When I use cublasSgemm(handle, CUBLAS_OP_N, CUBLAS_OP_N, m, n, k, alpha, A, lda, B, ldb, beta, C, ldc); for matrix multiplication, I see kernels ending with _nn which means both were not transposed.
I would like to see _nt kernels and I want to know can I feed the matrices to get that. Any comment is appreciated.