Mark 17, the latest version of the Numerical Algorithms Group (NAG) subroutine library, is now available on all RCS machines. (The new licensing agreement with NAG makes the NAG library available on any Rensselaer-owned computer.)
To access the NAG library, enter the following commands from a UNIX prompt, using those appropriate to your machine type:
IBM systems: setup nag-new xlf mycode.f -L$SETUP_TempDir/lib -lnag -lblas SGI systems: setup nag-new f77 mycode.f -n32 -L$SETUP_TempDir/lib -lnag Sun systems: setup nag-new f77 mycode.f -L$SETUP_TempDir/lib -lnag
One new chapter, F11 - Sparse Linear Algebra, has been introduced with the Mark 17 version. Containing ten routines, the F11 chapter currently addresses the solution of real sparse symmetric linear systems using iterative techniques. (The chapter will be extended to treat nonsymmetric systems at the next Mark.) The F11 chapter also includes conjugate gradient and Lanczos algorithms, incomplete Cholseky, SSOR, and Jacobi preconditioning, and support routines.
Extended coverage in the differential equations chapters now includes the following topics. (The stated chapters appear in parentheses.)
NAG Mark 17 also uses eight new Black Box routines, which exploit existing F08 routines, for computing. (The stated chapters appear in parentheses.)
Users can find thirteen of the new routines in the Statistics chapters. These routines include facilities for the following, with the stated chapters appearing in parentheses:
NAG 17 will become the system default on Tuesday, January 6, 1998, and you will no longer have to set it up. (In addition, the NAG 17 library will be in /usr/local/lib and will not need to be "setup" when NAG 17 becomes the default.) You will still be able to access the current version of the program, NAG 16, until March 1, 1998, by entering the command setup nag-old from a UNIX prompt before running the program.
If you have any further questions about using the NAG libraries, please contact the consultants at the ACS Help Desk locations, or direct electronic mail to consult@rpi.edu.