Current implementation of rand() does not generate half of the numbers

[WheretIB]

In the current implementation, lowest two bits of the rand() result are always either 00/11 or 01/10 depending on the starting seed.

the following code produces an endless stream of either 0 3 0 3 0 3 0 3... or 1 2 1 2 1 2 1 2...:

srand(time(NULL));
while(1)
printf("%d\n", rand() % 4);

This happens because the multiplier does not follow the proper conditions of a good Linear congruential generator (rule #3 is violated):
http://en.wikipedia.org/wiki/Linear_congruential_generator#Period_length

[juj]

When I was studying LCGs from Knuth several years ago, I seem to have documented the following LCG configurations:

https://github.com/juj/MathGeoLib/blob/master/src/Algorithm/Random/LCG.cpp#L32

and then settled to using

https://github.com/juj/MathGeoLib/blob/master/src/Algorithm/Random/LCG.h#L69

i.e. a multiplier of 69621, increment 0, and modulus 0x7FFFFFFF, which was referred by Knuth as "Park and Miller generator nr. 3". Perhaps that would have a better randomness?