diff --git a/mlir/lib/Conversion/ShapeToStandard/ShapeToStandard.cpp b/mlir/lib/Conversion/ShapeToStandard/ShapeToStandard.cpp
--- a/mlir/lib/Conversion/ShapeToStandard/ShapeToStandard.cpp
+++ b/mlir/lib/Conversion/ShapeToStandard/ShapeToStandard.cpp
@@ -49,6 +49,23 @@
   }
 };
 
+class GetExtentOpConversion : public OpConversionPattern<shape::GetExtentOp> {
+public:
+  using OpConversionPattern<shape::GetExtentOp>::OpConversionPattern;
+
+  LogicalResult
+  matchAndRewrite(shape::GetExtentOp op, ArrayRef<Value> operands,
+                  ConversionPatternRewriter &rewriter) const override {
+    shape::GetExtentOpOperandAdaptor transformed(operands);
+    auto shapeOfOp =
+        cast<shape::ShapeOfOp>(transformed.shape().getDefiningOp());
+    rewriter.replaceOpWithNewOp<DimOp>(op.getOperation(),
+                                       rewriter.getIndexType(), shapeOfOp.arg(),
+                                       transformed.dim());
+    return success();
+  }
+};
+
 class IndexToSizeOpConversion
     : public OpConversionPattern<shape::IndexToSizeOp> {
 public:
@@ -63,6 +80,21 @@
   }
 };
 
+class ShapeOfOpConversion : public OpConversionPattern<shape::ShapeOfOp> {
+public:
+  using OpConversionPattern<shape::ShapeOfOp>::OpConversionPattern;
+
+  LogicalResult
+  matchAndRewrite(shape::ShapeOfOp op, ArrayRef<Value> operands,
+                  ConversionPatternRewriter &rewriter) const override {
+    // Simply delete `shape_of`.
+    // The lowering of all operations that depends on a `!shape.shape` ensure
+    // correct lowering.
+    rewriter.eraseOp(op.getOperation());
+    return success();
+  }
+};
+
 class SizeToIndexOpConversion
     : public OpConversionPattern<shape::SizeToIndexOp> {
 public:
@@ -147,7 +179,9 @@
       BinaryOpConversion<AddOp, AddIOp>,
       BinaryOpConversion<MulOp, MulIOp>,
       FromExtentTensorOpConversion,
+      GetExtentOpConversion, 
       IndexToSizeOpConversion,
+      ShapeOfOpConversion, 
       SizeToIndexOpConversion,
       ToExtentTensorOpConversion>(ctx);
   // clang-format on
diff --git a/mlir/test/Conversion/ShapeToStandard/shape-to-standard.mlir b/mlir/test/Conversion/ShapeToStandard/shape-to-standard.mlir
--- a/mlir/test/Conversion/ShapeToStandard/shape-to-standard.mlir
+++ b/mlir/test/Conversion/ShapeToStandard/shape-to-standard.mlir
@@ -75,3 +75,27 @@
   // CHECK-NEXT: muli %[[LHS]], %[[RHS]] : index
   return
 }
+
+// -----
+
+// Lower `get_extent` such that is does not rely on the result of `shape_of`.
+// CHECK-LABEL: @get_extent
+// CHECK-SAME: (%[[ARG:.*]]: tensor<2x?xf32>, %[[IDX:.*]]: index) -> index
+func @get_extent(%arg : tensor<2x?xf32>, %idx : !shape.size) -> !shape.size {
+  // CHECK-NEXT: %[[RESULT:.*]] = dim %[[ARG]], %[[IDX]] : tensor<2x?xf32>
+  // CHECK-NEXT: return %[[RESULT]]
+  %shape = shape.shape_of %arg : tensor<2x?xf32>
+  %result = shape.get_extent %shape, %idx
+  return %result : !shape.size
+}
+
+// -----
+
+// Erase `shape_of` operation.
+// CHECK-LABEL: @shape_of
+func @shape_of(%arg : tensor<2x?xf32>) {
+  // CHECK-NOT shape_of
+  %shape = shape.shape_of %arg : tensor<2x?xf32>
+  return
+}
+