How to optimize

[FredCabral]

Hi all !

I'm new using Intel MKL routines. I have the following function that multiplies two dense matrices:

void multiplyMatrices(double* A, double* B, double* C, int N) {

double alpha = 1.0, beta = 0.0;

cblas_dgemm(CblasRowMajor, CblasNoTrans, CblasNoTrans,

N, N, N, alpha, A, N, B, N, beta, C, N);

}

I compile with the followin flags

g++ -std=c++11 -O3  -I/opt/intel/oneapi/2022/mkl/latest/include  c laicoMult.cpp -L/opt/intel/oneapi/2022/mkl/latest/lib/intel64 -Wl,-rpath,/opt/intel/oneapi/mkl/2024.0/lib/intel64 -lmkl_intel_lp64 -lmkl_sequential -lmkl_core -lpthread -lm -ldl -fopenmp -march=native -fopt-info-vec -ffast-math -ftree-vectorize -DARMA_DONT_USE_WRAPPER -o laicoMult

At the moment I'm using a Intel(R) Core(TM) i7-10700 CPU @ 2.90GHz processor with 16GB DDR4.

The code is pretty fast but I'd like to know if there is something I could do to make this code faster.

Can anyone help me ?

Thanks in advance !

[Gennady_F_Intel · Accepted solution]

yes, you could link against threaded version of MKL instead of sequential (-lmkl_sequential ) once. Please check with MKL Linker Adviser ( https://www.intel.com/content/www/us/en/developer/tools/oneapi/onemkl-link-line-advisor.html ) how to link against OpenMP or TBB threaded runtimes. You also might check MKL Developer Guide where you could find out many of such examples.