linalg.misc.mult_matvec protential bug for Fortran Matrix-vector element wise mulitplication

[yumais2]

Problem

Potential Bugs in cuda code of Fortran array. The case where input Matrix is Fortran array, and axis == 0. The kernel code seems to use the wrong indexing. Here is our workable version, please confirm.

        if axis == 0:
            row_kernel(x_gpu, a_gpu, out, n, m,
                       block=block, grid=grid, stream=stream,
                       shared=24*x_gpu.dtype.itemsize)

__global__ void opRowVecToMat(const ${type}* mat, const ${type}* vec, ${type}* out,
                                   const int n, const int m){
        const int tx = threadIdx.x;
        const int ty = threadIdx.y;
        const int tidx = blockIdx.x * blockDim.x + threadIdx.x;
        const int tidy = blockIdx.y * blockDim.y + threadIdx.y;

        extern __shared__ ${type} shared_vec[];

        if ((ty == 0) & (tidx < n))
            shared_vec[tx] = vec[tidx];
        __syncthreads();

        if ((tidy < m) & (tidx < n)) {
            out[tidy*n+tidx] = mat[tidy*n+tidx] ${binary_op} shared_vec[tx];
        }
    }

We haven't verify, but we think axis == 1 case for Fortran array is also buggy. The following are what we think how the code supposed to be like.

        elif axis == 1:
            col_kernel(x_gpu, a_gpu, out, n, m,
                       block=block, grid=grid, stream=stream,
                       shared=24*x_gpu.dtype.itemsize)

    __global__ void opColVecToMat(const ${type} *mat, const ${type} *vec, ${type} *out,
                                   const int n, const int m){
        const int tx = threadIdx.x;
        const int ty = threadIdx.y;
        const int tidx = blockIdx.x * blockDim.x + threadIdx.x;
        const int tidy = blockIdx.y * blockDim.y + threadIdx.y;

        extern __shared__ ${type} shared_vec[];

        if ((tx == 0) & (tidy < m))
            shared_vec[ty] = vec[tidy];
        __syncthreads();

        if ((tidy < m) & (tidx < n)) {
            out[tidy*n+tidx] = mat[tidy*n+tidx] ${binary_op} shared_vec[ty];
        }
    }

Environment

CUDA - 11.0.0 skcuda - 0.5.2 pycuda - 2020.1 python - 3. 7. 4 Linux - ubuntu- 18.04

[yuanz4]

Same problem