//   SRC -> I
    //   SRC -> V
    //   Const -> Identity
    //   Const -> InplaceAdd [label="x"]
    //   I -> InplaceAdd [label="i"]
    //   V -> InplaceAdd [label="v"]
    //   InplaceAdd -> Identity [style=dotted]
    // }
    //
    // then the value produced by `Identity` is Const+I*V since InplaceAdd
    // modifies the tensor in place.  However, if we clone `Const` and turn the
    // graph into:

    "FakeQuantWithMinMaxVarsPerChannel",
    "FakeQuantWithMinMaxVarsPerChannelGradient",
    "Fill",
    "Fingerprint",
    "Gather",
    "GuaranteeConst",
    "Identity",
    "IdentityN",
    "InplaceAdd",
    "InplaceSub",
    "InplaceUpdate",
    "InvertPermutation",
    "ListDiff",
    "MatrixBandPart",
    "MatrixDiag",
    "MatrixDiagPart",
    "MatrixDiagPartV2",
    "MatrixDiagPartV3",

  {
    Output tr0 = ops::Transpose(on_gpu.WithOpName("tr0"), in0, perm);
    Output tr1 = ops::Transpose(on_gpu.WithOpName("tr1"), in1, perm);
  }

  Output in_place_add =
      ops::InplaceAdd(on_cpu.WithOpName("tr0"), perm,
                      ops::Placeholder(on_cpu.WithOpName("i"), DT_INT32), perm);

  std::unique_ptr<Graph> result;
  TF_ASSERT_OK(CloneConstantsForBetterClustering(root, &result));

  );

  let results = (outs);

  TF_DerivedOperandTypeAttr Tkey = TF_DerivedOperandTypeAttr<1>;
  TF_DerivedOperandTypeAttr Tval = TF_DerivedOperandTypeAttr<2>;
}

def TF_InplaceAddOp : TF_Op<"InplaceAdd", [Pure, TF_AllTypesMatch<["x", "y"]>]> {
  let summary = "Adds v into specified rows of x.";

  let description = [{
Computes y = x; y[i, :] += v; return y.
  }];

  let arguments = (ins