[Mlir-commits] [mlir] [MLIR] Add fusability query to TilingInterface (PR #166502)

[llvmlistbot at llvm.org]

================
@@ -359,6 +359,52 @@ struct LinalgOpTilingInterface
     /// Inline the op payload and store the result.
     return inlinePayload(builder, linalgOp, ivs, indexedValues);
   }
+
+  bool isOpFusableWithConsumerSlice(Operation *op, unsigned resultNumber,
+                                    ArrayRef<OpFoldResult> offsets,
+                                    ArrayRef<OpFoldResult> sizes) const {
+    return !cast<LinalgOp>(op).getShapesToLoopsMap();
+  }
+
+  bool isOpFusableWithProducerSlices(
+      Operation *op, ArrayRef<unsigned> operandNumbers,
+      ArrayRef<SmallVector<OpFoldResult>> allOffsets,
+      ArrayRef<SmallVector<OpFoldResult>> allSizes) const {
+
+    auto linalgOp = cast<LinalgOp>(op);
----------------
MaheshRavishankar wrote:

This is kind of based on how the ShapesToLoopsMap works. you could have dimensions used in any arbitrary AffineExpr, but the shapes of the loop are determined by those accesses that are accessed through AffineDimExpr. I am doing the same logic in that method https://github.com/llvm/llvm-project/blob/a7bf45a914f6e0d85d678aa5eb1864b35c0198e4/mlir/lib/Dialect/Linalg/Transforms/TilingInterfaceImpl.cpp#L156 . If you feel the documentation needs to be updated, please do so.

https://github.com/llvm/llvm-project/pull/166502