From: Matthias Springer <springerm@google.com>
Date: Tue, 4 May 2021 01:43:10 +0000 (+0900)
Subject: [mlir] Fix bug in TransferOpReduceRank when all dims are broadcasts
X-Git-Tag: llvmorg-14-init~7720
X-Git-Url: http://review.tizen.org/git/?a=commitdiff_plain;h=aa58281979710a1cee6f1635dc1074c4dfc44650;p=platform%2Fupstream%2Fllvm.git

[mlir] Fix bug in TransferOpReduceRank when all dims are broadcasts

TransferReadOps that are a scalar read + broadcast are handled by TransferReadToVectorLoadLowering.

Differential Revision: https://reviews.llvm.org/D101808
---

diff --git a/mlir/lib/Dialect/Vector/VectorTransforms.cpp b/mlir/lib/Dialect/Vector/VectorTransforms.cpp
index 2c8b3379..3804295 100644
--- a/mlir/lib/Dialect/Vector/VectorTransforms.cpp
+++ b/mlir/lib/Dialect/Vector/VectorTransforms.cpp
@@ -2970,6 +2970,9 @@ struct TransferOpReduceRank : public OpRewritePattern<vector::TransferReadOp> {
       return failure();
     SmallVector<int64_t> newShape = llvm::to_vector<4>(
         originalVecType.getShape().take_back(reducedShapeRank));
+    // Vector rank cannot be zero. Handled by TransferReadToVectorLoadLowering.
+    if (newShape.empty())
+      return failure();
     VectorType newReadType =
         VectorType::get(newShape, originalVecType.getElementType());
     ArrayAttr newInBounds =