LeptonFlurry (as an open-source product of Eightomic) is the evolved CBRNG (non-cryptographic) that has 2⁶⁴ key-based independent parallel bitstreams that each have a counter-based period of 2⁶⁴, excellent statistical randomness quality test results (counter/key iterated bitstreams passed Dieharder 3.31.1 dieharder -Y 1 -a -g 200 -k 2, NIST STS 2.1.2 assess 1000000 with 100 bitstreams, PractRand 0.96 RNG_test stdin -tlmin 1KB -tlmax 32TB, TestU01 1.2.3 BigCrush and cryptographic SAC tests that each had 2¹⁸ bit-flipped output pairs with near-perfect avalanche scores between 0.4991 and 0.5009), low-footprint implementation (efficient memory usage and small code size), no division/modulus/multiplication operators, no internal state transitions (the counter/key is the implementation-managed state) and ultra-fast speed (relative to the aforementioned constraints).
Each mention of LeptonFlurry refers to each of the following 6 variants individually (leptonflurry32x1, leptonflurry32x2, leptonflurry32x4, leptonflurry64x1, leptonflurry64x2 and leptonflurry64x4) implemented in C (requiring the stdint.h header to define unsigned integral types for both a 32-bit uint32_t and a 64-bit uint64_t).
Each of the following speed benchmark results (with gcc from an AMD A4-9120C) log the fastest process execution speed (in milliseconds) among several repetitions of generating 10 million pseudorandom output integers (uint32_t for leptonflurry32 variants and uint64_t for leptonflurry64 variants) in a blocking loop.
The leptonflurry32x1 function generates (from a uint64_t counter integer that must be seeded and a uint64_t key integer that must be seeded) a deterministic pseudorandom uint32_t integer as the return value.
Elapsed Counter/Key/Mixing/Output Bits
leptonflurry32x1 1196ms 224
squares32 2266ms 352
leptonflurry32x2, leptonflurry32x4, leptonflurry64x1, leptonflurry64x2 and leptonflurry64x4 aren't ready to publish yet.