Excel Benchmarks

Gathering benchmarks for Excel add-ins. Inspired by The Benchmarks Game and some LinkedIn posts.

Featuring

• Python via xlOil - thanks for the help @cunnane!
• Python via xlwings
• C# via Excel-DNA - thanks for the help @govert!
• C++ via xlladdins - credits to @keithalewis!
• VBA
• Native Excel Formulas

Unfortunately I couldn't make a JavaScript addin to work. Suggestions welcome!

Benchmarks

Sum a 1.000 x 1.000 matrix (1 million cells)

Time in miliseconds:

Name	Excel SUM(...)	C++ (xlladdins)	C# (Excel-DNA)	Python + Numpy (xlOil)	Python + Numba (xlOil)	Python Only (xlOil)	VBA	Python + Numpy (xlwings)
Median	3.2	10.1	13.1	36.7	38.6	117.9	239.5	893.5

How it works

I used the VBA code provided by Microsoft for performance measurement, in order to calculate the runtime of each cell with formulas. I changed it to run a 100 times in order to compute the statistics.

How to build (Windows only)

• Download Visual Studio Community 2022 (free for personal use)
• Install with .NET and C++ support
• Install Python 3.10 (pyenv is a nice way to control versions)
• Run the following commands in a terminal:

pip install xloil xlwings numba icc-rt cffi --upgrade
xlwings addin install
xloil install

• Clone the repo. On the instructions below, ROOT will be the repo folder path.
• Python - xlOil (has to be done before opening the workbook):
  • Open %APPDATA%\xlOil\xlOil.ini.
  • On the parameter PYTHONPATH, add ;ROOT\Python to the end
  • Change parameter: LoadModules=["xloil_addin"]
• Open ROOT\Benchmarks.xlsm
• C#:
  • Build CSharpAddin project with Release configuration.
  • Drag this file to Excel: ROOT\CSharp\bin\Release\net6.0-windows\CSharpAddin64.xll
• C++:
  • Build CppAddin project with Release configuration.
  • Drag this file to Excel: ROOT\Cpp\x64\Release\CppAddin.xll
• Python - xlwings:
  • On the xlwings tab inside Excel, add ROOT\Python to PYTHONPATH
  • On UDF Modules parameter, write xlwings_addin
  • Click Import Functions
• Make sure the formulas for each framework work (yellow cells).
• Click the button Run to compute the times