Fork me on GitHub

Intel Math Kernel Library

Intel’s Math Kernel Library (MKL) provides highly optimized, threaded and vectorized functions to maximize performance on each processor family. It Utilises de-facto standard C and Fortran APIs for compatibility with BLAS, LAPACK and FFTW functions from other math libraries.

Interactive usage

The MKL was installed along with the Intel compilers as part of Intel Parallel Studio. The MKL can be used with or without other Parallel Studio packages (such as the Intel compilers).

To activate just the MKL, use one of :

module load libs/binlibs/intel-mkl/2017.0

module load libs/binlibs/intel-mkl/11.2.3

Note that 2017.0 is newer than 11.2.3.

Sample C and Fortran programs demonstrating matrix multiplication are available for version 2017.0 in the directory $MKL_SAMPLES:

$ ls $MKL_SAMPLES/
mkl_c_samples  mkl_fortran_samples

To compile one of these you need to copy the samples to a directory you can write to, ensure the MKL and the Intel compilers are both loaded, change into the directory containing the relevant sample program (C or Fortran) then run make to compile:

$ qrsh
$ cp -r $MKL_SAMPLES/ ~/mkl_samples
$ module load libs/binlibs/intel-mkl/2017.0
$ module load libs/binlibs/intel-mkl/2017.0
$ cd ~/mkl_samples/mkl_fortran_samples/matrix_multiplication
$ make

ifort -c src/dgemm_example.f -o release/dgemm_example.o
ifort release/dgemm_example.o -mkl -static-intel -o release/dgemm_example
ifort -c src/dgemm_with_timing.f -o release/dgemm_with_timing.o
ifort release/dgemm_with_timing.o -mkl -static-intel -o release/dgemm_with_timing
ifort -c src/matrix_multiplication.f -o release/matrix_multiplication.o
ifort release/matrix_multiplication.o -mkl -static-intel -o release/matrix_multiplication
ifort -c src/dgemm_threading_effect_example.f -o release/dgemm_threading_effect_example.o
ifort release/dgemm_threading_effect_example.o -mkl -static-intel -o release/dgemm_threading_effect_example

You should then find several compiled programs in the release directory:

$ ./release/matrix_multiplication
 This example measures performance of computing the real
 matrix product C=alpha*A*B+beta*C using
 a triple nested loop, where A, B, and C are matrices
 and alpha and beta are double precision scalars

 Initializing data for matrix multiplication C=A*B for
 matrix A( 2000 x  200) and matrix B(  200 x 1000)

 Intializing matrix data

 Making the first run of matrix product using
 triple nested loop to get stable run time
 measurements

 Measuring performance of matrix product using
 triple nested loop

 == Matrix multiplication using triple nested loop ==
 == completed at    184.42246 milliseconds ==

 Example completed.

For documentation and a tutorial, start an interactive session using qsh, load MKL, then run:

$ firefox $MKL_SAMPLES/mkl_fortran_samples/matrix_multiplication/tutorial/en/index.htm

Installation Notes

The following notes are primarily for system administrators.

Intel MKL 2017.0

Installed as part of Parallel Studio Composer Edition 2017.

This modulefile was installed as /usr/local/modulefiles/libs/binlibs/intel-mkl/2017.0.

Intel MKL 11.2.3

Installed as part of Intel Parallel Studio Composer Edition 2015 Update 3.

This modulefile was installed as /usr/local/modulefiles/libs/binlibs/intel-mkl/11.2.3