{"id":16865,"date":"2025-09-05T16:41:19","date_gmt":"2025-09-05T20:41:19","guid":{"rendered":"https:\/\/blogs.mathworks.com\/pick\/?p=16865"},"modified":"2025-09-09T05:21:13","modified_gmt":"2025-09-09T09:21:13","slug":"benchmark-random-number-generators-in-matlab-with-comparerng","status":"publish","type":"post","link":"https:\/\/blogs.mathworks.com\/pick\/2025\/09\/05\/benchmark-random-number-generators-in-matlab-with-comparerng\/","title":{"rendered":"Benchmark Random Number Generators in MATLAB with CompareRNG"},"content":{"rendered":"
\r\n
Mike<\/a>'s Pick this week is CompareRNG<\/a> by Ben Tordoff<\/a><\/div>\r\n
Let's see how long it takes to generate 20 million random numbers on my laptop<\/div>\r\n
\r\n
\r\n
tic;vec=rand(1,20000000);toc<\/span><\/div>\r\n
\r\n
\r\n
Elapsed time is 0.077252 seconds.<\/div>\r\n<\/div>\r\n<\/div>\r\n<\/div>\r\n<\/div>\r\n
That's pretty fast but I can do even better by switching to a different algorithm<\/div>\r\n
\r\n
\r\n
rng(\"simdTwister\"<\/span>)<\/span><\/div>\r\n<\/div>\r\n
\r\n
tic;vec=rand(1,20000000);toc<\/span><\/div>\r\n
\r\n
\r\n
Elapsed time is 0.038272 seconds.<\/div>\r\n<\/div>\r\n<\/div>\r\n<\/div>\r\n<\/div>\r\n
MATLAB contains several high quality random number generation algorithms and the one it uses is controlled by the rng<\/span><\/a> function as I show here. It turns out that I can do even better than the above!<\/div>\r\n
\r\n
\r\n
rng(\"threefry\"<\/span>)<\/span><\/div>\r\n<\/div>\r\n
\r\n
tic;vec=rand(1,20000000);toc<\/span><\/div>\r\n
\r\n
\r\n
Elapsed time is 0.026482 seconds.<\/div>\r\n<\/div>\r\n<\/div>\r\n<\/div>\r\n<\/div>\r\n
Showing these three algorithms is the beginning of a deep<\/span> rabbit hole that I go down during a course that I give on High Performance Computing in MATLAB. It turns out, that the fastest algorithm is machine dependent and I discovered this while asking people to try something like the above and to shout out which was the best algorithm on their machine.<\/div>\r\n

Which is the fastest random number generation algorithm on your machine?<\/h3>\r\n
This is where today's pick of the week comes in. Developed by MathWorker, Ben Tordoff<\/a>, CompareRNG<\/a> runs all of the high quality random number generation algorithms in MATLAB and shows you which is best on your machine. Here's the result for my Apple M2 Mac<\/div>\r\n
\"\"<\/div>\r\n
and here is the result for my aging Intel desktop<\/div>\r\n
\"\"<\/div>\r\n
in both cases, the 'Threefry'<\/span> algorithm is the fastest but this isn't always the case. When I've been on the road with this benchmark, some people find that 'simdTwister'<\/span> or 'philox'<\/span> is better, for example.<\/div>\r\n
The app also benchmarks GPU random number generators and the difference between those and the CPU-based generators is enormous!<\/div>\r\n
\"\"<\/div>\r\n

Parallelization and random number generation algorithms<\/h3>\r\n
So what is it that makes the fastest random number generation algorithm machine dependent? Part of the answer is 'parallelization'. Some algorithms (e.g. 'twister'<\/span> and 'simdTwister'<\/span>) only use one CPU core, no matter how many you actually have. Others, such as 'threefry'<\/span> run in parallel and the results can be very impressive at high core counts.<\/div>\r\n
This is where Ben's second new entry comes in useful. RNGMultiThreading<\/span><\/a> show's how the performance of each algorithm varies as we increase the number of CPU cores. Here are the results on my Mac<\/div>\r\n
\"\"<\/div>\r\n
We now get some insight into why we have machine dependent results.'philox'<\/span> and 'threefry'<\/span> does great when you use a lot of CPU cores but is not so great at lower core counts.<\/div>\r\n
Give these new apps a try on your own machines and see which of these algorithms are best for you.<\/div>\r\n<\/div>\r\n