simple_fastfloat_benchmark

This repository contains benchmarking code related to the fast_float number parsing library. It supports at least Linux, macOS and Windows (Visual Studio).

If you have a recent version of CMake (3.15 or better) under linux, you can simply go in the directory and type the following commands:

cmake -B build .
cmake --build build
./build/benchmarks/benchmark 

While the fast_float library only requires C++11, for this benchmark, we are assuming that your compiler supports C++17.

You may use docker to run these benchmarks easily on a variety of platforms: see https://github.com/lemire/docker_programming_station

Windows

Usage under Windows is similar. After installing cmake and Visual Studio 2019, one might type in the appropriate shell:

cmake -B build .
cmake --build build --config Release
.\build\benchmarks\Release\benchmark.exe

Advanced Usage

Parse the strings (one per line) included in a text file:

./build/benchmarks/benchmark -f data/canada.txt

Parse strings generated from floats in (0,1):

./build/benchmarks/benchmark

Parse strings generated from floats in (0,1), using a few digits as possible during the serialization:

./build/benchmarks/benchmark -c

Parse strings generated from floats in the int_e_intmodel (credit @alexey-milovidov), using a few digits as possible during the serialization:

./build/benchmarks/benchmark -m int_e_int

There a different models available, they are printed out in the console.

Example output

GNU GCC 9 (Linux):

$ ./build/benchmarks/benchmark 
# parsing random numbers in the range [0,1)
volume = 2.09808 MB 
netlib                                  :   271.18 MB/s (+/- 1.2 %)    12.93 Mfloat/s  
doubleconversion                        :   225.35 MB/s (+/- 1.2 %)    10.74 Mfloat/s  
strtod                                  :   190.94 MB/s (+/- 1.6 %)     9.10 Mfloat/s  
abseil                                  :   430.45 MB/s (+/- 2.2 %)    20.52 Mfloat/s  
fastfloat                               :  1042.38 MB/s (+/- 9.9 %)    49.68 Mfloat/s  

$ ./build/benchmarks/benchmark -f data/canada.txt 
# read 111126 lines 
volume = 1.93374 MB 
netlib                                  :   287.92 MB/s (+/- 1.5 %)    16.55 Mfloat/s  
doubleconversion                        :   203.95 MB/s (+/- 1.4 %)    11.72 Mfloat/s  
strtod                                  :   139.51 MB/s (+/- 2.0 %)     8.02 Mfloat/s  
abseil                                  :   399.62 MB/s (+/- 1.7 %)    22.96 Mfloat/s  
fastfloat                               :   912.10 MB/s (+/- 2.3 %)    52.42 Mfloat/s  

$ ./build/benchmarks/benchmark -f data/mesh.txt 
# read 73019 lines 
volume = 0.536009 MB 
netlib                                  :   309.97 MB/s (+/- 2.8 %)    42.23 Mfloat/s  
doubleconversion                        :   221.93 MB/s (+/- 1.5 %)    30.23 Mfloat/s  
strtod                                  :   117.69 MB/s (+/- 1.8 %)    16.03 Mfloat/s  
abseil                                  :   243.55 MB/s (+/- 2.8 %)    33.18 Mfloat/s  
fastfloat                               :   657.04 MB/s (+/- 4.6 %)    89.51 Mfloat/s