From 1c2dcd1125d4a542a7b4b291e8de1f3b3cdc86a0 Mon Sep 17 00:00:00 2001 From: Gertjan van den Burg Date: Thu, 30 Jul 2015 18:48:58 +0200 Subject: By passing numbers between R and C as doubles we can avoid the limitations of the R integer type. This allows us to create a uniform RNG. --- README.md | 13 +++---------- 1 file changed, 3 insertions(+), 10 deletions(-) (limited to 'README.md') diff --git a/README.md b/README.md index 4d4c244..ac9a83e 100644 --- a/README.md +++ b/README.md @@ -52,18 +52,11 @@ You'll notice that the random numbers are indeed the same. Notes ----- -Since R is not capable of reliably handling integers larger than `2^32 - 1`, -the random numbers are internally capped at this value (using a bitwise and -with `0x7FFFFFF`), this influences the quality of the random numbers. The -random numbers are no longer uniformly distributed on `[0, 2^32 -1]`. For the -intended use of SyncRNG this is not a problem, but it is a compromise worth -considering when using SyncRNG. SyncRNG should definitely not be used for any -cryptographic purposes. +The random numbers are uniformly distributed on `[0, 2^32 - 1]`. TODO ---- -Future versions may include a random number generator that does not need -capping, and is uniform. It may also provide easier system-wide installation -through an R package and a Python module. +It may be easier to provide system-wide installation through an R package and +a Python module. -- cgit v1.2.3