Developer Reference
Intel® oneAPI Math Kernel Library Vector Statistics Random Number Generator Performance Data
ID
772985
Date
12/04/2020
Public
Visible to Intel only — GUID: GUID-E212C9D0-247F-47A1-822F-E8389AEE7216
Uniform (VSL_RNG_METHOD_UNIFORM_STD)
Gaussian (VSL_RNG_METHOD_GAUSSIAN_ICDF)
Gaussian (VSL_RNG_METHOD_GAUSSIAN_BOXMULLER)
Performance
Notes:
Gaussian (VSL_RNG_METHOD_GAUSSIAN_BOXMULLER2)
GaussianMV (VSL_RNG_METHOD_GAUSSIANMV_ICDF)
GaussianMV (VSL_RNG_METHOD_GAUSSIANMV_BOXMULLER)
GaussianMV (VSL_RNG_METHOD_GAUSSIANMV_BOXMULLER2)
Exponential (VSL_RNG_METHOD_EXPONENTIAL_ICDF)
Laplace (VSL_RNG_METHOD_LAPLACE_ICDF)
Weibull (VSL_RNG_METHOD_WEIBULL_ICDF)
Cauchy (VSL_RNG_METHOD_CAUCHY_ICDF)
Rayleigh (VSL_RNG_METHOD_RAYLEIGH_ICDF)
Lognormal (VSL_RNG_METHOD_LOGNORMAL_ICDF)
Gumbel (VSL_RNG_METHOD_GUMBEL_ICDF)
Gamma (VSL_RNG_METHOD_GAMMA_GNORM)
Beta (VSL_RNG_METHOD_BETA_CJA)
ChiSquare (VSL_RNG_METHOD_CHISQUARE_CHI2GAMMA)
Uniform (VSL_RNG_METHOD_UNIFORM_STD)
UniformBits (VSL_RNG_METHOD_UNIFORMBITS_STD)
Bernoulli (VSL_RNG_METHOD_BERNOULLI_ICDF)
Geometric (VSL_RNG_METHOD_GEOMETRIC_ICDF)
Binomial (VSL_RNG_METHOD_BINOMIAL_BTPE)
Hypergeometric (VSL_RNG_METHOD_HYPERGEOMETRIC_H2PE)
Poisson (VSL_RNG_METHOD_POISSON_PTPE)
Poisson (VSL_RNG_METHOD_POISSON_POISNORM)
PoissonV (VSL_RNG_METHOD_POISSONV_POISNORM)
NegBinomial (VSL_RNG_METHOD_NEGBINOMIAL_NBAR)
Multinomial (VSL_RNG_METHOD_MULTINOMIAL_MULTPOISSON)
Visible to Intel only — GUID: GUID-E212C9D0-247F-47A1-822F-E8389AEE7216
Gaussian (VSL_RNG_METHOD_GAUSSIAN_BOXMULLER)
Performance
| FP precision | BRNG | Intel® Xeon® Processor E5-2699 v4 | Intel® Xeon® Platinum Processor 8280L |
|---|---|---|---|
| single | MCG31m1 | 9,09 | 5,93 |
| R250 | 12,25 | 10,86 | |
| MRG32k3a | 24,12 | 20,21 | |
| MCG59 | 9,18 | 4,88 | |
| WH | 14,28 | 7,98 | |
| SOBOL | 6,47 | 3,87 | |
| NIEDERR | 6,44 | 3,84 | |
| MT19937 | 8,25 | 4,29 | |
| MT2203 | 8,27 | 4,69 | |
| SFMT19937 | 7,28 | 5,02 | |
| Philox4x32-10 | 11,83 | 5,91 | |
| ARS-5 | 8,07 | 6,11 | |
| double | MCG31m1 | 17,62 | 7,96 |
| R250 | 23,10 | 15,44 | |
| MRG32k3a | 33,74 | 22,48 | |
| MCG59 | 17,88 | 7,24 | |
| WH | 23,22 | 10,46 | |
| SOBOL | 17,33 | 7,78 | |
| NIEDERR | 17,28 | 7,77 | |
| MT19937 | 19,52 | 7,94 | |
| MT2203 | 18,84 | 7,68 | |
| SFMT19937 | 17,73 | 9,19 | |
| Philox4x32-10 | 22,09 | 9,34 | |
| ARS-5 | 17,64 | 8,82 |
Notes:
The following hardware is used for measurements:
Intel® Xeon® Processor E5-2699 v4: 2 twenty two-core CPUs (2.20 GHz Processor Base Frequency), Red Hat* Enterprise Linux* Server 7.2
Intel® Xeon® Platinum Processor 8280L: 2 twenty-eight-core CPUs (2.70 GHz Processor Base Frequency), Red Hat* Enterprise Linux* Server 7.5
Performance is measured over the vector length of 1000 elements.
Parent topic: Continuous Distributions