gauge is a flexible C++ benchmarking tool.
C++ Python
Latest commit d20a6fe Apr 7, 2017 @petya2164 petya2164 Print results
Failed to load latest commit information.



gauge is a flexible C++ benchmarking tool.


We have tested gauge on various versions of Windows, Linux and Mac. You can see the status by selecting the gauge project on the Steinwurf buildbot page:


We use the waf build-system to build the gauge static library. With some additional tools which may be found at waf

To build gauge run the following command:

./waf configure --bundle-path=~/dev/bundle_dependencies
./waf build

Substitute the ~/dev/bundle_dependencies with the path where you wish waf to download gauge's dependencies. You may omit this option then waf will download the dependencies into the gauge project folder.

You should now have the libgauge.a static lib. It also build the boost dependencies as static libs. In the gauge directory type find . -name "*.a" to see the libraries produced. We need to use the path to the libraries when using gauge with our own applications as shown in the following.

When building the static lib, waf will also build the gauge_example executable.

Depending on your platform you should be able to launch it by running:


Example Use

See various use-cases in the examples folder. The following will be used to explain the basic concepts of gauge. To try it out save the following code in a file called main.cpp

#include <gauge/gauge.hpp>
#include <gauge/console_printer.hpp>
#include <vector>

BENCHMARK(MyTest, RunThis, 100)
    std::vector<int> integers;
    for(int i = 0; i < 2048; ++i)

    // This is where the clock runs
        for(uint i = 1; i < integers.size(); ++i)
            integers[i] += integers[i-1];

int main(int argc, const char* argv[])

    gauge::runner::run_benchmarks(argc, argv);
    return 0;

In the above we use the BENCHMARK macro which takes 3 parameters:

  1. The name of the test-case in this case MyTest
  2. The name of the benchmark in this case RunThis
  3. The number of runs to complete in this case 100.

The measurement will not start until we hit the RUN macro. Depending on the type of benchmark (the default is time) the code inside RUN will be executed several times (we refer to this as the number of iterations). When gauge is satisfied with the measurement we exit the run loop. For every BENCHMARK we may only call RUN once.

Using g++ the example code may be compiled as:

g++ main.cpp -o benchmark --std=c++0x -I../path_to_gauge/ -L../path_to_libguage -lgauge -L../path_to_libboostxyz -lboost_chrono -lboost_program_options -lboost_system -lrt

We assume you are using the boost libs built together with gauge. In the future we will attempt also support boost libraries installed using your package manager or similar.

You should now be able to run the benchmark using:



gauge is available under the BSD license, see the LICENSE.rst file.


We have create gauge to fit our specific purpose, however we hope that others may also find it useful. When designing gauge we found inspiration in these other nice projects:

  • Qt's benchmark tools part of QTestLib.
  • Nick Bruun's Hayai C++ benchmark tool, who also provided the nice terminal colors.
  • The Google Test framework.

Thanks for all the fish.