Differences between C++ random number genertators

[Dobbsjr]

What is the difference between rand, srand, default_random_engine?

[NeatSquidYT]

They're just from different libraries. iirc, std::default_random_engine is the more up to date one to use

[Dobbsjr]

They're just from different libraries. iirc, std::default_random_engine is the more up to date one to use

Do they all have similar functionality?

[NeatSquidYT]

Do they all have similar functionality?

Yeah, I think so.

 

EDIT: S rand sets the seed, whereas rand doesn't http://stackoverflow.com/questions/15188049/rand-and-srand-in-c

[manon_gfx]

Acutally (almost) all random number generators are what we call pseudo-random. They work by setting a seed. Then do some calculations on that and return a number, the next time to generate a number, is will use the last generated number as a seed.

 

rand() is usually considered bad, since the distribution is not what we call 'uniform'. A uniform distribution is where all numbers have an equal chance to get generated.

 

C++ offers the <random> header file, containing better random number generators, like std::default_random_engine.

If you want to be able to recreate the same values with the same seed everytime, it is recommended that you don't use the std::default_random_engine, but a different one, or that you write your own.

If you need to generate a truly random number (not pseudo), there are ways to do that. x86 offers and extension that allows you to read a truly random number generated on the CPU (it had something to with crystals and stuff iirc). To use this include <immintrin.h> and call _rdrand16_step, _rdrand32_step, or _rdrand64_step. Note that these functions are very slow. The intel intrinsics guide says that it takes about 200 cycles on Ivy Bridge, and doesn't specify a cycle count on Haswell.

This video has a great explanation on how pseudo random number generators work: