1 / 21

TAU Performance System tau.uoregon

TAU Performance System http://tau.uoregon.edu. tau-team@cs.uoregon.edu. TAU Parallel Performance System. http://tau.uoregon.edu/ Multi-level performance instrumentation Multi-language automatic source instrumentation Flexible and configurable performance measurement

thu
Télécharger la présentation

TAU Performance System tau.uoregon

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. TAU Performance Systemhttp://tau.uoregon.edu tau-team@cs.uoregon.edu

  2. TAU Parallel Performance System • http://tau.uoregon.edu/ • Multi-level performance instrumentation • Multi-language automatic source instrumentation • Flexible and configurable performance measurement • Widely-ported parallel performance profiling system • Computer system architectures and operating systems • Different programming languages and compilers • Support for multiple parallel programming paradigms • Multi-threading, message passing, mixed-mode, hybrid

  3. TAU Port to IBM BG/P • Supports automatic instrumentation at: • Source level (PDT, tau_instrumentor; KOJAK, opari) • MPI • Flexible and configurable performance measurement • Support for profiling and tracing • Support for PAPI counters on BG/P • Uses bgxlC_r, bgxlc_r, bgxlf90_r as compilers • To configure TAU: • ./installtau -arch=bgp -mpi -pdt=<dir> -pdt_c++=xlC -papi=<dir> • ./tau_validate --html --build bgp >& results.html • Parallel Profile Analysis: • Paraprof profile browser • PerfDMF profile database • Perfexplorer cross-experiment data analysis toolkit

  4. Using TAU on IBM BGP (surveyor.alcf.anl.gov) • Choose measurement configuration % ls /soft/apps/tau/tau_latest/bgp/lib/Makefile.* Makefile.tau-mpi-pdt Makefile.tau-mpi-pdt-trace Makefile.tau-callpath-mpi-pdt Makefile.tau-callpath-mpi-compensate-pdt Makefile.tau-depthlimit-mpi-pdt Makefile.tau-mpi-compensate-pdt Makefile.tau-multiplecounters-mpi-papi-pdt Makefile.tau-multiplecounters-mpi-papi-pdt-trace Makefile.tau-multiplecounters-papi-pdt Makefile.tau-multiplecounters-pthread-papi-pdt Makefile.tau-pdt Makefile.tau-phase-multiplecounters-mpi-compensate-papi-pdt Makefile.tau-phase-multiplecounters-mpi-papi-pdt Makefile.tau-pthread-pdt … % setenv TAU_MAKEFILE /soft/apps/tau/tau-2.17/bgp/lib/Makefile.tau-mpi-pdt % set path=(/soft/apps/tau/tau-2.17/ppc64/bin $path) # Front-end binaries • Replace mpixlf90_r with tau_f90.sh and compile your application • Use tau_cxx.sh and tau_cc.sh for C++ and C compilers respectively

  5. Using TAU on IBM BGP (surveyor.alcf.anl.gov) • Choose measurement configuration % ls /soft/apps/tau/tau_latest/bgp/lib/Makefile.* Makefile.tau-mpi-pdt Makefile.tau-mpi-pdt-trace Makefile.tau-callpath-mpi-pdt Makefile.tau-callpath-mpi-compensate-pdt Makefile.tau-depthlimit-mpi-pdt Makefile.tau-mpi-compensate-pdt Makefile.tau-multiplecounters-mpi-papi-pdt Makefile.tau-multiplecounters-mpi-papi-pdt-trace Makefile.tau-multiplecounters-papi-pdt Makefile.tau-multiplecounters-pthread-papi-pdt Makefile.tau-pdt Makefile.tau-phase-multiplecounters-mpi-compensate-papi-pdt Makefile.tau-phase-multiplecounters-mpi-papi-pdt Makefile.tau-pthread-pdt … % setenv TAU_MAKEFILE /soft/apps/tau/tau-2.17/bgp/lib/Makefile.tau-mpi-pdt % set path=(/soft/apps/tau/tau-2.17/ppc64/bin $path) # Front-end binaries • Replace mpixlf90_r with tau_f90.sh and compile your application • Use tau_cxx.sh and tau_cc.sh for C++ and C compilers respectively • Visualize performance data with paraprof, pprof, vampir, jumpshot

  6. TAU’s ParaProf 3D Profile Browser: Matmult

  7. Profiling FLASH3 on IBM BG/P

  8. Sedov 2D Auto • Initial test run did not include a load balanced problem • Small problem: too little work for 1024 processor • Proof of concept to validate porting of tools

  9. PerfExplorer: Cross Experiment Analysis

  10. TAU PerfExplorer: Runtime Breakdown MPI_Barrier IO_OUTPUT

  11. Relative Efficiency

  12. Relative Speedup for One Event

  13. TAU’s PerfExplorer: IBM BG/P

  14. TAU Portal • TAU portal supports the FLASH regression testing • Allows groups to share profiling data in a secure way • Allows users to launch TAU performance tools (paraprof, perfexplorer) • Nightly regression testcases uploaded to the database automatically • SVN checkout each night • TAU: http://www.cs.uoregon.edu/research/tau • TAU Portal: http://tau.nic.uoregon.edu

  15. Portal: Nightly Performance Regression Testing

  16. TAU Portal: Launch ParaProf/PerfExplorer

  17. PerfExplorer: Regression Testing

  18. PerfExplorer: Limiting Events (> 3% ), Oct 2007

  19. PerfExplorer: Exclusive Time for Events (2007)

  20. ParaProf: 3D Visualization

  21. Support Acknowledgements • Department of Energy (DOE) • Office of Science • LLNL, LANL, ASC • Argonne National Laboratory • University of Chicago • Department of Defense • NSF

More Related