This change interleaves the vector used for storing bits and the vector used for storing rank blocks. Functionally identical, but faster because the two pieces of information required at each rank step are co-located in memory (aka reducing page-ins).
This change interleaves the vector used for storing bits and the vector used for storing rank blocks. Functionally identical, but faster because the two pieces of information required at each rank step are co-located in memory (aka reducing page-ins).