Fixes in non-reduce implementation for MPI #9
Description
I found certain bugs when I looked at the code for benchmarking MPI-only against easyLambda.
- Why every process is sending the results to the 0th process N number of times here?
- Why 0th process is receiving from 0th process as well, here? I think this behavior is implementation defined and might result in forever wait (I think it does with intel implementation).
- Doesn't received[0] contain garbage value here?
- Why seed is fixed for all the processes. I understand this is just an example and this issue must at least be pointed out in comments for the readers to know.
Since, I've fixed these issues and have tested for correctness, I will send a PR if the above points make sense.