Performance deep dive

In this article, we'll find out how to most efficiently calculate if a year is a leap year in C++.

The approaches page lists two idiomatic approaches and mentions three further ideas to solving this exercise:

1. Using a boolean chain with best use of short-circuiting
2. Using a ternary operator
3. Using the chrono library (requires C++20)
4. Using the boost libraries
5. Using a boolean chain with worst use of short-circuiting as comparison.

Benchmarks

To benchmark the approaches, we wrote a small benchmark application using Google's benchmark library.

The benchmark prepares a vector of 1000 random elements in the range of 1400 to 9999 (limited by the boost library). Each approach implementation is called and timed with those same 1000 vector elements. The benchmark is repeated 100 times via the --benchmark_repetitions flag.

The aggregated runs give the following results:

You can see that the boolean chain has the lowest average cpu time. On the test system the ternary approach was a few nanoseconds slower, which translates to about 27 %. As expected the worst-case chaining is the slowest approach. Calling boost::gregorian::gregorian_calendar::is_leap_year is almost as slow, probably because a boost::gregorian::gregorian_calendar::year object is constructed for the test.

Here is a box-plot if you want more statistical information:

This article discusses very low level differences that might be altered by the compiler or even the processor's own optimizations. While it is interesting to compare functions in this detail on a theoretical level, they are seldomly the bottleneck at runtime.