This PR rolls back the vectorized popcount that have been added, as it seems to give me a address fault under certain conditions (only noticed in a test). Moreover, it improves the Sum(), Min() and Max() aggregates when traversing filled bitmaps, ~2x improvement over previous state.
This PR rolls back the vectorized popcount that have been added, as it seems to give me a address fault under certain conditions (only noticed in a test). Moreover, it improves the
Sum()
,Min()
andMax()
aggregates when traversing filled bitmaps, ~2x improvement over previous state.