Linking MKL to R on Windows

💰 I’m still using this on R version 4.5 and it works fine.

🐦 The Bottom Line (at the top)

Revolutions R came along around 2007 and provided more powerful R computing… it was freeware, but you had to register to get it. A few years later, Microsoft bought it and called it Microsoft R Open. Unfortunately, it is gone with the wind.

Here is a work around from the discussion on Stackoverflow - Linking Intel’s Math Kernel Library (MKL) to R on Windows.

These notes are for Windows. If you run Linux, go to MKL4DEB for details (as mentioned there, you’ll want to scroll down to the Appendix first).

Before you start, do this and record how flippin’ long it takes to run:

# Singular Value Decomposition
m <- 10000
n <- 2000
A <- matrix (runif (m*n), m, n)
system.time (S <- svd (A, nu=0, nv=0))

Now close R and let the magic begin …

---

🐦 Steps to Enhance R version 4

This is an easy update as described in the discussion on Stack Overflow previously mentioned.

• Download the library files from here (it’s an archive ~ 250 MB with 2 directories, compiler and mkl): Intel_Libraries.zip and extract everything.

• Go to the directory where R is installed, something like C:\Program Files\R\R-4.xxx\bin\x64\, and change Rlapack.dll and Rblas.dll to Rlapack.dll.bak and Rblas.dll.bak, respectively, as backups.

• From the extracted folder in step 1, paste all the content from the folders complier and mkl into the directory where R is installed.

• Inside the destination folder, create 2 (additional) copies of mkl_rt.dll and rename the new files as Rlapack.dll and Rblas.dll and keep mkl_rt.dll.

• Start R and rerun the SVD code and be amazed 😃 at the difference in computation time.

---

---

​