if you are using C and GSL.
This, however, will not help if you are running distributed computations on a cluster, and you need to get an independent random sequence for each of the processes. The solution in this case is to generate a unique random seed for each of the processes. Adding the rank of the current job to the current time value works the best for me when using MPI. The solution looks like the following:
gsl_rng_set(r,time(NULL) + MPI::COMM_WORLD.Get_rank());
for C++, and
int rank;if you are using C.
gsl_rng_set(r,time(NULL) + rank);
If you are not using MPI, but rather run independent jobs through the execution queue, you cannot compute such a rank, but you can use hostname checksum / process id combination instead. It is extremely unlikely (I am speaking about the most UNIX architectures) that two instances of your program will be run on the same host within the same second and have exactly the same process id. If it is the case, you have to think a little, but you already have the general idea.