[Mlir-commits] [llvm] [mlir] [MLIR] Generalize expand_shape to take shape as explicit input (PR #69267)

[llvmlistbot at llvm.org]

================
@@ -757,7 +757,10 @@ pushDownUnPackOpThroughExpandShape(tensor::UnPackOp unPackOp,
   ArrayRef<int64_t> innerDimsPos = unPackOp.getInnerDimsPos();
   ArrayRef<int64_t> outerDimsPerm = unPackOp.getOuterDimsPerm();
 
-  ArrayRef<int64_t> dstShape = expandOp.getType().getShape();
+  auto expandTy = expandOp.getType().dyn_cast<RankedTensorType>();
----------------
MaheshRavishankar wrote:

Why did you have to change this? This should have always been a ranked tensor type. If so then something is off with the pack op. Fine to leave as is, just noticing.

https://github.com/llvm/llvm-project/pull/69267