[Mlir-commits] [mlir] [mlir][sparse] implement sparse_tensor.reorder_coo (PR #68916)

Yinying Li llvmlistbot at llvm.org
Thu Oct 12 10:28:53 PDT 2023


================
@@ -374,6 +376,24 @@ class SparseTensorStorage final : public SparseTensorStorageBase {
   /// Partially specialize lexicographical insertions based on template types.
   void lexInsert(const uint64_t *lvlCoords, V val) final {
     assert(lvlCoords && "Received nullptr for level-coordinates");
+    // TODO: get rid of this! canonicalize all-dense "sparse" array into dense
+    // tensors.
+    bool allDense = true;
+    for (DimLevelType lt : getLvlTypes()) {
+      if (!isDenseDLT(lt)) {
+        allDense = false;
+        break;
+      }
+    }
----------------
yinying-lisa-li wrote:

Maybe use std::all_of?

https://github.com/llvm/llvm-project/pull/68916


More information about the Mlir-commits mailing list