sitmo

NAME Nim implementation of the Sitmo parallel random number generator

DESCRIPTION This implementation in Nim follows the original C++ implementation from Sitmo. The implementation is based on a paper by John Salmon and Mark Moraes and described in their paper “Parallel Random Numbers: As Easy as 1, 2, 3”. (Proceedings of 2011 International Conference for High Performance Computing, Networking, Storage and Analysis). The algorithm is based on the Threefish cryptographic cipher. It is usable for large scale parallel processing:

      1. O(1) and neglect-able skip ahead cost for block and
         leap-frogging parallelism.
      2. 32 bit seed offers 2^32 (4 billion) parallel random
         streams of length 2^256.
      3. Streams are guaranteed to be non-overlapping.

 The object, `sitmo', contains the state of the RNG.  The
 procedure `newsitmo' creates such object.  `seed' seeds the
 RNG, and `random' returns a `uint32' random number while
 advancing the internal state of the RNG.  There is only O(1)
 skip ahead cost, and the procedure `skip' advances the state
 of the RNG by a given amount (`uint64').  For details, see
 the documentation file `doc/sitmo.html'.

SYNOPSIS type sitmo = object s,k,o: array[4,uint64] oCounter: uint16 proc seed(r:var sitmo) proc seed(r:var sitmo, s:uint32) proc newsitmo:sitmo proc newsitmo(s:uint32):sitmo proc newsitmo(r:sitmo):sitmo proc random(r:var sitmo):uint32 proc skip(r:var sitmo, z:uint64) proc ==(x,y:sitmo):bool template !=(x,y:sitmo):bool proc setKey(r:var sitmo, k0,k1,k2,k3:uint64 = 0) proc setCounter(r:var sitmo, s0,s1,s2,s3:uint64 = 0; oCounter:uint16 = 0)

EXAMPLES import sitmo var r = newsitmo() r.seed(0x7FFFFFFFu32) echo r.random r.skip(0xFFFFFFFFFFFFFFFFu64) echo r.random

LICENSE This work is licensed under the MIT license. See file LICENSE for details.

SEE ALSO Sitmo PRNG: https://www.sitmo.com/?p=1206