> Matrices have to be truly spectacularly large to benefit from Strassens I don't have personal experience with other mmult methods, but Wikipedia says Strasse

> Matrices have to be truly spectacularly large to benefit from Strassens I don't have personal experience with other mmult methods, but Wikipedia says Strassen's algorithm is at least beneficial at matrix sizes that fit in cache. I have had occasion to deal with matrices of large dimension (though usually sparse, and usually not square) and those numbers don't seem crazy at all. Are they that far off-base? (Wikipedia does say that Coppersmith-Winograd tends to never be a useful improvement.) In any case, perhaps a more friendly response might be to point out that Strassen's algorithm tends to "bail out" into regular old matrix multiplication for small `n` anyway, so this still helps provide a speedup.