> On 12 Aug 2022, at 18:24, David Carter via Std-Discussion <std-discussion_at_[hidden]> wrote:
>
> It's something like this:
>
> for (size_t i=0; i < numThreads; ++) {
> threadPool_.push_back(std::thread(&ClassName::ClassFn, this, arg1, arg2, arg3));
> }
> . . .
> queue work
> . . .
> condition_variable stuff
>
> threadPool_ is a private member of ClassName and of type std::vector<std::thread>

I use a lambda to pass to the thread-pool, below shown with Boehm GC registration. The latter puts locks on each allocation, so may not be thread effective. The function eval_thread then picks tasks off a priority queue.

  data& C::eval_main(D& dr, size_type n) {
    std::vector<std::thread> threads;

    // Create up to N threads:
    for (int k = 0; k < N; ++k)
      threads.push_back(std::thread(
        // Lambda wrap for 'eval_thread' with GC register/unregister:
        [this, &dr, n] {
          // Register thread with GC:
          GC_stack_base sb;
          GC_get_stack_base(&sb);
          GC_register_my_thread(&sb);

          eval_thread(dr, n);

          // Unregister thread with GC:
          GC_unregister_my_thread();
        }));

    // Joining threads.
    for (auto& i: threads)
      i.join();

    return data_;
  }