HP MLIB for PA-RISC and Itanium HP-UX Version 9.5.6 Release Note

What’s in This Version

MPI Requirements

The MPI parallel libraries ScaLAPACK and SuperLU_DIST in HP MLIB were built and

tested with HP-MPI Version 2.1.1. Refer to “MPI Parallelism” on page 14.

Benefits and Features

HP MLIB V9.5.6 is optimized for use on both PA-RISC 2.0 and Itanium-based platforms

running HP-UX.

HP MLIB V9.5.6 incorporates algorithmic improvements and several tunable parameter

adjustments for improved execution performance. VECLIB is optimized by using

highly-efficient implementations of both dense and sparse BLAS—levels 1, 2, and 3. LAPACK

fully conforms with the public-domain version 3.0 of LAPACK. The key computational kernels

in LAPACK are optimized to take full advantage of the respective architecture. The HP

version of ScaLAPACK is tuned on and built with HP-MPI V2.1.1 and fully conforms to V1.7

of the ScaLAPACK standard. (Refer to http://www.hp.com/go/mpi for information on the HP

MPI product.) This implementation provides the SuperLU_DIST V2.0 library designed for

solving non-symmetric sparse linear systems on distributed memory systems. VMATH

supports vector math routines.

HP MLIB V9.5.6 incorporates algorithmic improvements, and several tunable parameters

have been adjusted for improved execution performance. Features in this version include

• VECLIB

• LAPACK

• ScaLAPACK

• Distributed SuperLU

• SOLVERS

•VMATH

• SMP Parallelism

•MPI Parallelism

VECLIB

VECLIB is a collection of Fortran-callable subprograms optimized for use on Hewlett-Packard

servers, providing mathematical software and computational kernels for engineering and

scientific application programs. This library contains subprograms for

• Dense vector operations, including the Level 1 BLAS

• Matrix operations, including the Level 2 and Level 3 BLAS