[Getfem-users] parallelized Dot product of sparse vectors

[Tarek Elsayed]

Hi all,

I noticed when implementing the dot product between two sparse vectors in a multi-threaded for loop, that the efficiency of the multi-threading (as measured by the ratio between the CPU time and wall-time) is very bad. 

I tried to implement the dot product myself as below, in a multi-threaded way, but still with no improvement!

Any suggestions ?!

Code: 

float Dot(Svec v1, Svec& v2)   // Svec is the GMM Sparse vector

{

typename gmm::linalg_traits<Svec>::const_iterator its = gmm::vect_const_begin(v1);

typename gmm::linalg_traits<Svec>::const_iterator ite = gmm::vect_const_end(v1);

typename gmm::linalg_traits<Svec>::const_iterator it;

float result=0.0f; 

omp_set_num_threads(8);

#pragma omp parallel for default(shared)

for(it=its;it!=ite;it++)

{

 int i=it.index() ;

#pragma omp atomic 

 result += real( conj(v1[i]) * v2[i] );  

}

return result;

}