This a sort benchmark in support of this related blog post.
Just run make.
Currently it builds on Linux with clang or gcc, or on Windows Subsystem for Linux (WSL) with the same compilers.
In principle it could work fine on Windows with other compilers, but I haven't had time to sort out my Visual Studio Community license.
It should work on OSX with clang or gcc, although I haven't tried it.
Patches which add platform support are welcome.
Run it like ./bench to run all the benchmarks, or like ./bench BENCH_NAME to run a specific benchmark (e.g., ./bench qsort to run the qsort benchmark). The benchmark names are the same as those shown in the first row when you run ./bench.
You too can generate the figures from the blog post on your hardware, using the scripts in the scripts directory.
First generate the data by running scripts/data.sh. This updates the CSV file data/all.csv.
Then you can generate all the figures by running scripts/allfigs.sh. This generates the .svg figures in figures. You can also generate any figure directly by calling say scripts/fig2.sh to generate Figure 2 (which by default opens an interactive window to display the figure).
Some code in the bench.cpp and wyhash.h originates from this code written by Daniel Lemire for a blog post.
wyhash courtesy of Wang Yi.