Is increased-strided DLTensor stored in column-major?

Questions
2

I found in wiki link, that by definition, a row-major tensor should have decreased strides.

So, A or B here can either be row-major or column-major, that’s why we need IsInPlaceTransposed . (Although in tvm, tensors are most commonly stored in row-major.)

Finally I understand that my confusion actually comes from the necessity of the arguments transa and transb in ./python/tvm/contrib/cblas.py.

Since A is DLTensor*, I think transa can be determined by A->strides:

  1. if A->strides[0] > A->strides[1], A is inferred to be row major, transa=False.

  2. otherwise, A is inferred to be column major, transa=True.

So as to transb.

PS.

In C interface of the CBLAS, the matmul interface is given as follows:

void cblas_sgemm(const CBLAS_LAYOUT Layout, const CBLAS_TRANSPOSE TransA,
                               const CBLAS_TRANSPOSE TransB, const MKL_INT M, const MKL_INT N,
                               const MKL_INT K, const float alpha, const float *A,
                               const MKL_INT lda, const float *B, const MKL_INT ldb,
                               const float beta, float *C, const MKL_INT ldc) NOTHROW;

we need TransA and TransB here because A and B here are just pointers instead of DLTensor*, there is no clue whether A is column-major or row-major.