[Mlir-commits] [mlir] Remove assertion in ExpandShapeOp::build (PR #91361)

[Benoit Jacob]

https://github.com/bjacob created https://github.com/llvm/llvm-project/pull/91361

Unblocking downstream integrate where an expected-to-fail test was expecting this to be a runtime verifier error, not a compiler crash: https://github.com/llvm/torch-mlir/pull/3279.

>From 8990e76df4a4cfd969805e9044274d996d9c90a6 Mon Sep 17 00:00:00 2001
From: Benoit Jacob <jacob.benoit.1 at gmail.com>
Date: Tue, 7 May 2024 13:07:42 -0400
Subject: [PATCH] no assert in ExpandShapeOp::build

---
 mlir/lib/Dialect/Tensor/IR/TensorOps.cpp | 10 ++++++----
 1 file changed, 6 insertions(+), 4 deletions(-)

diff --git a/mlir/lib/Dialect/Tensor/IR/TensorOps.cpp b/mlir/lib/Dialect/Tensor/IR/TensorOps.cpp
index 4c65045084dc..7a13f7a7d135 100644
--- a/mlir/lib/Dialect/Tensor/IR/TensorOps.cpp
+++ b/mlir/lib/Dialect/Tensor/IR/TensorOps.cpp
@@ -1676,10 +1676,12 @@ void ExpandShapeOp::build(OpBuilder &builder, OperationState &result,
   auto tensorResultTy = cast<RankedTensorType>(resultType);
   FailureOr<SmallVector<OpFoldResult>> outputShape = inferOutputShape(
       builder, result.location, tensorResultTy, reassociation, inputShape);
-  // Failure of this assertion usually indicates presence of multiple
-  // dynamic dimensions in the same reassociation group.
-  assert(succeeded(outputShape) && "unable to infer output shape");
-  build(builder, result, tensorResultTy, src, reassociation, *outputShape);
+  SmallVector<OpFoldResult> outputShapeOrEmpty;
+  if (succeeded(outputShape)) {
+    outputShapeOrEmpty = *outputShape;
+  }
+  build(builder, result, tensorResultTy, src, reassociation,
+        outputShapeOrEmpty);
 }
 
 SmallVector<AffineMap, 4> CollapseShapeOp::getReassociationMaps() {