The library consists actually of six libraries at /usr/local/lib/
libnagmpi.a
libnagmpiz.a
libnagblas.a
libnagmpiblacs.a
libnagtools.a
libnagfls.a
and can be used quite simply following the example script
/usr/local/lib/fdcre02sm/scripts/nscexample
The documentation is available at NSC as printed manuals.
You can also read some of it on the network directly from Oxford.
The example programs are easy to use through the nscexample script. On your own directory you give the command
/usr/local/lib/fdcre02sm/scripts/nscexample NAG_namewhere NAG_name is the name of the routine you wish to test, for example d01atfp for adaptive quadrature on a finite interval, allowing for badly behaved integrands. The example program, together with its data file (if any), will be copied to your present directory, compiled, linked, and run. The result file will be put in the subdirectory newresults.
I recommend that you introduce an alias for the long command string required to use nscexample.
The example program can quite often be used as the basis for the application program you are about to write. It includes the initial call of Z01AAFP to initialise a 2-D-grid for MPI and the final call of Z01ABFP to un-define the grid.
There are no official C header files available for using the parallel library from C or C++.
NAG also has a serial Fortran 77 Library (installed on our T3E), and a new numerical library written in full Fortran 90 (not installed on our T3E).
|
![]()
|