(This article was first published on

**PlanetFlux**, and kindly contributed to R-bloggers)I did some comparisons of the generic BLAS with Intel's MKL (both sequential and parallel) on a Dell PowerEdge 610 server with dual hyperthreading 6-core 3.06GHz Xeon X5675 processors. Here are the results from an R benchmarking script (Normal R indicates the generic BLAS, sMKL is the sequential (single core Intel MKL, and pMKL is the parallel Intel MKL using all 24 threads available on this system). Times are in seconds, lower is better.

R Benchmark 2.5

===============

Number of times each test is run__________________________: 3

I. Matrix calculation

--------------------- Normal R sMKL pMKL

Creation, transp., deformation of a 2500x2500 matrix (sec): 0.592 0.583 0.585

2400x2400 normal distributed random matrix ^1000____ (sec): 0.425 0.411 0.427

Sorting of 7,000,000 random values__________________ (sec): 0.787 0.778 0.777

2800x2800 cross-product matrix (b = a' * a)_________ (sec):

**11.543 1.875****0.283** Linear regr. over a 3000x3000 matrix (c = a \ b')___ (sec):

**5.367 0.910****0.214** Trimmed geom. mean (2 extremes eliminated): 1.358 0.743 0.414

II. Matrix functions

--------------------

FFT over 2,400,000 random values____________________ (sec): 0.422 0.451 0.435

Eigenvalues of a 640x640 random matrix______________ (sec): 0.949 0.443 0.414

Determinant of a 2500x2500 random matrix____________ (sec):

**4.864****0.967****0.352** Cholesky decomposition of a 3000x3000 matrix________ (sec):

**4.131****0.865****0.179** Inverse of a 1600x1600 random matrix________________ (sec):

**4.011****0.751****0.277** Trimmed geom. mean (2 extremes eliminated): 2.505 0.667 0.343

III. Programmation

------------------

3,500,000 Fibonacci numbers calculation (vector calc)(sec): 0.787 0.824 0.841

Creation of a 3000x3000 Hilbert matrix (matrix calc) (sec): 0.456 0.465 0.431

Grand common divisors of 400,000 pairs (recursion)__ (sec): 2.196 2.386 1.927

Creation of a 500x500 Toeplitz matrix (loops)_______ (sec): 0.616 0.612 0.596

Escoufier's method on a 45x45 matrix (mixed)________ (sec): 0.470 0.425 0.447

Trimmed geom. mean (2 extremes eliminated): 0.611 0.617 0.607

--------------------------------------------

Total time for all 15 tests_________________________ (sec):

**37.62****12.76****8.18**Overall mean (sum of I, II and III trimmed means/3)_ (sec):

**1.28****0.67****0.44**

So you can see there are some significant gains, especially for the slowest tasks (see bolded items above). For example, the parallel MKL resulted in a 40x speedup of the cross-product calculation. Across all jobs

*the parallel MKL version was ~4.5x faster than the generic BLAS on this system*. When running models that take days to fit, that is significant!Here's how I installed it on Ubuntu 12.04:

- Download and install the Intel MKL from here.
First install some libraries needed by R:sudo apt-get install libreadline6 libreadline6-dev xserver-xorg xserver-xorg-dev gfortran-devel

libxt-dev- Download the latest R, I used v2.15, available here
- Compile R from source using something like this (detailed instructions here):

export MKL_LIB_PATH=/opt/intel/composer_xe_2011_sp1.7.256/mkl/lib/intel64export optim_flags="-O3 -funroll-loops -march=native"# from http://cran.r-project.org/doc/manuals/R-admin.html#MKL=" -L${MKL_LIB_PATH} -Wl,--start-group ${MKL_LIB_PATH}/libmkl_gf_lp64.a \ ${MKL_LIB_PATH}/libmkl_gnu_thread.a ${MKL_LIB_PATH}/libmkl_core.a \ -Wl,--end-group -lgomp -lpthread"sudo ./configure --enable-R-shlib --with-blas="$MKL" CC="gcc" CFLAGS="$optim_flags" CXX="g++" CXXFLAGS="$optim_flags" F77="gfortran" FFLAGS="$optim_flags" FC="gfortran" FCFLAGS="$optim_flags"

To

**leave a comment**for the author, please follow the link and comment on his blog:**PlanetFlux**.R-bloggers.com offers

**daily e-mail updates**about R news and tutorials on topics such as: visualization (ggplot2, Boxplots, maps, animation), programming (RStudio, Sweave, LaTeX, SQL, Eclipse, git, hadoop, Web Scraping) statistics (regression, PCA, time series, trading) and more...