Is this "bitmap" method the current path for Wayland + Qt?

Is this "bitmap" method the current path for Wayland + Qt?

yes

mmm ... I'm a bit worried about what this means for memory.

Previously, if you had say 10 canvases, they'd be rendered to one by one, and after each draw, the texture is released, allowing the GPU to re-use that memory for e.g. the texture target for the next canvas. But now all canvases hold onto their texture and a buffer of the same size. 🤔 This easily happens in e.g. a notebook.

Solved this by using a shared copy-buffer. It looks like re-using the texture does not do much; re-using the temporary buffer was what contributed by far the most to the speed-up.

e-using the temporary buffer was what contributed by far the most to the speed-up.

can you point to the one you are referring to in the diff (for my own learning)

This does not seem to clash with #673, except for a change in tests_mem/testutils.py which seems to be exactly the same.

I tested performance on Windows and Linux, and did not observe a difference. That's on a machine with integrated graphics. Will test on a Windows machine with a GPU later today.

edit: tested again, now with the same 4k monitor. The integrated graphics have trouble rendering that many pixels already ... it looks like maybe there is a tiny improvement from 18-ish to 19-ish fps, compared to 35 fps when rendering to screen.

Tested on a laptop with an nvidia gpu. Can observe a similar performance enhancement as for macos.