diff --git a/mlir/include/mlir/Dialect/Vector/IR/VectorOps.td b/mlir/include/mlir/Dialect/Vector/IR/VectorOps.td
--- a/mlir/include/mlir/Dialect/Vector/IR/VectorOps.td
+++ b/mlir/include/mlir/Dialect/Vector/IR/VectorOps.td
@@ -311,6 +311,7 @@
   // TODO: Migrate to assemblyFormat once `AllTypesMatch` supports optional
   // operands.
   let hasCustomAssemblyFormat = 1;
+  let hasCanonicalizer = 1;
   let hasVerifier = 1;
 }
 
diff --git a/mlir/lib/Dialect/Vector/IR/VectorOps.cpp b/mlir/lib/Dialect/Vector/IR/VectorOps.cpp
--- a/mlir/lib/Dialect/Vector/IR/VectorOps.cpp
+++ b/mlir/lib/Dialect/Vector/IR/VectorOps.cpp
@@ -488,6 +488,30 @@
   return llvm::to_vector<4>(getVectorType().getShape());
 }
 
+namespace {
+struct ElideSingleElementReduction : public OpRewritePattern<ReductionOp> {
+  using OpRewritePattern::OpRewritePattern;
+
+  LogicalResult matchAndRewrite(ReductionOp reductionOp,
+                                PatternRewriter &rewriter) const override {
+    if (reductionOp.getVectorType().getDimSize(0) != 1)
+      return failure();
+    if (reductionOp.getAcc())
+      return failure();
+
+    rewriter.replaceOpWithNewOp<ExtractOp>(reductionOp, reductionOp.getType(),
+                                           reductionOp.getVector(),
+                                           rewriter.getI64ArrayAttr(0));
+    return success();
+  }
+};
+} // namespace
+
+void ReductionOp::getCanonicalizationPatterns(RewritePatternSet &results,
+                                              MLIRContext *context) {
+  results.add<ElideSingleElementReduction>(context);
+}
+
 //===----------------------------------------------------------------------===//
 // ContractionOp
 //===----------------------------------------------------------------------===//
diff --git a/mlir/test/Dialect/Vector/canonicalize.mlir b/mlir/test/Dialect/Vector/canonicalize.mlir
--- a/mlir/test/Dialect/Vector/canonicalize.mlir
+++ b/mlir/test/Dialect/Vector/canonicalize.mlir
@@ -1528,3 +1528,32 @@
   %1 = vector.extractelement %v[%i : i32] : vector<4xi32>
   return %1 : i32
 }
+
+// -----
+
+// CHECK-LABEL: func @reduce_one_element_vector
+//  CHECK-SAME: (%[[V:.+]]: vector<1xf32>)
+//       CHECK:   %[[S:.+]] = vector.extract %[[V]][0] : vector<1xf32>
+//       CHECK:   return %[[S]] : f32
+func @reduce_one_element_vector(%a : vector<1xf32>) -> f32 {
+  %s = vector.reduction <add>, %a : vector<1xf32> into f32
+  return %s : f32
+}
+
+// -----
+
+// CHECK-LABEL: func @reduce_one_element_vector
+//       CHECK: vector.reduction
+func @reduce_one_element_vector(%a : vector<1xf32>, %b: f32) -> f32 {
+  %s = vector.reduction <add>, %a, %b : vector<1xf32> into f32
+  return %s : f32
+}
+
+// -----
+
+// CHECK-LABEL: func @reduce_one_element_vector
+//       CHECK: vector.reduction
+func @reduce_one_element_vector(%a : vector<4xf32>) -> f32 {
+  %s = vector.reduction <add>, %a : vector<4xf32> into f32
+  return %s : f32
+}