Supercharge R, with Multi-Threaded Maths
Open up R, either in your RStudio, or your standard RGui, and run the following (matrix multiplication ensues);
system.time({ x <- replicate(5e3, rnorm(5e3)); tcrossprod(x) })
On my system, using either R or RStudio (using the same base R engine), the user time was about 50 seconds. That, due to the single threaded processing nature of R.
Time to supercharge R maths
Microsoft has come out with a replacement default RGui, MRO (Microsoft R Open), which combined with the MKL (math kernel library), brings about multi-threaded processing.Go here and download MRO and the MKL
https://mran.revolutionanalytics.com/download/#download
Your new processing speed will look something more like this now;
If you're using RStudio, be sure you modify the R version in the options to point to your new-fangled R installation.
No comments:
Post a Comment