NAG - Numerical Algorithms Group library

The NAG Fortran library is an extensive and robust library of numerical algorithms that has been tested over many years. It is reliable and comprehensive.

NAG's own on-line documentation :

NAG Ltd. home page

Fortran numerical library

index to library

There are also 8 volumes of manuals in book form (for Version 14) (somewhere). We seem to have Version 17 at the time of writing, but do ls -l /opt/lib/libnag.a to get an up-to-date view. Source code not available. Not currently on Linux.

Examples (in Fortran) of how to use each routine can also be found in somewhere like here.

Calling NAG Fortran routines from C/C++:

Here is a simple example in C++ for the NAG complex division function A02ACF. The source code is:

#include 

extern "C" {
  void a02acf_(double &XR, double &XI, double &YR, double &YI, double &ZR, double &ZI);
}

int main() {
  double XR=1.0, XI=1.0, YR=0.0, YI=1.0, ZR, ZI;
  a02acf_(XR,XI,YR,YI,ZR,ZI);
  cout << "z = (" << ZR << "," << ZI << ")\n";
}

Notes:

the NAG function must be renamed: A02ACF has become a02acf_
a suitable function declaration is needed for each function (wrapped in extern "C" { } if using C++)
all arguments are passed in by reference (C programmers use pointers)
2D array indices in Fortran are the opposite way round from C/C++
strings: my impression is that they are stored like they are in C (i.e. with trailing null). Any more information would be welcome.
more complex arguments - try Jonathan Deane

Linking under Solaris:

using QMake, add nag to the USESLIBs variable
using ld, add -lnag to the library path

NAG with SGI:

On SGI there are 2 libraries: nag3000 and nag4000, corresponding to two different cpu models (neither compatible with AMMA). I'm not sure we have it installed correctly.

Acknowledgements:

Thanks to Jonathan Deane, who supplied most of the information on this page.

Bill Christmas

Last modified: Mon Oct 27 14:27:25 GMT 2003