Before the main content: I am creating an R Community on Google Groups. You can join the group using this form.
Have you ever heard someone claim that R is “too slow”?
As a Linux user who does almost everything on the command line, I used to find this idea hard to believe. R has always been quite efficient for my work. But recently, everything clicked.
While attending an R Sprint, I had to run R inside a Windows virtual machine to test that my suggested improvements to R worked there too. Linear models and some mathematical functions ran much slower compared to another Linux virtual machine setup.
Here is why: When you install Linux, something naive such as “pacman -S r” manages high-performance numerical libraries (like OpenBLAS) and makes sure everything is neat. On Windows, base R doesn’t ship with these optimizations out of the box, and setting them up requires an intricate separate installation.
A while back, Microsoft R Open solved this exact bottleneck by shipping with Intel MKL built-in. MRO is now retired and it closed a gap.
To bridge this gap again, I created R Batteries Included. It is a custom R distribution for Windows that ships pre-configured with OpenBLAS and Intel MKL. No messy setups and no extra steps. Just high-performance R on Windows, right out of the box.
G. N. Gueye suggested that my R distribution that makes R fast on Windows should be openly available with sponsorships. I followed that advice and R Batteries Included is here:
See the plots showing the speed gains. I used the famous AT&T Benchmark.
Please re-post, try it, and add a star to the repository. If you can, consider sponsoring this work on: