refactor _to_sparse_semi_structure()

aten/src/ATen/native/sparse/cuda/SparseSemiStructuredOps.cu

@@ -911,20 +911,20 @@ _to_sparse_semi_structured(const Tensor& dense) {
 #if defined(_MSC_VER) || (defined(CUDA_VERSION) && CUDA_VERSION < 11080)
   AT_ERROR(__func__, " : CUTLASS not supported");
   return std::make_tuple(Tensor{}, Tensor{});
-#else
+#elif defined(USE_ROCM)
   // Check dimensions of the dense matrix.
   TORCH_CHECK(dense.dim() == 2,
               __func__, " : Expected dense argument to be 2D tensor, got ",
               dense.dim(), " dims");
 
-#if defined(USE_ROCM)
   // Generate sparse tensor using cuSPARSELt compression
   auto sparse = torch._cslt_compress(dense);    
 
   // Extract the compressed data and metadata
   auto compressed_data = sparse.values();
   auto metadata = sparse.indices();
-
+  // Print to console that we are using hipSPARSELt
+  printf("Using hipSPARSELt for sparse semi-structured conversion\n");
   return std::make_tuple(compressed_data, metadata);
 
 #else  // Determine PyTorch datatype for the metadata matrix.
@@ -1052,7 +1052,6 @@ _to_sparse_semi_structured(const Tensor& dense) {
   return std::make_tuple(sparse_cpu.to(dense.device()),
                          meta_reordered_cpu.to(dense.device()));
 #endif
-#endif
 }
 
 }  // namespace at::native