fallback to our spmv when single strided vector

cusparse SpMM has the issue when SM >= 70 and nvcc >= 11.6

cuda/matrix/csr_kernels.cu

@@ -383,6 +383,14 @@ bool try_general_sparselib_spmv(std::shared_ptr<const CudaExecutor> exec,
         cusparse::destroy(vecb);
         cusparse::destroy(vecc);
     } else {
+#if CUDA_VERSION >= 11060
+        if (b->get_size()[1] == 1 && exec->get_major_version() >= 7) {
+            // cusparseSpMM seems to take the single strided vector as column
+            // major without considering stride and row major (SM >= 70 and
+            // cuda 11.6)
+            return false;
+        }
+#endif  // CUDA_VERSION >= 11060
         cusparseSpMMAlg_t alg = CUSPARSE_SPMM_CSR_ALG2;
         auto vecb =
             cusparse::create_dnmat(b->get_size(), b->get_stride(), b_val);