For linear algebra operations, we recommend that you use the vendor versions of LAPACK and ScaLAPACK routines that are tuned by the vendors for their platforms, such as LIBSCI on the Cray T3E, ESSL and PESSL on the IBM SP, and CHALLENGEcomplib or SCSL on the SGI/Cray Origin 2000. You should use the netlib version of a LAPACK or ScaLAPACK routine only if that routine is not available in the vendor version. For a comparison of netlib and vendor versions of LAPACK and ScaLAPACK, see the paper ScaLAPACK Evaluation and Performance at the DoD MSRCs (postscript), also available as CEWES MSRC Technical Report 98-33 (in Adobe PDF).

Please send comments or questions to
Clay Breshears or Shirley Browne.

Since March 9, 1999, you are visitor :

Search the Catalog

• Benchmark programs
• Distributed Processing Tools
• Math Libraries
• Parallel Processing Tools
  • Communication Libraries
  • Debuggers
  • Parallel I/O Systems
  • Parallel Programming Languages and Compilers
  • Performance Analysis Tools

Benchmark programs

LLCbench Low-Level Characterization Benchmarks

Distributed Processing Tools

Java Development Kit for compiling and running applets and applications written in Java

Math Libraries

Cray Scientific Library library of mathematical subroutines optimized for UNICOS and UNICOS/mk platforms

IBM ESSL 2.2 Engineering and Scientific Subroutine Library

IBM ESSL 3.1 Engineering and Scientific Subroutine Library

IBM PESSL 1.2 Parallel Engineering and Scientific Subroutine Library

IBM PESSL 2.1 Parallel Engineering and Scientific Subroutine Library

LAPACK transportable library of Fortran 77 routines for solving numerical linear algebra problems

ScaLAPACK dense linear algebra routines for distributed memory computers

SCSL SGI Cray Scientific Library

SGI CHALLENGEcomplib optimmized library for linear algebra operations, FFTs, and convolutions

SGI libfastm highly optimized versions of a subset of the libm math library

SuperLU set of subroutines for direct solution of sparse linear systems

SuperLU_MT parallel extension to the serial Super LU library

Parallel Processing Tools

Communication Libraries

Debuggers

Cray TotalView source-level, multiprocess debugger for Cray Unicos and Unicos/mk platforms

Etnus TotalView The TotalView debugger is a source-level, window-oriented, multiprocess debugger

pdbx IBM's extension for parallel programs of the dbx command-line debugger

SGI dbx a source-level command-line debugger with some multiprocess and thread support

SGI Workshop Debugger graphical source-level debugging tool for SGI IRIX platforms

XMPI graphical interface for starting and monitoring MPI applications

Parallel I/O Systems

GPFS General Parallel File System for the IBM SP

IBM PIOFS Parallel I/O File System for AIX

MPI-IO/PIOFS implementation of MPI-IO for IBM SP

ROMIO a high-performance, portable implementation of MPI-IO

Parallel Programming Languages and Compilers

KAP/Pro for OpenMP combines a full implementation of OpenMP with unique parallel correctness and performance tools

OpenMP supports portable shared memory parallel programming in C/C++ and Fortran

SGI Parallel Analyzer (cvpav) graphically displays information about Fortran programs

Performance Analysis Tools

Autopilot infrastructure for real-time adaptive control of parallel and distributed computing resources

Dynamic Probe Class Library (DPCL) client-server extension of the dyninst per-node dynamic instrumentation library

MPP Apprentice performance analysis tool for the Cray T3E

Paradyn Parallel Performance Tool real-time performance measurement of large-scale parallel programs

PAT performance analysis tool for the Cray T3E

perfex command line interface to the R10000 hardware performance counters

SpeedShop set of tools for running experiments and generating reports to analyize performance on SGI IRIX platforms

VAMPIR Visualization and Analysis of MPI Resources

Vampirtrace MPI profiling and trace data generation library

Virtue a collaborative virtual environment for direct software manipulation

browne@cs.utk.edu