The implementation just proposed is only slightly different than what we did in the previous recipe. Threads were replaced with asynchronous functions, starting with std::async(), and results were made available through the returned std::future. The number of asynchronous functions that are launched concurrently is equal to the number of threads the implementation can support. This is returned by the static method std::thread::hardware_concurrency(), but this value is only a hint and should not be considered very reliable.
There are mainly two reasons for taking this approach:
- Seeing how a function implemented for parallel execution with threads can be modified to use asynchronous functions and, therefore, avoid lower-level ...
