[mercury-users] random.m and integer overflow

[Thomas Charles CONWAY]

Martin Stein, you write:
> Hi,
> 
> I've just investigated the random.m problem:
> 
> The 3 constants A, C and M aren't appropriate for a 32 bit signed int! One needs
> a 32 bit unsigned int. A*M is arround 2^32. Using a seed a little less than M
> should give an integer overflow error. That's all.
> Properly A,C and M are a good choice, but not int.
> 

Here are some values from Press et al (originally due to Knuth):
M		A		C
6075		106		1283
11979		430		2531
53125		171		11213
134456		141		28411
81000		421		17117
86436		1093		18257
120050		2311		25367
139968		3877		29573

Hope these are useful.
-- 
Thomas Conway <conway at cs.mu.oz.au>
Nail here [] for new monitor.  )O+