CalcPi

An attempt to calculate Pi accurately and efficiently.

Branches:

Solution 1

GlobalSharedSum => Every time a sum is calculated, it is immediately added to the SharedSum (which is synchronized). Not very efficient

DelayedSharedSum => The local sums of each thread group are summed up locally and only when the group is finished does its sum get added to the SharedSum. More efficient than the above

ReductionSum => Local sums are stored in an array on which a reduction algorithm is ran (no tree).

Solution 2

DitrSum => Distributed model in which the clients are served one after another.

DistrSumMultConns => Distributed model in which the clients are served "at the same time" (Multi-threaded).