Neither block Lanczos nor block Wiedemann have a tutorial article available that explains them step by step; also, neither are amenable to small examples, as both methods require a matrix several times larger than the block size in order to need several iterations. Just making the block size 2 isn't safe either, as solving a matrix that small can work by random chance even if the implementation is broken.
Pretty much the only BW available that can scale to large problems is in CADONFS. Unfortunately scaling to large problems disqualifies the implementation for tutorial applications. The same is true of the BL code in Msieve.
Last fiddled with by jasonp on 20151115 at 23:46
