Contents

• 1 Introduction
• 2 Related Work
• 3 Source Code Instrumentation for Regions
  • 3.1 Regions
  • 3.2 Subprograms
  • 3.3 User Regions
  • 3.4 Instrumentation of OpenMP Programs
  • 3.5 Instrumentation of MPI Programs
  • 3.6 Instrumentation of HPF Programs
  • 3.7 Region Information Files
  
  
• 4 Source Code Instrumentation for Data Structures
  • 4.1 Array Runtime Descriptors
  • 4.2 Instrumentation for Data Profiling
  • 4.3 Information about Data Layout
  
  
• 5 Compiling and Linking of Instrumented Programs
  • 5.1 Using the ADAPTOR Compiler Driver
  • 5.2 Linking for Performance Monitoring
  • 5.3 Linking for Tracing
  
  
• 6 Counting Performance Events For Regions
  • 6.1 The Idea of Event Counting
  • 6.2 Event Counting with ADAPTOR
  • 6.3 Brutto and Netto Counting
  • 6.4 Composed Counters
  • 6.5 Derived Counter
  • 6.6 Output of Region Event Couting
  
  
• 7 Tracing of Performance Events and Visualization
  • 7.1 The Idea of Tracing
  • 7.2 Tracing With ADAPTOR
  • 7.3 Performance Visualization With VAMPIR
  
  
• 8 Profiling of Communication
  • 8.1 Counters for Communication
  • 8.2 Communication Statistics
  • 8.3 Tracing of Communication
  
  
• 9 Data Profiling
  • 9.1 The Idea of Data Profiling
  • 9.2 Data Profiling Support for the Intel Pentium 4
  • 9.3 Data Profiling Support with ADAPTOR
  • 9.4 Example of Data Profiling
  
  
• 10 Inconveniences and Extensions
• Bibliography
• A. Appendix: Configuration File for Performance Monitoring
• B. Appendix: PCL Counters
• C. Appendix: PAPI Counters
• D. Appendix: Supported Events on Linux Intel Pentium 4 Platforms
• About this document ...