I think we should be testing with smaller, more typical user buffer sizes to verify the performance when the split is disabled or removed.

Reply via email to