The TAU Performance Technology for Complex Parallel .ppt

《The TAU Performance Technology for Complex Parallel .ppt》由会员分享，可在线阅读，更多相关《The TAU Performance Technology for Complex Parallel .ppt（35页珍藏版）》请在麦多课文档分享上搜索。

1、The TAU Performance Technology for Complex Parallel Systems (Performance Analysis Bring Your Own Code Workshop, NRL Washington D.C.) Sameer Shende, Allen D. Malony, Robert Bell University of Oregon sameer, malony, bertiecs.uoregon.edu,Outline,Motivation Part I: Instrumentation Part II: Measurement P

2、art III: Analysis Tools Conclusion,TAU Performance System Framework,Tuning and Analysis Utilities Performance system framework for scalable parallel and distributed high-performance computing Targets a general complex system computation model nodes / contexts / threads Multi-level: system / software

3、 / parallelism Measurement and analysis abstraction Integrated toolkit for performance instrumentation, measurement, analysis, and visualization Portable, configurable performance profiling/tracing facility Open software approach University of Oregon, LANL, FZJ Germany http:/www.cs.uoregon.edu/resea

4、rch/paracomp/tau,TAU Performance System Architecture,paraprof,TAU Analysis,Parallel profile analysis pprof parallel profiler with text-based display paraprof Graphical, scalable, parallel profile analysis and display Trace analysis and visualization Trace merging and clock adjustment (if necessary)

5、Trace format conversion (ALOG, SDDF, VTF, Paraver) Trace visualization using Vampir (Pallas/Intel),Pprof Output (ESMF CoupledFlowSolver),IBM AIX F95, C+, C, MPI Profile - Node - Context - Thread Events - code - MPI,Terminology Example,For routine “int main( )”: Exclusive time 100-20-50-20=10 secs In

6、clusive time 100 secs Calls 1 call Subrs (no. of child routines called) 3 Inclusive time/call 100secs,int main( ) /* takes 100 secs */f1(); /* takes 20 secs */f2(); /* takes 50 secs */f1(); /* takes 20 secs */* other work */ /* Time can be replaced by counts */,Performance Analysis and Visualization

7、,Analysis of parallel profile and trace measurement Parallel profile analysis ParaProf Cube Profile Browser (UTK, FZJ) Profile generation from trace data Performance data management framework (PerfDMF) Parallel trace analysis Translation to VTF 3.0 and EPILOG Integration with VNG (Technical Universi

8、ty of Dresden) Online parallel analysis and visualization,TAUs ParaProf Framework Architecture,Portable, extensible, and scalable tool for profile analysis Try to offer “best of breed” capabilities to analysts Build as profile analysis framework for extensibility,Profile Manager Window,Structured AM

9、R toolkit (SAMRAI+), LLNL,Paraprof: CoupledFlowApp (ESMF) on 4 Nodes,Paraprof Mean Profile (4 nodes),Individual Node (0) Profile in Paraprof,MPI Routines,Text Profile Window,k-Level Callpath Implementation in TAU,TAU maintains a performance event (routine) callstack Profiled routine (child) looks in

10、 callstack for parent Previous profiled performance event is the parent A callpath profile structure created first time parent calls TAU records parent in a callgraph map for child String representing k-level callpath used as its key “a( )=b( )=c()” : name for time spent in “c” when called by “b” wh

11、en “b” is called by “a” Map returns pointer to callpath profile structure k-level callpath is profiled using this profiling data Set environment variable TAU_CALLPATH_DEPTH to depth Build upon TAUs performance mapping technology Measurement is independent of instrumentation Use PROFILECALLPATH to co

12、nfigure TAU,k-Level Callpath Implementation in TAU,Examining Callpaths,Unique Callpaths,Gprof Style Parent, Routine, Children Display,Clickable Callpath Entities,Paraprof,Tracking I/O on Node 0 in ESMF,Calling Path for MPI_Recv( ),CUBE (UTK, FZJ) Browser Sept. 2004,Using TAU with Vampir (Intel Trace

13、 Analyzer),Configure TAU with -TRACE option % configure TRACE mpi Execute application % poe CoupledFlowApp procs 4 This generates TAU traces and event descriptors Merge all traces using tau_merge % tau_merge *.trc app.trc Convert traces to Vampir Trace format using tau_convert % tau_convert pv app.t

14、rc tau.edf app.pv Note: Use vampir instead of pv for multi-threaded traces Load generated trace file in Vampir % vampir app.pv,Global Timeline Display with Parallelism View,Vampir: Zooming In,Vampir: IO on Node 0,Vampir: Communication Matrix Display,Vampir: Calltree View,Summary Chart,TAU Performanc

15、e System Status,Computing platforms (selected) IBM SP / pSeries, SGI Origin 2K/3K, Cray T3E / SV-1 / X1, HP (Compaq) SC (Tru64), Sun, Hitachi SR8000, NEC SX-5/6, Linux clusters (IA-32/64, Alpha, PPC, PA-RISC, Power, Opteron), Apple (G4/5, OS X), Windows Programming languages C, C+, Fortran 77/90/95,

16、 HPF, Java, OpenMP, Python Thread libraries pthreads, SGI sproc, Java,Windows, OpenMP Compilers (selected) Intel KAI (KCC, KAP/Pro), PGI, GNU, Fujitsu, Sun, Microsoft, SGI, Cray, IBM (xlc, xlf), Compaq, NEC, Intel,Concluding Remarks,Complex parallel systems and software pose challenging performance

17、analysis problems that require robust methodologies and tools To build more sophisticated performance tools, existing proven performance technology must be utilized Performance tools must be integrated with software and systems models and technology Performance engineered software Function consisten

18、tly and coherently in software and system environments TAU performance system offers robust performance technology that can be broadly integrated,Support Acknowledgements,Department of Energy (DOE) Office of Science contracts University of Utah DOE ASCI Level 1 sub-contract DOE ASCI Level 3 (LANL, LLNL) NSF National Young Investigator (NYI) award Research Centre Juelich John von Neumann Institute for Computing Dr. Bernd Mohr Los Alamos National Laboratory,