This project is read-only.

ampblas_sgemm, ampblas_dgemm do not appear to work for me

Dec 14, 2013 at 2:32 AM
hello all.

i have a not been successful getting ampblas_sgemm and ampblas_dgemm to work. i am confident that my call is correct, because the correspondong call to intel's MKL sgemm and dgemm work coorectly. my calls are below; for reference, my matrices are row major. below,

A has dimension 1 x nft
B has dimension nft x nft
C has dimension 1 x nft

intel mkl function call:

dgemm("n", "n", &onei, &nft, &nft, &onef, A, &onei, B, &nft, &zerof, C, &onei);

ampblas_dgemm function call

ampblas_dgemm(AmpblasRowMajor, AmpblasNoTrans, AmpblasNoTrans, onei, nft, nft, onef, A, onei, B, nft, zerof, C, onei);

as stated before, intel mkl dgemm works correctly, but ampblas_dgemm gives the following error:

** On entry to ampblas_dgemm, parameter number 11 had an illegal value,

where parameter 11 is the leading dimension of matrix B, which is veritably nft. do you have any ideas why this might be the case?