On 10/04/2013 11:18 AM, Aleksey Shipilev wrote: > On 10/04/2013 03:34 AM, Brian Burkhalter wrote: >> Here is an alternative solution: http://cr.openjdk.java.net/~bpb/7189139.2/. > > Seems OK with me, as long as Miller-Rabin test tolerates the > lower-entropy PRNG. > >> If this seems reasonable, does anyone have suggestions as to testing? > > I would like to see the functional tests. > > We can take the list of prime numbers [1] as set P, and then check: > a) numbers in intersect([2;N), P) return isProbablyPrime=true in more > than >(1 - 1/2^cert) cases; > b) all numbers in difference([2;N), P) return isProbablyPrime=false
Wait, this is one is wrong (too general). I meant to say all numbers in [2;N) that return isProbablyPrime=false are not in P. > N > 50.000.000 would sound convincing. I need this list for another reason, so I had preprocessed the first 50M primes here: http://cr.openjdk.java.net/~shade/scratch/primes-50M.txt.xz -Aleksey.
