gpucoder.rng

R2026b

Control random number generator in generated GPU code

Since R2026b

Syntax

gpucoder.rng("default")

gpucoder.rng(seed)

Description

gpucoder.rng initializes the random number generator in generated GPU code by using the NVIDIA^® cuRAND library. When you generate GPU code from gpucoder.rng, the generated code initializes the generator and uses it for subsequent calls to gpucoder.rand, gpucoder.randi, and gpucoder.randn. Use this function to reproduce sequences of random numbers or generate different sequences of random numbers. For more information, see Reproducibility.

If you call gpucoder.rng in MATLAB^®, gpucoder.rng initializes the MATLAB random number generator by using the rng function.

Note

Because the NVIDIA cuRAND library and MATLAB use different random number generators, the random numbers produced in MATLAB for gpucoder.rand, gpucoder.randn, and gpucoder.randi do not match the random numbers produced by generated GPU code.

gpucoder.rng("default") initializes the random number generator with the default seed of 0. In generated GPU code, the random number generator uses the CURAND_RNG_PSEUDO_XORWOW generator type. For more information, see the cuRAND documentation on the NVIDIA website.

gpucoder.rng(seed) sets the seed of the random number generator.

Examples

Use Default Generator Seed

Create a MATLAB function, randUsingDefault, that generates random numbers by using gpucoder.rand. To use the default random number generator seed in subsequent calls to GPU Coder™ random number generation functions, use the "default" argument.

function Y = randUsingDefault(n)
gpucoder.rng("default");
Y = gpucoder.rand(n);
end

Generate GPU code from randUsingDefault.

cfg = coder.gpuConfig("mex");
codegen randUsingDefault.m -config cfg -args {200}

Generate a 4-by-4 matrix by using the generated MEX function randUsingDefault_mex.

Y = randUsingDefault_mex(4)

Y = 4×4
0.4385	0.0530	0.4822	0.5126
0.4604	0.3377	0.0428	0.2643
0.2502	0.3968	0.5084	0.0520
0.4947	0.8744	0.6545	0.5790

Set and Reset Random Number Generator Seed

Create a MATLAB function named randnUsingSeed that generates a matrix by using the gpucoder.randn function. If the resetGenerator input argument is true, reset the random number generator seed to 2.

function Y = randnUsingSeed(sz1,sz2,resetGenerator)
if resetGenerator
    gpucoder.rng(2);
end
Y = gpucoder.randn(sz1,sz2);
end

Generate GPU code from randnUsingSeed.

cfg = coder.gpuConfig("mex");
codegen randnUsingSeed.m -config cfg -args {0,0,false}

Generate a 5-by-4 matrix by using the generated MEX function randnUsingSeed_mex. Specify the resetGenerator argument as true.

Y = randnUsingSeed_mex(5,4,true)

Y = 5×4
-1.3322	0.2694	-1.5536	1.9227
0.1524	-2.3717	-0.2757	0.3150
-0.2752	1.1509	1.8704	0.9156
0.3052	-0.2633	-0.3970	-0.7013
-0.1619	2.0606	1.0849	0.1884

Generate another 5-by-4 matrix without resetting the generator. The output is different.

Y = randnUsingSeed_mex(5,4,false)

Y = 5×4
-1.2467	0.6683	0.2895	-0.3010
-1.3132	0.3259	-1.2485	0.3757
0.9530	0.6921	0.0877	1.8518
-0.8706	0.4449	1.1088	-1.3921
-1.0971	-0.8106	1.1149	-1.1833

To repeat the sequence of random numbers, call randnUsingSeed_mex and reset the generator. The output matches the first matrix generated by randnUsingSeed_mex.

Y = randnUsingSeed_mex(5,4,true)

Y = 5×4
-1.3322	0.2694	-1.5536	1.9227
0.1524	-2.3717	-0.2757	0.3150
-0.2752	1.1509	1.8704	0.9156
0.3052	-0.2633	-0.3970	-0.7013
-0.1619	2.0606	1.0849	0.1884

Input Arguments

seed — Random number seed

nonnegative integer | "shuffle"

Random number seed, specified as a nonnegative integer less than 2^32 or as "shuffle".

• Specify seed as an integer, such as 1, to use that seed for the random number generator.

• Specify seed as "shuffle" to set the seed based on the current time.

Note

Using gpucoder.rng with the "shuffle" option sets the seed by using the current time in seconds. To set a new seed with the "shuffle" option, use gpucoder.rng at least one second after the last call to gpucoder.rng("shuffle").

Example: gpucoder.rng(23)

Example: gpucoder.rng("shuffle")

More About

Reproducibility

To generate a repeatable sequence of random numbers, specify the seed in your entry-point function by using gpucoder.rng("default") or gpucoder.rng(seed). Initializing the random number generator with the same seed causes the random number generator to produce the same results.

Alternatively, to generate a different sequence of random numbers on subsequent runs, set the generator seed based on the current time by using the "shuffle" option.

Extended Capabilities

GPU Code Generation
Generate CUDA® code for NVIDIA® GPUs using GPU Coder™.

Version History

Introduced in R2026b