Fix LadspaEffect mutex

6efad9b

Move outSum calculations to wrapper method

826b193

Fix Linux build

5dee396

Oops

ef53c0f

JohannesLorenz approved these changes Sep 13, 2024

View reviewed changes

include/Effect.h Outdated Show resolved Hide resolved

plugins/Eq/EqEffect.cpp Show resolved Hide resolved

plugins/StereoEnhancer/StereoEnhancer.cpp Outdated Show resolved Hide resolved

Apply suggestions from code review

5523488

Co-authored-by: Johannes Lorenz <[email protected]>

zonkmachine mentioned this pull request Sep 18, 2024

Define fpp_t and f_cnt_t to be of size_t #7363

Merged

sakertooth approved these changes Sep 18, 2024

View reviewed changes

plugins/PeakControllerEffect/PeakControllerEffect.cpp Outdated Show resolved Hide resolved

plugins/StereoEnhancer/StereoEnhancer.cpp Outdated Show resolved Hide resolved

messmerd and others added 2 commits September 20, 2024 12:17

Apply suggestions from code review

6f3ec8c

Co-authored-by: saker <[email protected]>

Merge branch 'master' of https://github.com/messmerd/lmms into effect…

a165c27

…-process-wrapper

messmerd merged commit 1825208 into LMMS:master Sep 21, 2024

messmerd deleted the effect-process-wrapper branch September 21, 2024 02:03

messmerd mentioned this pull request Jan 15, 2025

Tentative changelog for 1.3 #6949

Open

Uh oh!

Refactor Effect processing #7484

Refactor Effect processing #7484

Uh oh!

Conversation

messmerd commented Sep 2, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

sakertooth left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Rossmaxx commented Sep 3, 2024

Uh oh!

Uh oh!

JohannesLorenz left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

PhysSong commented Sep 5, 2024

Uh oh!

PhysSong commented Sep 5, 2024

Uh oh!

messmerd commented Sep 13, 2024

Uh oh!

messmerd commented Sep 13, 2024

Uh oh!

JohannesLorenz left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

bratpeki commented Sep 16, 2024

Uh oh!

messmerd commented Sep 16, 2024

Uh oh!

JohannesLorenz commented Sep 16, 2024

Uh oh!

bratpeki commented Sep 17, 2024

Uh oh!

bratpeki commented Sep 17, 2024

Uh oh!

messmerd commented Sep 18, 2024

Uh oh!

sakertooth left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

bratpeki commented Sep 18, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

messmerd commented Sep 18, 2024

Uh oh!

bratpeki commented Sep 18, 2024

Uh oh!

bratpeki commented Sep 20, 2024

Uh oh!

messmerd commented Sep 20, 2024

Uh oh!

messmerd commented Sep 21, 2024

Uh oh!

Reviewers

Assignees

Labels

messmerd commented Sep 2, 2024 •

edited

Loading

bratpeki commented Sep 18, 2024 •

edited

Loading