18 Mai
2007
18 Mai
'07
11:37
Here's the new code. I tested it on an Opteron multiprocessor box, too. I also changed the parameters so fft multiplication won't be used until 20,000 bits (from 6800 bits), but fast-gcd will be used down to 1400 bits. Perhaps these are consequences of testing on 64- bit machines with 32-bit mdigits (so naive multiplication is four times as fast as on a 32-bit machines with 16-bit mdigits) and native 64-bit adigits (so the shifts needed in fast gcd are significantly faster than on a 32-bit machine).
Brad