potrf.cpp

Go to the documentation of this file.

#include <shark/LinAlg/BLAS/blas.h>
#include <shark/LinAlg/BLAS/kernels/potrf.hpp>
#include <shark/Core/Timer.h>
#include <iostream>
using namespace shark;
using namespace std;

template<class MatA, class Triang>
double benchmark(
    blas::matrix_expression<MatA, blas::cpu_tag> const& A,
    Triang
){
    double minTime = std::numeric_limits<double>::max();
    volatile double res = 0;
    for(std::size_t i = 0; i != 20; ++i){
        typename blas::matrix_temporary<MatA>::type Acopy =A;
        Timer time;
        blas::kernels::potrf<Triang>(Acopy);
        minTime = min(minTime,time.stop());
        res += max(Acopy);
    }
    return (1.0/3.0*A().size1()*A().size1()*A().size1())/minTime/1024/1024;
}

int main(int argc, char **argv) {
    std::size_t size = 128;
    std::cout<<"Mega Flops"<<std::endl;
    for(std::size_t iter = 0; iter != 10; ++iter){
        blas::matrix<double,blas::row_major> Arow(size,size);
        for(std::size_t i = 0; i != size; ++i){
            for(std::size_t j = 0; j != size; ++j){
                Arow(i,j)  = 0.1/size*i+0.1/size*j;
            }
            Arow(i,i) += 1000.0;
        }
        blas::matrix<double,blas::column_major> Acol = Arow;
        std::cout<<size<<"\t upper\t"<<benchmark(Arow,blas::upper())<<"\t"<< benchmark(Acol,blas::upper())<<std::endl;
        std::cout<<size<<"\t lower\t"<<benchmark(Arow,blas::lower())<<"\t"<< benchmark(Acol,blas::lower())<<std::endl;
        size *=2;
    }
}