Matplot with multithreading

[SAtacker]

Thank you for this wonderful library!
I was playing with it to plot around 5.3 MB of data and tried plotting gantt chart like stuff using rectangle but because the data is too much I had to make it multithreaded. It's not possible for me to share the code and data but it looks something like this.

std::mutex mut;

    auto do_work = [&](decltype(cpu_cnt) cnt) {
      for (auto const [id, st, dt] : cnt) {
        const std::lock_guard<std::mutex> lock(mut);
        auto r2 = rectangle(st, 4, dt, height);
        r2->fill(true);
        r2->color({0.4f, 0.f, 0.4f, 1.0f});
        // hold(on);
      }
    };

    int parts = cpu_cnt.size() / 8;
    for (int i = 0; i < 8; i++) {
      if ((i + 1) * parts < cpu_cnt.size()) {
        auto part = decltype(cpu_cnt)(cpu_cnt.begin() + i * parts,
                                      cpu_cnt.begin() + (i + 1) * parts);

        vec_threads.push_back(std::thread(do_work, part));
      } else {
        auto part =
            decltype(cpu_cnt)(cpu_cnt.begin() + i * parts, cpu_cnt.end());

        vec_threads.push_back(std::thread(do_work, part));
      }
    }

    grid(on);
    hold(off);
    axis(tight);
    if (argv[3] != nullptr)
      save(argv[3]);
    else
      save("Temp.png");

    for (auto &thread : vec_threads) {
      thread.join();
    }

Without the lock guard inside the plot it just fails. Is there any better way to plot?
Thanks again!

[alandefreitas]

Yes. You'll need to lock the figure to make the operation thread-safe. Making the figure thread-safe by default is a bad idea here.
You could look for bulk operations to minimize the number of interactions with the figure but that's not possible there because rectangles are basic primitives.
Considering how much contention (basically 100% contention) you have there, I'm not sure the multi-thread variant is really helping you.
What I would do is put the figure in quiet mode while plotting the rectangles and undo it in the end.

[SAtacker]

Thanks! I agree with you.