"Transpose", "ZerosLike", "OnesLike", "BiasAdd" /*PW + Broadcast*/,
      "BroadcastArgs", "BroadcastGradientArgs", "OneHot", "Concat", "ConcatV2",
      "ConcatOffset", "Const", "MirrorPad", "MirrorPadGrad", "Pack", "Pad",
      "PadV2", "Reverse", "ReverseV2", "ReverseSequence", "Slice", "Split",
      "SplitV", "StridedSlice", "StridedSliceGrad",
      "ResourceStridedSliceAssign", "Tile", "Transpose", "InvertPermutation",

// with a TR1 tuple implementation.  NVIDIA's CUDA NVCC compiler
// pretends to be GCC by defining __GNUC__ and friends, but cannot
// compile GCC's tuple implementation.  MSVC 2008 (9.0) provides TR1
// tuple in a 323 MB Feature Pack download, which we cannot assume the
// user has.  QNX's QCC compiler is a modified GCC but it doesn't
// support TR1 tuple.  libc++ only provides std::tuple, in C++11 mode,

  Append(tmp_body_results, core_tpu_res_im1);
  Append(tmp_body_results, non_tpu_res_i);
  Append(tmp_body_results, {C_im1, C_i, C_ip1});

  llvm::SetVector<Value> new_body_results;
  // This should pack the same as deduping code above.
  new_body_results.insert(tmp_body_results.begin(), tmp_body_results.end());
  auto new_body_return_types = GetValueTypes(new_body_results);

// FcntlFlock performs a fcntl syscall for the F_GETLK, F_SETLK or F_SETLKW command.
func FcntlFlock(fd uintptr, cmd int, lk *Flock_t) error {
	// struct flock is packed on z/OS. We can't emulate that in Go so
	// instead we pack it here.
	var flock [24]byte
	*(*int16)(unsafe.Pointer(&flock[0])) = lk.Type
	*(*int16)(unsafe.Pointer(&flock[2])) = lk.Whence
	*(*int64)(unsafe.Pointer(&flock[4])) = lk.Start

    tf.TPUReshardVariablesOp(%rvar, %default_format, %rstate)
  }
```
### `-tf-unroll-batch-matmul`

_Unroll TF BatchMatMul op into Reshape, Slice, MatMul, Pack ops._

### `-tf-verify-for-export`

_Verify module is suitable for export back to TF Graph_

Verifies whether all functions in module are of single tf_executor.graph and