std::notify_all_at_thread_exit, llvm libc++ implements.

gistfile1.cpp

template <class _Fp>
void*
__thread_proxy(void* __vp)
{
    __thread_local_data().reset(new __thread_struct);
    std::unique_ptr<_Fp> __p(static_cast<_Fp*>(__vp));
    (*__p)();
    return nullptr;
}

template <class _Fp>
thread::thread(_Fp __f)
{
    std::unique_ptr<_Fp> __p(new _Fp(__f));
    int __ec = pthread_create(&__t_, 0, &__thread_proxy<_Fp>, __p.get());
    if (__ec == 0)
        __p.release();
    else
        __throw_system_error(__ec, "thread constructor failed");
}

class __thread_specific_ptr
{
    pthread_key_t __key_;
}

template <class _Tp>
__thread_specific_ptr<_Tp>::__thread_specific_ptr()
{
    int __ec = pthread_key_create(&__key_, &__thread_specific_ptr::__at_thread_exit);
}
template <class _Tp>
void
__thread_specific_ptr<_Tp>::__at_thread_exit(void* __p)
{
    delete static_cast<pointer>(__p);
}

template <class _Tp>
void
__thread_specific_ptr<_Tp>::reset(pointer __p)
{
    pointer __p_old = get();
    pthread_setspecific(__key_, __p);
    delete __p_old;
}

class _LIBCPP_TYPE_VIS __thread_struct
{
    __thread_struct_imp* __p_;

    __thread_struct(const __thread_struct&);
    __thread_struct& operator=(const __thread_struct&);
public:
    __thread_struct();
    ~__thread_struct();

    void notify_all_at_thread_exit(condition_variable*, mutex*);
    void __make_ready_at_thread_exit(__assoc_sub_state*);
};

__thread_specific_ptr<__thread_struct>&
__thread_local_data()
{
    static __thread_specific_ptr<__thread_struct> __p;
    return __p;
}

class _LIBCPP_HIDDEN __thread_struct_imp
{
    typedef vector<__assoc_sub_state*,
                          __hidden_allocator<__assoc_sub_state*> > _AsyncStates;
    typedef vector<pair<condition_variable*, mutex*>,
               __hidden_allocator<pair<condition_variable*, mutex*> > > _Notify;

    _AsyncStates async_states_;
    _Notify notify_;

    __thread_struct_imp(const __thread_struct_imp&);
    __thread_struct_imp& operator=(const __thread_struct_imp&);
public:
    __thread_struct_imp() {}
    ~__thread_struct_imp();

    void notify_all_at_thread_exit(condition_variable* cv, mutex* m);
    void __make_ready_at_thread_exit(__assoc_sub_state* __s);
};

__thread_struct_imp::~__thread_struct_imp()
{
    for (_Notify::iterator i = notify_.begin(), e = notify_.end();
            i != e; ++i)
    {
        i->second->unlock();
        i->first->notify_all();
    }
    for (_AsyncStates::iterator i = async_states_.begin(), e = async_states_.end();
            i != e; ++i)
    {
        (*i)->__make_ready();
        (*i)->__release_shared();
    }
}

void
__thread_struct_imp::notify_all_at_thread_exit(condition_variable* cv, mutex* m)
{
    notify_.push_back(pair<condition_variable*, mutex*>(cv, m));
}