Benchmarks for modular multiplications

Torbjorn Granlund tg at gmplib.org
Wed Jul 21 18:29:07 CEST 2010


Fabien.Herbaut at unice.fr writes:

  112bits 11s
  128bits 9.95s (!)
  160bits 28.63s
  192bits 27.37s
  224bits 32.27s
  256bits 37.6s 
  384bits 45.03s
  
  Please, would you know how to explain the slow-growing ? Or could you tell us a
  reference about such benchmarks ?
  
The time for a multiplication will always contain a constant term and a
linear term, plus one ore more super-linear terms which varies with
operand size as GMP changes multiplication algorithm.  For very small
multiplies, the constant term will dominate, which is what you're
seeing.

-- 
Torbjörn


More information about the gmp-discuss mailing list