// that stops being the case, we can cap this func to a fixed number of
	// architectures being generated at once.

	tasks := []func(){
		genOp,
		genAllocators,
	}
	for _, a := range archs {
		a := a // the funcs are ran concurrently at a later time
		tasks = append(tasks, func() {
			genRules(a)
			genSplitLoadRules(a)
			genLateLowerRules(a)
		})
	}

    const std::map<int, const Tensor*>& resource_vars) {
  se::Stream* stream =
      ctx->op_device_context() ? ctx->op_device_context()->stream() : nullptr;
  Allocator* allocator = ctx->device()->GetAllocator({});

  // Computation output should always be a tuple.
  VLOG(2) << "Result tuple shape: " << output.on_host_shape().DebugString();
  VLOG(2) << "Result tuple shape (on device): "

}
#undef EXPECT_TF_SIZE

class DummyDevice : public DeviceBase {
 public:
  explicit DummyDevice(Env* env) : DeviceBase(env) {}
  Allocator* GetAllocator(AllocatorAttributes /*attr*/) override {
    return cpu_allocator();
  }
};

TEST(TestKernel, TestInputAndOutputCount) {
  const char* node_name = "InputOutputCounterKernel";
  const char* op_name = "BarOp";