[Mlir-commits] [mlir] [mlir][sparse] implement lowering rules for ExtractIterSpaceOp. (PR #89143)

[Yinying Li]

================
@@ -0,0 +1,78 @@
+
+#include "Utils/CodegenUtils.h"
+#include "Utils/SparseTensorIterator.h"
+
+#include "mlir/Dialect/SCF/IR/SCF.h"
+#include "mlir/Dialect/SparseTensor/IR/SparseTensor.h"
+#include "mlir/Dialect/SparseTensor/Transforms/Passes.h"
+#include "mlir/Transforms/OneToNTypeConversion.h"
+
+using namespace mlir;
+using namespace mlir::sparse_tensor;
+
+void convertLevelType(SparseTensorEncodingAttr enc, Level lvl,
+                      SmallVectorImpl<Type> &fields) {
+  // Position and coordinate buffer in the sparse structure.
+  if (enc.getLvlType(lvl).isWithPosLT())
+    fields.push_back(enc.getPosMemRefType());
+  if (enc.getLvlType(lvl).isWithCrdLT())
+    fields.push_back(enc.getCrdMemRefType());
+  // One index for shape bound (result from lvlOp)
----------------
yinying-lisa-li wrote:

```suggestion
  // One index for shape bound (result from lvlOp).
```

https://github.com/llvm/llvm-project/pull/89143