The code hasn't been updated in about a year: https://github.com/wbhart/bsdnt

The author appears to be planning a restart of the project, but that won't happen until next year: http://wbhart.blogspot.com/2011_10_01_archive.html

One big interruption is that I have been writing an FFT implementation completely from scratch that can eventually be used in BSDNT. You can find the code here:

https://github.com/wbhart/mpir-fft

For the time being it is implemented against the MPIR bignum library. But once BSDNT is up to it I will install it in BSDNT.

In my experience imath is the best (near) BSD licensed arbitrary precision integer arithmetic library. It also has a bunch of special purpose optimized functions for scientific computing.

imath: http://spinning-yarns.org/michael/sw/imath/

For my purposes I need a bignum library with an FFT, assembly optimisation, a bignum format based on full 32/64 bit words and the latest algorithms for performance.

If I don't want this to become a multi-decade effort I have to compromise somewhere. The particular compromise I make is that the library will initially compile with GCC only and will use inline assembly.