Benchmark

Usage

Generate Benchmark Target

using factory function to make benchmark target.

there are several default factory functions:

• bp::real_time: real time benchmark
• bp::proc_time: process time benchmark, including user time and system time

auto target = bp::real_time(bench_func, pre_func, post_func);

Benchmark

using bp::bench function to benchmark.
rounds is the number of times to run the benchmark target.
d_meth is the method to use to calculate result. see Result

for multiple targets, the execution order will be shuffled randomly.

bp::bench(rounds, d_meth, target1, target2, ...);

Result

d_meth is used to calculate result.

there are serveral default methods:

• bp::avg: average of all rounds
• bp::min: minimum of all rounds
• bp::max: maximum of all rounds
• min_max: minimum and maximum of all rounds
• bp::median: median of all rounds
• bp::stddev: standard deviation of all rounds
• bp::sum: sum of all rounds
• bp::avg_stddev: average and standard deviation of all rounds
• bp::excl_avg: average of all rounds excluding outliers
• bp::full: all results

bp::bench returns array of result if there is multiple target.
result type is return type of d_meth.

Extending Benchmark Library

New Factory Function

using std::bind to make factory function.
pre_func and post_func can be omitted.

// T_val is type return by clock_func
auto factory(std::function<T(void)> func) {
  return std::bind(time<T_val>, clock_func, clock_diff_func, func, pre_func,
                   post_func);
}

New Result Method

the calculate function should take std::span<std::chrono::nanoseconds>

for example, you want top 3 fastest results

using nanos = std::chrono::nanoseconds;
std::array<nanos> top3(const std::span<nanos> span) {
  if(span.size() < 3) {
    throw std::runtime_error("not enough results");
  }
  std::nth_element(span.begin(), span.begin() + 3, span.end());
  std::array<nanos, 3> top3{span[0], span[1], span[2]};
  return top3;
}