These parallel implementations of map and fold are similar in several aspects:
- They both fall back to a sequential version if the number of elements in the range is smaller than 10,000.
- They both start the same number of threads. These threads are determined using the static function std::thread::hardware_concurrency(), which returns the number of concurrent threads supported by the implementation. However, this value is rather a hint than an accurate value and should be used with that in mind.
- No shared data is used to avoid synchronization of access. Even though all the threads work on the elements from the same range, they all process parts of the range that do not overlap.
- Both these functions are implemented as function ...
