Eightomic
EntropyPRNG 16 B
It's the fastest sequential PRNG versus slower, weaker 16-bit rand() implementations.
Algorithm
Source
#include <stdint.h>
struct eightomic_prng_16_b_s {
uint32_t a;
uint32_t b;
};
uint16_t eightomic_prng_16_b(struct eightomic_prng_16_b_s *s) {
s->a = ((s->a << 13) | (s->a >> 19)) ^ s->b;
s->b += 1111111;
return s->a;
}
License
It's free and open source.
Reference
eightomic_prng_16_b() is the randomization function that accepts the following argument.
1: s is the struct eightomic_prng_16_b_s pointer with 2 32-bit unsigned integers s->a and s->b. Each must be initialized with any combination of values including 0.
The return value data type is uint16_t.
It returns the 16-bit unsigned integer pseudorandom number result.
Explanation
It's the fastest deterministic, sequential PRNG that passes stdin16 PractRand tests with up to 8 million randomized 16-bit numbers.
Furthermore, it guarantees a minimum cycle of 2³² numbers with only 64 bits of state by summing an odd-numbered constant to s->b which results in 2³² different numbers to XOR with s->a.
It's much faster than rand() in standard C implementations without security compliance violations.
A modern implementation of rand() passes PractRand tests up to 512000 randomized numbers, which is 7% of the high-quality randomized numbers that Eightomic PRNG 16 B generates.
Both Xorshift16 "798" and Xorshift32 pass PractRand tests up to 64000 randomized numbers, which is less than 1% of the high-quality randomized numbers that Eightomic PRNG 16 B generates.
When higher-quality 16-bit numbers are required with the same minimum cycle, the lower bits of Eightomic PRNG 32 A and Eightomic PRNG 32 B are suitable.
