Choosing a PRNG Algorithm
For applications (such as physical simulation, machine learning, and games) that don't require a cryptographically secure algorithm, the java.util.random package provides multiple implementations of interface RandomGenerator that focus on one or more PRNG properties, which include speed, space, period, accidental correlation, and equidistribution.
Note:
As PRNG algorithms evolve, Java SE may add new PRNG algorithms and deprecate older ones. It's recommended that you don't use deprecated algorithms; they may be removed from a future Java SE release. Check if an algorithm has been deprecated by calling either the RandomGenerator.isDeprecated() or RandomGeneratorFactory.isDeprecated() method.Cryptographically Secure
For applications that require a random number generator algorithm that is cryptographically secure, use the SecureRandom class in the java.security package.
See The SecureRandom Class in Java Platform, Standard Edition Security Developer's Guide for more information.
General Purpose
For applications with no special requirements, L64X128MixRandom balances speed, space, and period well. It's suitable for both single-threaded and multithreaded applications when used properly (a separate instance for each thread).
Single-Threaded, High Performance
For single-threaded applications, Xoroshiro128PlusPlus is small, fast, and has a sufficiently long period.
32-Bit Applications
For applications running in a 32-bit environment and using only one or a small number of threads, L32X64StarStarRandom or L32X64MixRandom are good choices.
Multithreaded Applications with Static Threads
For applications that use many threads that are allocated in one batch at the start of computation, consider a jumpable generator such as Xoroshiro128PlusPlus or Xoshiro256PlusPlus or a splittable generator such as L64X128MixRandom or L64X256MixRandom. If your application uses only floating-point values from a uniform distribution where no more than 32 bits of floating-point precision is required and exact equidistribution is not required, then MRG32k3a, a classic and well-studied algorithm, may be appropriate.
Multithreaded Applications with Dynamic Threads
For applications that create many threads dynamically, perhaps through the use of spliterators, a splittable generator such as L64X128MixRandom or L64X256MixRandom is recommended.
If the number of generators created dynamically may be very large (millions or more), then using generators such as L128X128MixRandom or L128X256MixRandom will make it much less likely that two instances use the same state cycle.
Tuples of Consecutively Generated Values
For applications that use tuples of consecutively generated values, consider a generator that is k-equidistributed such that k is at least as large as the length of the tuples being generated. For example, the generator L64X256MixRandom is shown to be 4-equidistributed, which means that you can have a sequence of tuples that contain four values, and these tuples will be uniformly distributed (there’s an equal chance that any 4-tuple will appear in the sequence). It’s also shown that L64X1024MixRandom is 16-equidistributed.
Large Permutations
For applications that generate large permutations, consider a generator whose period is much larger than the total number of possible permutations; otherwise, it will be impossible to generate some of the intended permutations. For example, if the goal is to shuffle a deck of 52 cards, the number of possible permutations is 52! (52 factorial), which is approximately 2225.58, so it may be best to use a generator whose period is roughly 2256 or larger, such as L64X256MixRandom, L64X1024MixRandom, L128X256MixRandom, or L128X1024MixRandom.