Re: [Boost-users] run-time integer log2

Tim Blechmann wrote:

i see, maybe it would be an interesting addition to boost.integer to provide a generic and efficient version, based on bsr/ffs or the like ...

There is a lot of branchless or SWAR algorithm to salvaged from there and to be indeed made generic or at least platform independant w/r to various things. Other thign we never did is benching all of the versions of a given algorithm (with/without branch) on different platform to add a proper #ifdef/else to maximize performance.