I am a beginner in MPI and have done some work in OPENMP type parallelization. I need your advice on the following. I want to my program in main thread to start a particular dlls subroutine calls in separate processors? Is this a correct architecture to go with ?
Also at different points in my main program thread I should be able to access a particular processor instance of dll subroutines. I.e., main thread starts multiple processer calls with different inputs, waits for completion of first phase of calls from all threads, then continues with its execution, then sends another call for second processing (received results) and so on. Pls let me know if you have done similar examples.
I would also be interested to see if using executables instead of dlls will give an easier and already proven option as well. So pls post any examples where one program called multiple copies of another and keeps the other processor copies to wait until you finalize all threads.