The version of kreed's blitter I used was an allegro C port, which I modified to use with DirectDraw subsequently. It is extremely slow, my Pentium III 650 runs it just okay, not too fast.
The only other version of his blitter that I can use is a 16-bit only MMX port, which should be faster. But I assume most people run 32-bit? So it wouldn't help? Dunno I guess.