Disable buffer swapping

[brisvag]

Fixes/Closes

Closes #5734.

Description

This change drastically improves performance on some hardware by sacrificing some quality (tearing is more likely). However, several people tested it and didn't notice any appreciable degradation of the rendering, so it seems like a worthwile exchange. See #5734 for more details.

We might be able to fix this on the vispy side, but for now this works well on the napari side.

Type of change

• Bug-fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• This change requires a documentation update

cc @ksofiyuk

[Czaki]

Could we control this using environment variable?

[codecov]

Codecov Report

Merging #5741 (cfe9bf3) into main (95356e4) will decrease coverage by 1.45%.
The diff coverage is 100.00%.

❗ Current head cfe9bf3 differs from pull request most recent head 3770faf. Consider uploading reports for the commit 3770faf to get more accurate results

@@            Coverage Diff             @@
##             main    #5741      +/-   ##
==========================================
- Coverage   89.83%   88.39%   -1.45%     
==========================================
  Files         608      608              
  Lines       51881    51882       +1     
==========================================
- Hits        46608    45861     -747     
- Misses       5273     6021     +748     

Impacted Files	Coverage Δ
napari/_qt/qt_viewer.py	79.23% <ø> (-0.39%)	⬇️
napari/settings/_experimental.py	100.00% <100.00%> (ø)

... and 39 files with indirect coverage changes

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[brisvag]

Could we control this using environment variable?

I would make it disabled by default in that case, though. Can you point me to where the env variable stuff is normally set?

[Czaki]

The user could set it in his .bashrc, .zshrc .profile, etc

[brisvag]

I meant where in the napari code we read all the env variables and store them for later usage in the code :P Or would you just read it when the canvas is created?

[Czaki]

I do not know the such place.

I prefer using it when creating Canvas. It is rare enough for me not to think about performance.

[brisvag]

Something like this? I suppose this should be somewhere in the docs as well.

[Czaki]

Yes. This needs to be added to docs.

[brisvag]

I see some mentions of environment variable regarding rendering in docs/guides/rendering.md, as well as docs/guides/rendering-explanation.md. I suppose one of those places could work? Opinions from the docs experts? @melissawm @psobolewskiPhD

[Czaki]

I prefer to have a single place with list of all Environment variables use by napari. Best, next to settings.

[psobolewskiPhD]

I think this section is prolly the best? But I think the whole thing at some point needs to be rewritten—I think it's mostly autogenerated?
https://napari.org/stable/guides/preferences.html
I mean you could also add this experimental preferences pane, like the async stuff? 🤷‍♂️

[brisvag]

I moved it to the experimental settings. Now it works both from there and as a env variable!

[brisvag]

I think it's mostly autogenerated?

Yep, so this should get added automatically now that it's a preference.

Maybe experimental is not quite the right place (advanced? :P) but for now I think it works.

[ksofiyuk]

I've tested it, works great for me, @brisvag thank you for the fix!

[psobolewskiPhD]

I can't see any performance difference nor any visual effect, but I like the implementation.
I wonder whether it should be true or false by default, given that the existing behavior is true.

[brisvag]

My reasoning was: it's more likely that a new user affected by the bug will simply think "oh well napari has terrible performance, bye" rather than opening an issue or testing experimental settings. And the cost seems basically zero otherwise 🤷

[alisterburt]

+1 on False as the new default - if we get reports of tearing we can reevaluate