Crystal implementations of PCG32, XorShift1024* and Xoroshiro128+, WELL512, ChaCha12, ChaCha20, ISAAC+ pseudorandom generators.
See http://www.pcg-random.org for more details about PCG
See http://xoroshiro.di.unimi.it for more details about xorshift family
This shard provides additional implementations of Random module, so usage is identical to usage of Random in stdlib.
Also, PCG32 and ChaCha12/20 implements jump method that allows to instantly travel along generated sequence forward and backward. See examples in spec
Benchmark results on my laptop:
[andrew@lenovo prngs]$ crystal -v
Crystal 0.23.1 (2017-07-17) LLVM 4.0.1
[andrew@lenovo prngs]$ crystal bench.cr
Warning: benchmarking without the `--release` flag won't yield useful results
Random::MT19937 2.08M (479.94ns) (± 1.07%) 2.83× slower
Random::ISAAC 1.88M (532.21ns) (± 2.59%) 3.14× slower
Random::ISAACPlus 1.84M (543.98ns) (± 1.85%) 3.20× slower
Random::Xorshift1024star 5.89M (169.76ns) (± 2.22%) fastest
Random::Xoroshiro128plus 5.69M ( 175.7ns) (± 2.41%) 1.03× slower
Random::ShardPCG32 5.04M (198.51ns) (± 2.43%) 1.17× slower
Random::WELL512 2.15M (464.87ns) (± 2.58%) 2.74× slower
Random::ChaCha(20) 137.62k ( 7.27µs) (± 1.15%) 42.80× slower
Random::ChaCha(12) 225.96k ( 4.43µs) (± 0.64%) 26.07× slower
SecureRandom 1.77M (566.23ns) (± 1.27%) 3.34× slower
65554439.77777062
[andrew@lenovo prngs]$ crystal bench.cr --release
Random::MT19937 21.99M ( 45.47ns) (± 4.23%) 4.12× slower
Random::ISAAC 35.42M ( 28.24ns) (± 6.12%) 2.56× slower
Random::ISAACPlus 34.13M ( 29.3ns) (± 6.96%) 2.65× slower
Random::Xorshift1024star 70.57M ( 14.17ns) (± 6.36%) 1.28× slower
Random::Xoroshiro128plus 90.55M ( 11.04ns) (± 4.40%) fastest
Random::ShardPCG32 64.03M ( 15.62ns) (± 6.25%) 1.41× slower
Random::WELL512 22.76M ( 43.94ns) (± 2.04%) 3.98× slower
Random::ChaCha(20) 22.1M ( 45.24ns) (± 4.17%) 4.10× slower
Random::ChaCha(12) 28.69M ( 34.85ns) (± 4.22%) 3.16× slower
SecureRandom 2.01M (497.42ns) (± 1.41%) 45.04× slower
973336292.1701912