Skip to content

martinus/nanobench

master
Switch branches/tags

Name already in use

A tag already exists with the provided branch name. Many Git commands accept both tag and branch names, so creating this branch may cause unexpected behavior. Are you sure you want to create this branch?
Code

Latest commit

* Check for failures in parseFile()

Check for failures when opening/reading the CPU frequency related sysfs
files, to avoid needless warning messages.

Note that if `scaling_min_freq` and `scaling_max_freq` both fail, then
`minFreq` and `maxFreq` will be zeroes and no warning is generated.

* Fallback to PERF_COUNT_HW_CPU_CYCLES

Not all platforms support the reference cycles, so fallback to the cycle
counter, it's better than nothing.

* Add HW perf events before SW events

The mixed order of HW and SW events causes EINVAL failures from some of
the perf_event_open() syscalls in one embedded device with 4.14 kernel.
Fixed by adding all the HW events first, then the SW events.
a0e1b34

Git stats

Files

Permalink
Failed to load latest commit information.
Type
Name
Latest commit message
Commit time
 
 
 
 
 
 
 
 
src
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

ankerl::nanobench

ankerl::nanobench logo

Release GitHub license Travis CI Build Status Appveyor Build Status Join the chat at https://gitter.im/nanobench/community

ankerl::nanobench is a platform independent microbenchmarking library for C++11/14/17/20.

#define ANKERL_NANOBENCH_IMPLEMENT
#include <nanobench.h>

int main() {
    double d = 1.0;
    ankerl::nanobench::Bench().run("some double ops", [&] {
        d += 1.0 / d;
        if (d > 5.0) {
            d -= 5.0;
        }
        ankerl::nanobench::doNotOptimizeAway(d);
    });
}

The whole executable runs for ~60ms and prints

|               ns/op |                op/s |    err% |          ins/op |          cyc/op |    IPC |         bra/op |   miss% |     total | benchmark
|--------------------:|--------------------:|--------:|----------------:|----------------:|-------:|---------------:|--------:|----------:|:----------
|                7.52 |      132,948,239.79 |    1.1% |            6.65 |           24.07 |  0.276 |           1.00 |    8.9% |      0.00 | `some double ops`

Which github renders as

ns/op op/s err% ins/op cyc/op IPC bra/op miss% total benchmark
7.52 132,948,239.79 1.1% 6.65 24.07 0.276 1.00 8.9% 0.00 some double ops

The benchmarked code takes 7.52 nanoseconds to run, so ~133 million times per seconds. Measurements fluctuate by 1.1%. On average 6.65 instructions are executed in 24.07 CPU cycles, resulting in 0.276 instructions per cycle. A single branch is in the code, which branch prediction missed in 8.9% of the cases. Total runtime of the benchmark with the name some double ops is 0.00, so just a few milliseconds.

Design Goals

Documentation

Extensive documentation is available.

More

  • Code of Conduct - Contributor Covenant Code of Conduct
  • I need a better logo. Currently I use a small bench. Nanobench. Ha ha.