For various performance reasons, it is surmised that canvases that are at least 256x256 can be hardware accelerated, so consider modifying the logic that deals with all offscreen canvases to use that as the minimum size (even if we only use a small part of it).
Need to first investigate:
what is the real benefit of hardware acceleration for non-visible, offscreen canvases?
which platforms/engines this is true for, and if the target dimensions differ in any meaningful way?
if you have a 256x256 canvas but you only use a small part of it, do you still get the HA if you're drawImage()ing from it and you only take that small part?
For various performance reasons, it is surmised that canvases that are at least 256x256 can be hardware accelerated, so consider modifying the logic that deals with all offscreen canvases to use that as the minimum size (even if we only use a small part of it).
Need to first investigate:
+@petkaantonov