fix some crashes in metal implementation when pathCount and eltCount are 0 #65
Loading…
Reference in New Issue
No description provided.
Delete Branch "macos_metal_fixes"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Running any of the samples under XCode on my old macbook with an integrated GPU resulted in asserts like this:
[MTLDebugComputeCommandEncoder dispatchThreads:threadsPerThreadgroup:]:1180: failed assertion `(threadsPerGrid.width(0) * threadsPerGrid.y(1) * threadsPerGrid.depth(1))(0) must not be 0.'This would happen when
pathCountand/oreltCountwere 0. These changes just protect against issuing GPU passes when there's no data.Couldn't we just early out if either
pathCountoreltCountare zero (similar tod3e27df818/src/graphics/gl_canvas.c (L1085))?Oh yeah definitely, I didn't realize the GL renderer did the same thing and was trying to be conservative since I wasn't sure if there were some passes that weren't dependent on
pathCountoreltCountthat needed to happen regardless.I've reverted the original change and added your suggested one.
2d3fe44239to9de9f936869de9f93686toceb4a3a95d