Parallel Monte Carlo on Beowulf Cluster

My project on parallel computing went quite well.

I implemented 3-dimension embarassingly pleasingly parallel Monte Carlo integration on Beowulf cluster consisting of mainly 6 units i3-540/550 and 10 units 10 E7500 for some extra experiments.


The project code is montice.
The paper (Indonesian): here.


I used Darren Wilkinson's blog which is great reference material with some basic benchmark (here), in which his result showed poor figures due to most likely memory bottleneck on his i7-940XM (which is quite insane CPU on a laptop).

The speedup on 6 units i3 is here:

As the figure shows, single/core dual core performs just ideal and multithreading (3 or 4 cores for each box) only shows 70% performance improvement over the single/dual core operation.