MRG32k3a generator

E760432

The MRG32k3a generator is a high-quality combined multiple recursive pseudorandom number generator widely used in scientific computing and simulations for its long period and good statistical properties.

Jump to: Statements Referenced by

Statements (48)

Predicate Object
instanceOf MRG (multiple recursive generator)
combined multiple recursive generator
pseudorandom number generator
hasApplication Monte Carlo simulation
financial risk simulation
queueing simulations
scientific computing
stochastic modeling
hasAuthor Pierre L’Ecuyer NERFINISHED
hasCategory multiple recursive generator with combination
uniform random number generator
hasCombinationRule z_n = (x_n - y_n) mod 4294967087
hasComponentCount 2
hasComponentType 3rd-order multiple recursive generator
hasDesignGoal high-quality random variates for simulation
support for parallel and distributed simulations
hasDimension 6
hasFeature ability to jump ahead in the sequence
well-defined stream and substream structure
hasModulus 4294944443
4294967087
hasOutputRange (0,1)
hasOutputType double-precision uniform variates
hasPeriodLength approximately 2^191
hasProperty good equidistribution properties
good statistical quality
long period
supports efficient substream generation
supports multiple independent streams
hasRecurrence x_n = (1403580 x_{n-2} - 810728 x_{n-3}) mod 4294967087
y_n = (527612 y_{n-1} - 1370589 y_{n-3}) mod 4294944443
hasSeedConstraint state must not be all zeros in each component
hasStateSize 6 integers
hasStreamCountPerSeed 2^64 streams (conceptual design)
hasSubstreamCountPerStream 2^64 substreams (conceptual design)
hasYearIntroduced late 1990s
isAlternativeTo Mersenne Twister NERFINISHED
linear congruential generators
isDescribedIn Pierre L’Ecuyer’s papers on combined multiple recursive generators
isImplementedIn Intel Math Kernel Library NERFINISHED
MATLAB random number generation toolbox NERFINISHED
NAG Library NERFINISHED
Python (via various simulation libraries)
R (via external packages and interfaces)
RngStreams library NERFINISHED
SSJ (Stochastic Simulation in Java) library NERFINISHED
isPreferredFor high-precision simulation studies
passesTestSuite TestU01 Crush battery (under recommended usage) NERFINISHED

Referenced by (1)

Full triples — surface form annotated when it differs from this entity's canonical label.

cuRAND provides MRG32k3a generator