Multithreaded Summation: Using std::lock_guard

You may have already guessed that using a shared variable for the summation with four threads is not optimal; the synchronization overhead will outweigh the performance benefit. Let me show you the numbers. The questions I want to answer are still the same.

1. What is the difference in performance between the summation using a lock and an atomic?
2. What is the difference in performance between single threaded and multithreaded execution of std::accumulate?

The simplest way to make the thread-safe summation is to use a std::lock_guard.

$$$$

Using a std::lock_guard