SGIs Scientific Computing Software Library
General
The Scientific Computing Software Library (SCSL) from SGI offers
highly optimized math routines for science and engineering on SGI
shared memory machines; at NSC these routines are available on Mozart.
See the product info from SGI for more details. [here].
The library contains:
- BLAS
- LAPACK
- Signal processing library with FFTs.
- Sparse direct solvers.
- Sparse iterative solvers.
- ScaLAPACK
- BLACS
Essentials
The SCSL is divided into two parts: scs and sdsm. The
former contains BLAS, LAPACK, FFT, and solvers and the latter contains
ScaLAPACK and BLACS. Both parts are available in integer sizes of 32
bits
libscs.so
ibsdsm.so
and 64 bits
libscs_i8.so
libsdsm_i8.so
In addition, the library is available in a parallel version, in which
the corresponding files are named:
libscs_mp.so libscs_i8_mp.so
libsdsm_mp.so libsdsm_i8_mp.so
The math routines are linked into your application. If you link an
application for a single processor, the link command may look like:
ifort -o myprogram.x myprogram.o -L/usr/lib -lscs -lsdsm
If you link an application for multi-processor usage, the link command
may look like:
ifort -o myprogram.x -L/usr/lib myprogram.o -lscs_mp -lsdsm_mp
Resource specifics
On Mozart you can also access SCSL by use of modules. You add the
appropriate module with the command:
module add scsl
and your appropriate path to the library will be added to the linker
via the variabel LD_LIBRARY_PATH. In this case, you need not to
specify the path in the above example.
Note:
- The variable OMP_NUM_THREADS is default set to 1 on Mozart. You
must set this variable to the number of processors that you intend to
use in the SCSL paralllel library routines.
Page last modified: 2006-05-23 14:10
For more information contact us at
info@nsc.liu.se.
| 
