You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Would it be possible to leverage the memory routines for things like slices, which may be more efficient than a for loop in some contexts, to optimize contiguous modifications like fill_contiguous?
I guess this is possible, and chances are good that this can enhance performance. Likely the FrameBufferBackend trait would have to be extended, but that's not an issue.
However, I currently don't have the time to do it myself, so this would require other (new?) contributors to take action.
Activity
jounathaen commentedon Aug 27, 2026
I guess this is possible, and chances are good that this can enhance performance. Likely the
FrameBufferBackendtrait would have to be extended, but that's not an issue.However, I currently don't have the time to do it myself, so this would require other (new?) contributors to take action.
iriswebb commentedon Aug 27, 2026
PR #31