Parallel matrix-vector and matrix-matrix multiplication

Problem:
Implement parallel multiplication of matrix with vector and matrix with matrix
using pthread, Openmp and MPI and compare the execution time between them and the serial program.
I will try to run them using a dual core machine, PS3 and the cluster.

To simplify the description below let's confider:
Matrix X Vector = A x V
Matrix X Matrix = A X B

How to make it Parallel:
With pthreads :
The main process is responsible to split the data, create thread and collect the results.
Every thread will multiply a row of A with V or a column of B and return the result to the main process.

OpenMP :
With OpenMP is easy to write the sequential algorithm and leave OpenMP to make the for statements parallel.

MPI :
With MPI is similar to the pthreads. The processor with id = 0 shares data to the other processors and collects the results.
For A x V a processor receives a row of A and V and multiplies them.
For A X B a processor receives (from processor 0) the number of times it will have to run.
After that in a for statement it receives a row of A and a column of B, make multiplication and sends the result back to processor 0.
Parallelization with MPI main problem:
How to split the data and synchronize processors to avoid useless data transfer.

Furthermore I may implement the A X B multiplication with the Strassen algorithm (probably with the three ways above) and also compare the execution time.