Good summary of results.

1: OK

2: OK

3: OK
task3.cpp:32
non initialzed memory for thread 0 (and the other threads)

4: OK

5:
CRS_Matrix::GetDiag  non parallel
CRS_Matrix::CalculateLaplace: 
  "omp critical" is slower than an "omp atomic" for the +=  [worth to test that]