Performance Tuning of Scientific Applications

With contributions from some of the most notable experts in the field, Performance Tuning of Scientific Applications presents current research in performance analysis. The book focuses on the following areas.

Performance monitoring: Describes the state of the art in hardware and software tools that are commonly used for monitoring and measuring performance and managing large quantities of data
Performance analysis: Discusses modern approaches to computer performance benchmarking and presents results that offer valuable insight into these studies
Performance modeling: Explains how researchers deduce accurate performance models from raw performance data or from other high-level characteristics of a scientific computation
Automatic performance tuning: Explores ongoing research into automatic and semi-automatic techniques for optimizing computer programs to achieve superior performance on any computer platform
Application tuning: Provides examples that show how the appropriate analysis of performance and some deft changes have resulted in extremely high performance

Performance analysis has grown into a full-fledged, sophisticated field of empirical science. Describing useful research in modern performance science and engineering, this book helps real-world users of parallel computer systems to better understand both the performance vagaries arising in scientific applications and the practical means for improving performance.
Read about the book on HPCwire and insideHPC

David Bailey is a chief technologist in the High Performance Computational Research Department at the Lawrence Berkeley National Laboratory. Dr. Bailey has published several books and numerous research studies on computational and experimental mathematics. He has been a recipient of the ACM Gordon Bell Prize, the IEEE Sidney Fernbach Award, and the MAA Chauvenet Prize and Merten Hasse Prize.
Robert Lucas is the director of computational sciences in the Information Sciences Institute and a research associate professor in computer science in the Viterbi School of Engineering at the University of Southern California. Dr. Lucas has many years of experience working with high-end defense, national intelligence, and energy applications and simulations. His linear solvers are the computational kernels of electrical and mechanical CAD tools.
Samuel Williams is a researcher in the Future Technologies Group at the Lawrence Berkeley National Laboratory. Dr. Williams has authored or co-authored thirty technical papers, including several award-winning papers. His research interests include high-performance computing, auto-tuning, computer architecture, performance modeling, and VLSI.

Introduction, David H. BaileyBackground "Twelve Ways to Fool the Masses"Examples from Other Scientific Fields Guidelines for Reporting High Performance Modern Performance Science

Parallel Computer Architecture, Samuel W. Williams and David H. BaileyIntroduction Parallel Architectures Processor (Core) Architecture Memory Architecture Network Architecture Heterogeneous Architectures

Software Interfaces to Hardware Counters, Shirley V. Moore, Daniel K. Terpstra, and Vincent M. WeaverIntroduction Processor Counters Off-Core and Shared Counter Resources Platform ExamplesOperating System Interfaces PAPI in Detail Counter Usage Modes Uses of Hardware Counters Caveats of Hardware Counters

Measurement and Analysis of Parallel Program Performance using TAU and HPCToolkit, Allen D. Malony, John Mellor-Crummey, and Sameer S. ShendeIntroduction Terminology Measurement ApproachesHPCToolkit Performance ToolsTAU Performance System

Trace-Based Tools, Jesus LabartaIntroduction Tracing and Its Motivation Challenges Data Acquisition Techniques to Identify StructureModels InteroperabilityThe Future

Large-Scale Numerical Simulations on High-End Computational Platforms, Leonid Oliker, Jonathan Carter, Vincent Beckner, John Bell, Harvey Wasserman, Mark Adams, Stéphane Ethier, and Erik SchnetterIntroduction HPC Platforms and Evaluated Applications GTC: Turbulent Transport in Magnetic Fusion GTC Performance OLYMPUS: Unstructured FEM in Solid Mechanics Carpet: Higher-Order AMR in Relativistic AstrophysicsCASTRO: Compressible Astrophysics MILC: Quantum Chromodynamics

Performance Modeling: The Convolution Approach, David H Bailey, Allan Snavely, and Laura CarringtonIntroduction Applications of Performance Modeling Basic Methodology Performance Sensitivity Studies

Analytic Modeling for Memory Access Patterns Based on Apex-MAP, Erich Strohmaier, Hongzhang Shan, and Khaled IbrahimIntroduction Memory Access CharacterizationApex-MAP Model to Characterize Memory Access Patterns Using Apex-MAP to Assess Processor Performance Apex-MAP Extension for Parallel Architectures Apex-MAP as an Application ProxyLimitations of Memory Access Modeling

The Roofline Model, Samuel W. WilliamsIntroductionThe Roofline Bandwidth CeilingsIn-Core CeilingsArithmetic Intensity WallsAlternate Roofline Models

End-to-End Auto-Tuning with Active Harmony, Jeffrey K. Hollingsworth and Ananta TiwariIntroduction OverviewSources of Tunable DataSearchAuto-Tuning Experience with Active Harmony

Languages and Compilers for Auto-Tuning, Mary Hall and Jacqueline ChameLanguage and Compiler TechnologyInteraction between Programmers and Compiler Triage Code Transformation Higher-Level Capabilities

Empirical Performance Tuning of Dense Linear Algebra Software, Jack Dongarra and Shirley MooreBackground and MotivationATLASAuto-Tuning for Multicore Auto-Tuning for GPUs

Auto-Tuning Memory-Intensive Kernels for Multicore, Samuel W. Williams, Kaushik Datta, Leonid Oliker, Jonathan Carter, John Shalf, and Katherine YelickIntroduction Experimental SetupComputational KernelsOptimizing PerformanceAutomatic Performance Tuning Results

Flexible Tools Supporting a Scalable First-Principles MD Code, Bronis R. de Supinski, Martin Schulz, and Erik W. DraegerIntroduction Qbox: A Scalable Approach to First-Principles Molecular DynamicsExperimental Setup and BaselinesOptimizing Qbox: Step by Step Customizing Tool Chains with PN MPI

The Community Climate System Model, Patrick H. WorleyIntroduction CCSM Overview Parallel Computing and the CCSM Case Study: Optimizing Interprocess Communication Performance in the Spectral Transform Method Performance Portability: Supporting Options and Delaying DecisionsCase Study: Engineering Performance Portability into the Community Atmosphere Model Case Study: Porting the Parallel Ocean Program to the Cray X1 Monitoring Performance Evolution Performance at Scale
Tuning an Electronic Structure Code, David H. Bailey, Lin-Wang Wang, Hongzhang Shan, Zhengji Zhao, Juan Meza, Erich Strohmaier, and Byounghak LeeIntroduction LS3DF Algorithm Description LS3DF Code Optimizations Test Systems Performance Results and Analysis Science Results
Bibliography
Index