Microsoft HPC Institute - University of Tennessee

University of Tennessee

Innovative Computing Laboratory

Computer Science Department

Jack Dongarra

Windows Cluster

Project

·    Detailed description of project exists

People

Jack Dongarra

George Bosilca

Dave Cronk

Graham Fagg

Julien Langou

Piotr Luszczek

Projects:

¨      Numerical Linear Algebra Algorithms and Software

Ø      LAPACK, ScaLAPACK, ATLAS

Ø      Self Adapting Numerical Algorithms (SANS) Effort

Ø       Generic Code Optimization

Ø      LAPACK For Clusters – easy access to clusters

¨      Heterogeneous Distributed Computing

Ø      NetSolve, FT-MPI, Open-MPI

¨      Performance Evaluation

Ø      PAPI, HPC Challenge, Top500

¨      Software Repositories

Ø      Netlib

·    Project Plan

·          has clear timeline and description of expected outcomes

LAPACK

•         Used by Matlab, Mathematica, Numeric Python,…

•         Tuned version provided by vendors: AMD, Apple, Compaq, Cray, Fujitsu, Hewlett-Packard, Hitachi, IBM, Intel, MathWorks, NAG, NEC, PGI, SUN, Visual Numerics, by Microsoft and most of Linux distribution (Fedora, Debian, Cygwin,...).

•         On going work: performance, accuracy, extended precision, ease of use

ScaLAPACK

•         Parallel implementation of LAPACK scaling on parallel hardware from 10’s to 100’s to 1000’s of processors

•         On going work: Match functionalities of current LAPACK

•         On going work: Target new architectures, new parallel environment. For example port to Microsoft HPC cluster solution

LAPACK for Clusters (LFC)

•         Most of ScaLAPACK functionality from serial clients (Matlab, Python, Mathematica)

FT-MPI and Open-MPI

·        Define the behavior of MPI in event a failure occurs at the process level.

·        FT-MPI based on MPI 1.3 (plus some MPI 2 features) with a fault tolerant model similar to what was done in PVM.

o       Complete reimplementation, not based on other implementations.

·        Gives the application the possibility to recover from a process-failure.

·        A regular, non fault-tolerant MPI program will run using FT-MPI.

·        What FT-MPI does not do:

o       Recover user data (e.g. automatic check-pointing)

o       Provide transparent fault-tolerance

Performance evaluation tools

·        Performance Application Programming Interface (PAPI)

o       A portable library to access hardware counters found on processors

o       Provides a standardized list of performance metrics

·        KOJAK (Joint with Felix Wolf)

o       Software package for the automatic performance analysis of parallel apps

§         Message passing and multi-threading (MPI and/or OpenMP)

§         Parallel performance

§         CPU and memory performance

• FT-MPI
• HPCC
• Kojak
• LAPACK / ScaLAPACK
• NetSolve / ActiveSheets
• NetSolve / .NET
• Open MPI
• PAPI
• top500