Daniel Lemire
Daniel Lemire
Indeed. The code could be faster in such cases.
I will document it better soon.
Sorry for the delay. Here are code samples that could be improved: ``` final Container c1 = x1.highLowContainer.getContainerAtIndex(pos1); final Container c2 = x2.highLowContainer.getContainerAtIndex(pos2); final Container c = c1.and(c2); if (c.getCardinality()...
@richardstartin Indeed. So there are multiple opportunities for optimization.
@richardstartin For clarity I turned your point into a distinct issue: https://github.com/RoaringBitmap/RoaringBitmap/issues/239
@richardstartin wrote: > I’d be tempted to verify the container isn’t scalarised before optimising this. If you get a significantly worse result and different allocation rates from -prof gc when...
@ppiotrow This does not sound bad. It might work!
@richardstartin That's possible too.
> It seems having two running iterators might be the 'best' way to do this? I would not call this approach "minimum overhead". Using iterators brings about a lot of...
@leventov This is valid code in Roaring: ``` RoaringBitmap rb = new RoaringBitmap(); rb.add(0, 1L