Hi Cenk,
thanks for bringing this up!
Can you add the POSIX perspective to the rand* calls? I guess it would be beneficial to remain close to what POSIX describes for an API.
Cheers, Thomas
Hey,
oh sure, here it follows:
The ISO C standard rand() and srand() functions allow per-process
pseudo-random streams shared by all threads. Those two functions need
not change, but there has to be mutual-exclusion that prevents
interference between two threads concurrently accessing the random
number generator.
With regard to rand(), there are two different behaviors that may be
wanted in a multi-threaded program:
1. A single per-process sequence of pseudo-random numbers that is shared
by all threads that call rand()
2. A different sequence of pseudo-random numbers for each thread that
calls rand()
This is provided by the modified thread-safe function based on whether
the seed value is global to the entire process or local to each thread.
See: http://pubs.opengroup.org/onlinepubs/9699919799/functions/rand.html
The above prose makes me wonder: our current global-state-PRNG-implementations do AFAIK not provide mutual exclusion.. So, two threads accessing the PRNG could get in each other's way (?). And this is not just about `random_init()`, but all calls that modify an existing PRNG sequence, such as a simple `rand()`.
Just as info: in the above link it says:
The drand48() and random() functions provide much more elaborate
pseudo-random number generators.
and those do actually allow to intuitively set custom states, or do mutual exclusion: http://pubs.opengroup.org/onlinepubs/9699919799/functions/random.html http://pubs.opengroup.org/onlinepubs/9699919799/functions/drand48.html
Cheers, Cenk