- Infos im HLRS Wiki sind nicht rechtsverbindlich und ohne Gewähr -
- Information contained in the HLRS Wiki is not legally binding and HLRS is not responsible for any damages that might result from its use -

MKL: Difference between revisions

From HLRS Platforms
Jump to navigationJump to search
mNo edit summary
No edit summary
Line 1: Line 1:
Intel's Math Kernel Library (MKL) is a library of optimized, math routines for science, engineering, and financial applications. Core math functions include BLAS, LAPACK, ScaLAPACK, Sparse Solvers, Fast Fourier Transforms and Vector Math.
== Fortran 95 Interfaces and Wrappers for LAPACK, BLAS, etc. ==
== Fortran 95 Interfaces and Wrappers for LAPACK, BLAS, etc. ==
The MKL fortran interfaces must be build seperately because different compiler return complex values differently. Due to the large number of combinations occuring from this the user has to do this on his own.
The MKL fortran interfaces must be build seperately because different compiler return complex values differently. Due to the large number of combinations occuring from this the user has to do this on his own.

Revision as of 12:40, 24 February 2010

Intel's Math Kernel Library (MKL) is a library of optimized, math routines for science, engineering, and financial applications. Core math functions include BLAS, LAPACK, ScaLAPACK, Sparse Solvers, Fast Fourier Transforms and Vector Math.

Fortran 95 Interfaces and Wrappers for LAPACK, BLAS, etc.

The MKL fortran interfaces must be build seperately because different compiler return complex values differently. Due to the large number of combinations occuring from this the user has to do this on his own.

# example how to build a fortran interface of the intel MKL
# set up your environment
module load numlib/intel # set up MKL environment
module load compiler/intel
module load mpi/impi # MPI needed by some fftw interfaces
# switch into the directory of the needed interface
cd $MKLROOT/interfaces/<interfacename>
# display build options like compiler and mpi library
make
# install the library into $HOME/lib
# By default, for interfaces without MPI the ifort compiler is used. 
# You can change it by an additional parameter FC=<compiler>.
make INSTALL_DIR=$HOME/lib libem64t [OPTIONS]


Example

Example program example_blas.f90:

program example_blas
  implicit none
  real, dimension(3) :: a = (/2.,1.,-1./), b = (/5., -2., 1.5/)
  real               :: c, sdot
  real(kind=8)       :: d, dnrm2
  ! *** skalar product ***
  ! sdot: s ... REAL, dot ... skalar product
  ! 1. Argument  ... dimensions of the vectors
  ! 2. und 4. A. ... the vectors
  ! 3. und 5. A. ... increment (here 1)
  c = sdot(3, a, 1, b, 1)
  write(*,*) c
  ! Result: 6.500000
  ! *** norm of the vectors ***
  ! dnrm2: d ... DOUBLE PRECISION, nrm2 ... (euklidian) norm
  ! 1. Argument: dimensions of the vectors
  ! 2. A.:       vektor
  ! 3. A.:       increment (here 1)
  d = dnrm2(3, dble(a), 1)
  write(*,*) d
  ! Result: 2.44948974278318
end program example_blas

Compile with

ifort example_blas.f90 -L$HOME/lib -lmkl_blas95 $MKLPATH/libmkl_intel_lp64.a -Wl,
      --start-group $MKLPATH/libmkl_intel_thread.a $MKLPATH/libmkl_core.a -Wl,--end-group 
      -L$MKLPATH -liomp5 -o example_blas

Links

http://software.intel.com/en-us/intel-mkl/