HP-UX 11i September 2002 Release Notes

HP-UX 11i Version 1.0 Operating Environment Applications
HP-UX 11i Technical Computing Operating Environment (TCOE)
Chapter 6
127
updated for
September 2002
New features for version B.08.03 include the following:
• ScaLAPACK
• Distributed SuperLU
• Nested SMP parallelism
ScaLAPACK is a library of high-performance linear algebra routines. This
implementation provides a version of ScaLAPACK tuned on HP servers and built with
HP’s Message Passing Interface (MPI). The ScaLAPACK library routines are callable
from Fortran 90 and C routines.
This implementation also provides the Distributed SuperLU library designed for
distributed memory parallel computers. The library is implemented in ANSI C, using HP
MPI for communication.
updated for
September 2001
New features for version B.08.01 include the following:
• 64-bit integer interface for VECLIB and LAPACK libraries
• Sparse BLAS functionality
• Sparse solver library improvements
• LAPACK 3.0 compliance
• Archive and shared libraries
new at 11i
original release
New features for version B.07.01 (described below) include the following:
• Shared libraries
• Basic Linear Algebra Subroutine (BLAS) Standard functionality
• LAPACK 3.0 tuned for HP PA-RISC 2.0 processors
• Simplified sparse solver interface with improved performance
• Improved performance of key routines
• Improved C and C++ usability
Impact
HP MLIB incorporates algorithmic improvements, and several tunable parameters are
adjusted for improved execution performance.
While you can use HP MLIB as archive or shared libraries, the performance of your
applications is better when you use archive libraries. However, if you need to keep
executable files small, you can use shared libraries on any PA-RISC 2.0 system running
the HP-UX 11.0 or later operating system.
VECLIB is optimized by using a highly efficient implementation of the Basic Linear
Algebra Subprograms (BLAS), Levels 1, 2, and 3, as well as a subset of the newly-defined
BLAS Standard. Performance for key BLAS routines has been improved.
MLIB fully conforms to the public domain version 3.0 of LAPACK in all user-visible
usage conventions. The internal workings of some subprograms have been tuned and
optimized for Hewlett-Packard computers.
This version simplifies sparse solver interface use and improves its performance. Version
4.0 of the METIS reordering technology has been incorporated.
You can now use the C or C++ compiler to link applications built with MLIB. (Previous to
this release, you were required to link using the Fortran compiler when using VECLIB
or LAPACK.)