How does std::random_device generate "true" random numbers

[abhishekm71]

I just read on this website that the function std::random_device() is not like other 'generators' that generate pseudo-random numbers.

It generates true stochastic (non-deterministic) random numbers based on the distribution, if it is present.

I'm just curious as to how such a function might be implemented internally. Also, why isn't it possible for all environments?

Anyone has any idea?

Thanks

[Lachlan Easton]

It will generally need to be connected to a hardware device or access true random numbers from an online service. I've heard of such devices measuring radioactive decay to retrieve random numbers for example. If a device is unavailable random_device will fallback to a pseudo random generator.

[MiiNiPaa]

I'm just curious as to how such a function might be implemented internally

For modern CPU this can be used for source of stochastic numbers:
http://stackoverflow.com/a/18004959/3410396

why isn't it possible for all environments?

Some enviroments does not have reliable source of non-deterministic numbers.

[abhishekm71]

Thanks for your replies @LachlanEaston and @MiiNiPaa.

So does the code in the function implementation look something like:

if (CPU is compatible)
  - generate true random number
else
  - fallback to pseudo random generator

[Cubbi]

Close to that, except there's also a "fallback to OS support" intermediate stage

See, for example, the GCC implementation, where the fallback from the Intel CPU command RDRND is to /dev/urandom, although the caller of random_device can opt for /dev/random: https://gcc.gnu.org/viewcvs/gcc/trunk/libstdc%2B%2B-v3/src/c%2B%2B11/random.cc?view=markup#l77 -- on Linux this "gathers environmental noise from device drivers and other sources"

[abhishekm71]

Aah, ok..

So the function random_device first tries to 'delegate' the random number generation to CPU. If it fails, then to the OS, and finally to the psuedo random generator.

Thanks for the link!