mkl_somatadd does nothing

ddavobsc — Fri, 11 Apr 2025 10:59:56 GMT

Perhaps I am not understanding correctly how it should be used, but mkl_?omatadd is not working for me.

I have tried with this test:

auto A = new float[4] {1,2,3,4}; auto B = new float[4] {1,2,3,4}; auto C = new float[4] {-1,-1,-1,-1}; auto E = new float[4] {2,4,6,8}; mkl_somatadd( CblasRowMajor, CblasNoTrans, CblasNoTrans, 2, 2, 1.0, A, 2, 1.0, B, 2, C, 2 ); ASSERT_THAT( std::vector<float>(A, A+2*2), testing::Pointwise(testing::FloatEq(), std::vector<float>(E, E+4)) );

The C matrix is not modified at all. After going through mkl_somatadd, C is still {-1, -1, -1, -1}

Re: mkl_somatadd does nothing

Fengrui — Mon, 14 Apr 2025 04:33:38 GMT

Hi,

Were there any errors captured by oneMKL error handler?

It might be caused by the wrong setting of the first three parameters. I tried

mkl_somatadd( 'R', 'N', 'N', 2, 2, 1.0, A, 2, 1.0, B, 2, C, 2 );

and got C={2, 4, 6, 8}.

Thanks,

Fengrui

topic Re: mkl_somatadd does nothing in Intel® oneAPI Math Kernel Library

mkl_somatadd does nothing

Re: mkl_somatadd does nothing