Blue Gene/L: Performance Analysis Tools

Blue Gene/L: Performance Analysis Tools

by Gary L. Mullen-Schultz
Released July 2006
Publisher(s): IBM Redbooks
ISBN: None

Read it now on the O’Reilly learning platform with a 10-day free trial.

O’Reilly members get unlimited access to books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.

Start your free trial

Book description

This IBM Redbooks publication is one in a series of IBM publications written specifically for the IBM System Blue Gene supercomputer, Blue Gene/L, which was developed by IBM in collaboration with Lawrence Livermore National Laboratory (LLNL). This book provides an overview of the application development performance analysis environment for Blue Gene/L.

This book explains some of the tools that are available to do application-level performance analysis. It devotes the majority of its content to Chapter 3, "External Performance Instrumentation Facility" and Chapter 4, "Performance Application Programming Interface."

Table of contents

  1. Notices
    1. Trademarks
  2. Preface
    1. The team that wrote this redbook
    2. Become a published author
    3. Comments welcome
  3. Chapter 1: Performance guidelines and tools
    1. Tooling overview
      1. IBM High Performance Computing Toolkit
    2. General performance testing
      1. Overview of the tools that are available on System p
      2. Overview of tools ported to Blue Gene/L
    3. Message passing performance
      1. MPI Tracer and Profiler
    4. CPU performance
      1. Hardware performance monitor
      2. Xprofiler
    5. I/O performance
      1. Modular I/O
    6. Visualization and analysis
      1. PeekPerf
    7. MASS and MASSV libraries
  4. Chapter 2: Comparison of performance tools
    1. External Performance Instrumentation Facility
    2. Performance Application Programming Interface
    3. Summary comparison of Perfmon and PAPI
  5. Chapter 3: External Performance Instrumentation Facility
    1. Overview of EPIF and Perfmon
      1. Objectives
      2. Goals and strategies
    2. Basic concepts
    3. EPIF commands
      1. perfmon
      2. dsp_perfmon (1/4)
      3. dsp_perfmon (2/4)
      4. dsp_perfmon (3/4)
      5. dsp_perfmon (4/4)
      6. ext_perfmon_data
      7. exp_perfmon_data
      8. imp_perfmon_data
      9. end_perfmon
    4. Building the necessary Python packages
    5. Typical command uses
      1. Options for EPIF
      2. Options for ext_perfmon_data
      3. dsp_perfmon
      4. exp_perfmon_data
      5. imp_perfmon_data
  6. Chapter 4: Performance Application Programming Interface
    1. PAPI implementation
      1. The linux-bgl PAPI substrate
      2. PAPI event mapping for Blue Gene/L
      3. Modifications to PAPI
    2. Examples of using hardware performance monitor libraries for Blue Gene/L
      1. PAPI library usage examples
      2. bgl_perfctr usage example (1/2)
      3. bgl_perfctr usage example (2/2)
    3. Conclusion
  7. Appendix A: Statement of completion
  8. Appendix B: Electromagnetic compatibility
  9. Appendix C: Perfmon database table specifications
    1. Database organization
    2. Performance collection instance table: BGLPERFINST
    3. Performance definition table: BGLPERFDEF
    4. Performance description table: BGLPERFDESC
    5. Performance job table: BGLPERFJOB
    6. Performance location table: BGLPERFLOCATION
    7. Performances samples definition table: BGLPERFSAMPLES
    8. Performance data file table: BGLPERFDATA
    9. BGLPERFDESC table (1/3)
    10. BGLPERFDESC table (2/3)
    11. BGLPERFDESC table (3/3)
    12. BGLPERFDEF and BGLPERFDESC table join (1/6)
    13. BGLPERFDEF and BGLPERFDESC table join (2/6)
    14. BGLPERFDEF and BGLPERFDESC table join (3/6)
    15. BGLPERFDEF and BGLPERFDESC table join (4/6)
    16. BGLPERFDEF and BGLPERFDESC table join (5/6)
    17. BGLPERFDEF and BGLPERFDESC table join (6/6)
  10. Appendix D: gmon support on Blue Gene/L
    1. How to enable gmon profiling
    2. Additional function in Blue Gene/L gmon support
      1. Multiple gmon.out.x files
      2. Enabling or disabling profiling within your application
      3. Collecting gmon data as set of program counter values instead of as histogram
    3. Enhancements to gprof in the Blue Gene/L toolchain
      1. Using gprof to read gmon.sample.x files
      2. Using gprof to merge a very large number of gmon.out.x files
  11. Glossary (1/2)
  12. Glossary (2/2)
  13. Related publications
    1. IBM Redbooks
    2. Other publications
    3. Online resources
    4. How to get IBM Redbooks
    5. Help from IBM
  14. Index
  15. Back cover

Product information

  • Title: Blue Gene/L: Performance Analysis Tools
  • Author(s): Gary L. Mullen-Schultz
  • Release date: July 2006
  • Publisher(s): IBM Redbooks
  • ISBN: None