1

What set of witnesses is sufficient for the Miller-Rabin test to be correct for all numbers up to 10¹⁸? I know that use of primes up to 17 as witnesses suffices for n < 341550071728321.

kinokijuf
  • 968
  • 1
  • 11
  • 33

3 Answers3

2

According to this record page, the set of 7 SPRP bases: {2, 325, 9375, 28178, 450775, 9780504, 1795265022} is sufficient for a deterministic test to at least n = 2^64 ( > 10^19).

Brett Hale
  • 21,653
  • 2
  • 61
  • 90
0

According to OEIS, use of witnesses up to 23 suffices for numbers up to 3825123056546413051

kinokijuf
  • 968
  • 1
  • 11
  • 33
0

If you're willing to use a Baillie-Wagstaff test instead of a Miller-Rabin test, it has been certified to be error-free in classifying primes up to 2^64. The coding is not much more complicated, the function executes more quickly than a Miller-Rabin test, and there are no known errors of classification.

user448810
  • 17,381
  • 4
  • 34
  • 59