fixest: A fast and feature-rich framework for econometric estimations in R -- Benchmarks

This page contains the necessary source code to replicate the two sets of benchmarks of the paper "fixest: A fast and feature-rich framework for econometric estimations in R" by Laurent Bergé, Kyle Butts and Grant McDermott (https://arxiv.org/abs/2601.21749).

It contains two archives:
- benchmark-sim: to replicate the results of Section 8.1 Simulated data
- benchmark-taxi: to replicate the results of Section 8.2 NYC taxi data

To replicate the benchmarks you need to have R, Julia and python/uv installed, then:
- unzip the archive
- read README.md
- open a terminal and set the working directory to the folder containing the archive's files
- run `make` (if you have `make` installed of course, otherwise, see README.md)

Note that the benchmarks run smoothly on computers with 32+GB of RAM, but specific settings may fail on 16-GB of RAM computers. To run smaller (but informative) benchmarks, use the environment variable `SMALL=1`, as in `make SMALL=1`.

Second note: the simulated benchmark may take several hours to run.