Random Namespace Reference

Functions
template<RNGSalt salt>
auto	philox_4_uint64s (uint64_t counter, uint32_t seed, int key1, int key2=0)
	get 4 random uint 64 from the Philox RNG
template<RNGSalt salt, std::size_t N = 3, std::enable_if_t<(N > 1) and(N<=4), int > = 0>
auto	noise_uniform (uint64_t counter, uint32_t seed, int key1, int key2=0)
	Generator for random uniform noise.
template<RNGSalt salt, std::size_t N = 3, class = std::enable_if_t<(N >= 1) and (N <= 4)>>
auto	noise_gaussian (uint64_t counter, uint32_t seed, int key1, int key2=0)
	Generator for Gaussian noise.
template<typename T >
std::mt19937	mt19937 (T &&seed)
	Mersenne Twister with warmup.

Function Documentation

mt19937()

template<typename T >

std::mt19937 Random::mt19937

(

T &&

seed

)

Mersenne Twister with warmup.

The first 100'000 values of Mersenne Twister generators are often heavily correlated [32]. This utility function discards the first 1'000'000 values.

Parameters

seed

RNG seed

Definition at line 194 of file random.hpp.

Referenced by draw_polymer_positions().

noise_gaussian()

template<RNGSalt salt, std::size_t N = 3, class = std::enable_if_t<(N >= 1) and (N <= 4)>>

auto Random::noise_gaussian	(	uint64_t	counter,
		uint32_t	seed,
		int	key1,
		int	key2 = 0
	)

Generator for Gaussian noise.

Mean = 0, standard deviation = 1.0. Based on the Philox RNG using 4x64 bits. The Box-Muller transform is used to convert from uniform to normal distribution. The transform is only valid, if the uniformly distributed random numbers are not zero (approx one in 2^64). To avoid this case, such numbers are replaced by std::numeric_limits<double>::min() This breaks statistics in rare cases but allows for consistent RNG counters across MPI ranks.

Template Parameters

salt	decorrelates different thermostat types

Parameters

counter	counter for random number generation
seed	seed for random number generation
key1	key for random number generation
key2	key for random number generation

Returns

Vector of Gaussian random numbers.

Definition at line 150 of file random.hpp.

noise_uniform()

template<RNGSalt salt, std::size_t N = 3, std::enable_if_t<(N > 1) and(N<=4), int > = 0>

auto Random::noise_uniform	(	uint64_t	counter,
		uint32_t	seed,
		int	key1,
		int	key2 = 0
	)

Generator for random uniform noise.

Mean = 0, variance = 1 / 12. This uses the Philox PRNG, the state is controlled by the counter, the salt and two keys. If any of the keys and salt differ, the noise is not correlated between two calls along the same counter sequence.

Template Parameters

salt	RNG salt
N	Size of the noise vector

Parameters

counter	counter for random number generation
seed	seed for random number generation
key1	key for random number generation
key2	key for random number generation

Returns

Vector of uniform random numbers.

Definition at line 112 of file random.hpp.

philox_4_uint64s()

template<RNGSalt salt>

auto Random::philox_4_uint64s	(	uint64_t	counter,
		uint32_t	seed,
		int	key1,
		int	key2 = 0
	)

get 4 random uint 64 from the Philox RNG

This uses the Philox PRNG, the state is controlled by the counter, the salt and two keys. If any of the keys and salt differ, the noise is not correlated between two calls along the same counter sequence.

Definition at line 75 of file random.hpp.

References Utils::u32_to_u64().