Application Performance on Commodity-class and High-end Computers

11/1/2004


Click here to start


Table of Contents

Application Performance on Commodity-class and High-end Computers

Outline

Capability and Capacity Computing - Cost and Performance

PNNL’s HPCS2 - 1960 Processor HP Supercluster

DisCo: Technical Progress in 2003-4

Working with the Community

Working with the Community II.

Commodity Systems (CSx) Prototype / Evaluation Hardware

Commodity Systems (CSx) II.

High-End Systems Evaluated

Applications Performance Overview

Performance Metrics: 1999-2001

Commodity Comparisons with High-end Systems

Commodity Comparisons with High-end Systems

SPEC CPU 2000 - SPECfp2000 Values relative to HP RX5670 Itanium2/1.5GHz

The GAMESS-UK Serial Benchmark

Single CPU performance - A Case Study

Interconnects and Networking

Communication Benchmarks PMB: Pallas MPI Benchmark Suite (V2.2) and B_EFF

PingPong Performance

Interconnect Benchmark - EFF_BW

Interconnect Benchmark - Latency

Collective Operations (Time - usec) - as function of no.of CPUs

MPI_allreduce Performance

64-CPU Relative Performance for Allreduce

128-CPU Relative Performance for Allreduce

64-CPU Relative Performance for Allgather

128-CPU Relative Performance for Allgather

64-CPU Performance for all collective operations

Application Codes

Performance Metrics: 2004

Molecular Simulation

DL_POLY: A Parallel MD Package

DL_POLY Parallel Benchmarks (Cray T3E/1200)

DL_POLY V2: Bench 4 - Commodity-based Systems

DL_POLY V2: Bench 5 - Commodity-based Systems

DL_POLY V2: Bench 7 - Commodity-based Systems

Migration from Replicated to Distributed data: DL_POLY-3 : Domain Decomposition

Migration from Replicated to Distributed data DL_POLY-3: Coulomb Energy Evaluation

DL_POLY3 Coulomb Energy Evaluation

DL_POLY3 Macromolecular Simulations

DL_POLY3 Macromolecular Simulations

DLMULTI

DLMULTI: High-end and Commodity-based Systems

DLMULTI: Commodity-based Systems

CHARMM

Parallel CHARMM Benchmark

Molecular Electronic Structure

CCP1: Molecular Electronic Structure

Molecular Electronic Structure

Distributed Data SCF

Global Arrays

Global Array Benchmark I. GET

Global Array Benchmark II. PUT

Global Array Benchmark III. ACCUMULATE

High-End Computational Chemistry The NWChem Software

Case Studies - Zeolite Fragments

DFT Coulomb Fit - NWChem 4.6

DFT Coulomb Fit - NWChem 4.6

Exploiting Global Memory: NWChem

Exploiting HPC: The PNNL Collaboration

GAMESS-UK

GAMESS-UK features 1.

GAMESS-UK features 2.

Parallel Implementation of GAMESS-UK

Parallel Implementation of GAMESS-UK

Parallel Implementations of GAMESS-UK

Parallel Linux Implementations of GAMESS-UK

Parallel Linear Algebra

PeIGS 3.0 Parallel Performance

Eigensolver Performance - “Small” case (IBM p690+)

PDSYEVD – IBM p690+ and SGI Altix 3700

Eigensolver Performance - “Largest” Case (IBM p690+)

GAMESS-UK ?SCF Performance IBM SP/p690, High-end and Commodity-based Systems

GAMESS-UK. DFT B3LYP Performance IBM SP/p690+, High-end and Commodity-based Systems

GAMESS-UK: DFT HCTH on Valinomycin. Impact of Coulomb Fitting

GAMESS-UK - GA-based Implementation

GAMESS-UK - GA-based Implementation

DFT BLYP Gradient: High-end and Commodity-based Systems

SCF and DFT Analytic 2nd Derivatives

SCF Analytic 2nd Derivatives Performance IBM SP/p690+, High-end and Commodity-based Systems

DFT Analytic 2nd Derivatives Performance Commodity-based Systems - HCTH functional

MP2 Gradient Algorithms

Performance of MP2 Gradient Module High-end and Commodity-based Systems

Distributed Data Implementation of GAMESS-UK

Coupled QM/MM Calculations

GAMESS-UK Interface with CHARMM

QM/MM Applications

Materials Simulation Codes

Materials Simulation. Plane Wave Methods: CASTEP, CPMD

CPMD - Car-Parrinello Molecular Dynamics

CPMD - Mixed Mode Programming

CPMD 3.7 - C120 Benchmark

CPMD 3.7 - C120 Benchmark

CPMD 3.7 - Si512 Benchmark

Computational Engineering

ANGUS: Combustion modelling (regular grid) High-end and Commodity-based Systems

ANGUS: Combustion modelling (regular grid) High-end and Commodity-based Systems

ANGUS: Combustion modelling (regular grid) Memory Bandwidth Effects: Pentium Xeon and Opteron Systems

ANGUS: Combustion modelling (regular grid) High-end and Commodity-based Systems

Computational Engineering: UK Turbulence Consortium

Direct Numerical Simulation: 3603 benchmark

DNS: 1203 benchmark. High-end and Commodity-based Systems

DNS: 1203 benchmark. High-end and Commodity-based Systems

Commodity Comparisons with High-end Systems

Commodity Comparisons with High-end Systems

Summary

Author: Martyn F. Guest

Email: m.f.guest@dl.ac.uk