Calculating the Sum of a Vector

What is the fastest way to add the elements of a std::vector? To get the answer, I fill a std::vector with one hundred million arbitrary but uniformly distributed numbers between 1 and 10. The task is to calculate the sum of the numbers in various ways. I use the performance of a single threaded addition as the reference execution time. I discuss atomics, locks, thread-local data and tasks.

Let’s start with the single threaded scenario.

Single Threaded addition of a Vector

The straightforward strategy is it to add the numbers in a range-based for loop.

Range-based for Loop

The summation takes place in line 27.

Summation of a vector in a range-based for loop

 1 // calculateWithLoop.cpp
 2 
 3 #include <chrono>
 4 #include <iostream>
 5 #include <random>
 6 #include <vector>
 7 
 8 constexpr long long size = 100000000;
 9 
10 int main(){
11 
12   std::cout << std::endl;
13 
14   std::vector<int> randValues...