Aniello - I am guessing that Yasin is looking for an algorithm that would be useful in translating existing algorithms into parallel algorithms. I wish him luck in finding such - because if he does, there is much $$ to be made from it.
I think that there are already projects about it, but it is quite challenging. There should a good code analysis (close to the human one) and then a good automatic rewriting of the code.
Parallelization is a huge domain in parallel programming;
Each serial algorithm to be parallelized must be at first assigned to an appropriate parralel computational model taking into account the complexity achieved in this model.
These are the challenges of the parallel algorithmic field.
using the algorithm translators from serial to parallel or parallel to parallel is not a good way in my sens. The main effort is to find the best parallel algorithm in terms of complexity and memory occupancy during its execution.
There are many parallel compilers but they aren't optimized in terms of use, they are strongly depending on the computer architecture.