} else {
        // Generalize differing axis lengths to "variable"/"unknown".
        for (int axis_index = 0; axis_index < combined_shape.dims();
             ++axis_index) {
          int64_t axis_length = combined_shape.dim_size(axis_index);
          if (axis_length != component_shape.dim_size(axis_index)) {
            axis_length = -1;
          }
          TF_RETURN_IF_ERROR(

// Return a 1-D INT32 tensor containing a single value 1.
TFE_TensorHandle* TestAxisTensorHandle(TFE_Context* ctx);

// Return an op taking minimum of `input` long `axis` dimension.
TFE_Op* MinOp(TFE_Context* ctx, TFE_TensorHandle* input,
              TFE_TensorHandle* axis);

// If there is a device of type `device_type`, returns true
// and sets 'device_name' accordingly.
// `device_type` must be either "GPU" or "TPU".

        // Non-parallel tensors from _EagerConst/tf.constant are implicitly
        // broadcast, i.e. set as the input to each parallel operation. This
        // allows code like "tf.constant(1.)" or "tf.reduce_sum(..., axis=1)"
        // (where the value starts on the host), without allowing other implicit
        // copies/broadcasts. Other implicit copies may be supported eventually,

  // Returns number of elements across all dimensions.
  virtual Status NumElements(int64_t* num_elements) const = 0;
  // Returns size of specified dimension
  //
  // -1 indicates an unknown axis length; this is unreachable for most standard
  // ImmediateExecutionTensorHandles, but comes up for example when computing
  // the shape of a parallel tensor with component shapes differing across
  // devices.

# A tape built on top of unified execution APIs.
load("//tensorflow/core/platform:rules_cc.bzl", "cc_library")

package(
    # copybara:uncomment default_applicable_licenses = ["//tensorflow:license"],
    licenses = ["notice"],
)

cc_library(
    name = "tape_context",
    srcs = ["tape_context.cc"],
    hdrs = [
        "tape_context.h",
    ],
    visibility = [
        "//tensorflow:internal",
    ],
    deps = [

    // the purposes of running this graph in a session. Thus, we don't
    // record the source node as being modified.
    RecordMutation(graph, *dst.oper, "updating input tensor");
  }
}

// Apis that are corresponding to python c api. --------------------------

void TF_AddOperationControlInput(TF_Graph* graph, TF_Operation* op,
                                 TF_Operation* input) {

  TFE_TensorHandle* axis = TestAxisTensorHandle(ctx);
  TFE_Op* minOp = MinOp(ctx, input, axis);
  TFE_TensorHandle* retvals[1] = {nullptr};
  int num_retvals = 1;
  TFE_Execute(minOp, &retvals[0], &num_retvals, status);
  EXPECT_EQ(TF_OK, TF_GetCode(status)) << TF_Message(status);
  TFE_DeleteOp(minOp);
  TFE_DeleteTensorHandle(input);
  TFE_DeleteTensorHandle(axis);
  ASSERT_EQ(1, num_retvals);

learning and neural networks. However, the framework is versatile enough to be
used in other areas as well.

TensorFlow provides stable [Python](https://www.tensorflow.org/api_docs/python)
and [C++](https://www.tensorflow.org/api_docs/cc) APIs, as well as a
non-guaranteed backward compatible API for
[other languages](https://www.tensorflow.org/api_docs).

Keep up-to-date with release announcements and security updates by subscribing
to

      context.get(), components, device_name, status.get());
  ASSERT_EQ(TF_GetCode(status.get()), TF_OK) << TF_Message(status.get());
  int num_axes = TFE_TensorHandleNumDims(combined_value.get(), status.get());
  ASSERT_EQ(TF_GetCode(status.get()), TF_OK) << TF_Message(status.get());
  EXPECT_EQ(num_axes, 1);
}

TEST(PARALLEL_DEVICE, TestNestedParallelDevices) {
  std::unique_ptr<TF_Status, decltype(&TF_DeleteStatus)> status(

struct TF_VariableInfo {
  TF_VariableInfo() = delete;
  // TF_VariableInfo is constructed here by TensorFlow, and will be passed to
  // plugin as a opaque pointer. Plugin will need to call C APIs below to
  // operate on TF_VariableInfo (such as allocate temp tensor for the `var` held
  // by the underlying tensorflow::VariableInfo.
  TF_VariableInfo(int index, const std::string& name, tensorflow::Var* var) {