typedef struct TF_StringStream TF_StringStream;
typedef struct TF_Thread TF_Thread;

typedef struct TF_ThreadOptions {
  // Thread stack size to use (in bytes), zero implies that the system default
  // will be used.
  size_t stack_size;

  // Guard area size to use near thread stacks to use (in bytes), zero implies
  // that the system default will be used.
  size_t guard_size;

tensorflow::ServerDef GetServerDef(const tensorflow::string& job_name,
                                   int num_tasks);

// Create a ServerDef with job name "localhost" and add `num_tasks` tasks in it.
tensorflow::ServerDef GetServerDef(int num_tasks);

// Create a multi-client ServerDef with the given `job_name`, add `num_tasks`

// parallel device creates a parallel tensor `x` with `a` on the first of
// `underlying_devices` and `b` on the second. Running `a_unpacked, b_unpacked =
// TPUReplicatedOutput(input=x, num_replicas=2)` un-packs the parallel tensor
// into its components.
//
// The filled `device` struct and the allocated `device_info` struct may be
// passed to TFE_RegisterCustomDevice. The `device_name` arguments must match.

  // Set up distributed execution environment on local and remote tasks.
  // When `reset_context` is true, initialize new cluster context state based
  // on cluster configurations provided in `server_def`; otherwise, update
  // existing context state with the provided `server_def`. Contexts created
  // on remote tasks will be considered stale and garbage collected after
  // `keep_alive_secs` of inactivity.

// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.

#include <string.h>

#include "_cgo_export.h"

void
callback(void *f)
{
	// use some stack space
	volatile char data[64*1024];

	data[0] = 0;
	goCallback(f);
        data[sizeof(data)-1] = 0;
}

void
callGoFoo(void)
{
	extern void goFoo(void);
	goFoo();
}

void
IntoC(void)

                                                        TF_Status* status);

// Some TF ops need a step container to be set to limit the lifetime of some
// resources (mostly TensorArray and Stack, used in while loop gradients in
// graph mode). Calling this on a context tells it to start a step.
TF_CAPI_EXPORT extern void TFE_ContextStartStep(TFE_Context* ctx);

  std::vector<int64_t> tensor_stack;
  tensor_stack.reserve(target.size());
  for (auto t : target) {
    tensor_stack.push_back(t);
  }
  BackpropInitialState<BackwardFunction, TapeTensor> result;
  while (!tensor_stack.empty()) {
    int64_t tensor_id = tensor_stack.back();
    tensor_stack.pop_back();
    auto op_id_it = tensor_tape.find(tensor_id);

  virtual Status InputLength(const char* input_name, int* length) = 0;
  virtual Status OutputLength(const char* output_name, int* length) = 0;

  // Set stack trace to be used for potential async error reporting.
  virtual void SetStackTrace(ManagedStackTrace stack_trace) = 0;

  virtual const tensorflow::AbstractOpAttrs* GetOpAttrs() const = 0;
  virtual void AddAttrs(const AbstractOpAttrs* op_attrs) = 0;

                                                     TF_Buffer* buf,
                                                     TF_Status* status);

// Get GraphDebugInfo containing stack traces mapping to node names
TF_CAPI_EXPORT extern void TFE_ContextGetGraphDebugInfo(
    TFE_Context* ctx, const char* function_name, TF_Buffer* buf,
    TF_Status* status);

// Extracts a TF_Function from the context.

  // all tasks in the cluster.
  // This call internally coordinates with other tasks to initialize the eager
  // context and TF server for multi-client execution.
  virtual Status EnableCollectiveOps(const ServerDef& server_def) = 0;

  // Set a distributed manager that helps set up, update, and check liveness
  // of member tasks in the cluster.
  virtual void SetDistributedManager(