TAU Parallel Performance System Allen D. Malony, Sameer S. Shende, Robert Bell Kai Li, Li Li, Kevin Huck {malony,sameer,bertie,likai,lili,khuck}@cs.uoregon.edu Department of Computer and Information Science Performance Research Laboratory University of Oregon

Outline Motivation TAU architecture and toolkit Example applications Instrumentation Measurement Analysis Example applications Users of TAU Conclusion

Problem Domain ASCI defines leading edge parallel systems and software Large-scale systems and heterogenous platforms Multi-model simulation Complex software integration Multi-language programming Mixed-model parallelism Complexity challenges performance analysis tools System diversity requires portable tools Need for cross-language support Coverage of parallel computation models Operate at scale

Research Motivation Tools for performance problem solving Empirical-based performance optimization process Performance technology concerns characterization Performance Tuning Performance Diagnosis Experimentation Observation hypotheses properties Experiment management Performance database Performance Technology Instrumentation Measurement Analysis Visualization Performance Technology

TAU Performance System Tuning and Analysis Utilities (11+ year project effort) 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 / parallelism Measurement and analysis abstraction Integrated toolkit for performance instrumentation, measurement, analysis, and visualization Portable performance profiling and tracing facility Open software approach with technology integration University of Oregon , Forschungszentrum Jülich, LANL

TAU Performance Systems Goals 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 Support for performance mapping Support for object-oriented and generic programming Integration in complex software systems and applications

General Complex System Computation Model Node: physically distinct shared memory machine Message passing node interconnection network Context: distinct virtual memory space within node Thread: execution threads (user/system) in context Interconnection Network * Inter-node message communication * Node Node Node Should we describe different parallel programming / computation models? physical view memory node memory SMP memory VM space … model view … Context Threads

TAU Performance System Architecture EPILOG Paraver

TAU Instrumentation Approach Support for standard program events Routines Classes and templates Statement-level blocks Support for user-defined events Begin/End events (“user-defined timers”) Atomic events Selection of event statistics Support definition of “semantic” entities for mapping Support for event groups Instrumentation optimization

TAU Instrumentation Flexible instrumentation mechanisms at multiple levels Source code manual automatic C, C++, F77/90/95 (Program Database Toolkit (PDT)) OpenMP (directive rewriting (Opari), POMP spec) Object code pre-instrumented libraries (e.g., MPI using PMPI) statically-linked and dynamically-linked Executable code dynamic instrumentation (pre-execution) (DynInstAPI) virtual machine instrumentation (e.g., Java using JVMPI)

TAU Source Instrumentation Automatic source instrumentation (TAUinstr) Routine entry/exit and class method entry/exit Block entry/exit and statement level (to be added) Uses an instrumentation specification file Include/exclude list for events and files Uses command line options for group selection Instrumentation event selection (TAUselect) Automatic generation of instrumentation specification file Instrumentation language to describe event constraints Event identity and location Event performance properties (e.g., overhead analysis) Create TAUselect scripts for performance experiments

Multi-Level Instrumentation Targets common measurement interface TAU API Multiple instrumentation interfaces Simultaneously active Information sharing between interfaces Utilizes instrumentation knowledge between levels Selective instrumentation Available at each level Cross-level selection Targets a common performance model Presents a unified view of execution Consistent performance events

Program Database Toolkit (PDT) Program code analysis framework develop source-based tools High-level interface to source code information Integrated toolkit for source code parsing, database creation, and database query Commercial grade front-end parsers Portable IL analyzer, database format, and access API Open software approach for tool development Multiple source languages Implement automatic performance instrumentation tools tau_instrumentor

Program Database Toolkit (PDT) Application / Library C / C++ parser Fortran parser F77/90/95 Program documentation PDBhtml Application component glue IL IL SILOON C / C++ IL analyzer Fortran IL analyzer C++ / F90/95 interoperability CHASM Program Database Files Automatic source instrumentation DUCTAPE TAU_instr

PDT 3.0 Functionality C++ statement-level information implementation for, while loops, declarations, initialization, assignment… PDB records defined for most constructs DUCTAPE Processes PDB 1.x, 2.x, 3.x uniformly PDT applications XMLgen PDB to XML converter (Sottile) Used for CHASM and CCA tools PDBstmt Statement callgraph display tool

PDT 3.0 Functionality (continued) Cleanscape Flint parser fully integrated for F90/95 Flint parser is very robust Produces PDB records for TAU instrumentation (stage 1) Linux x86, HP Tru64, IBM AIX Tested on SAGE, POP, ESMF, PET benchmarking codes Full PDB 2.0 specification (stage 2) [Q1 ‘04] Statement level support (stage 3) [Q3 ‘04] Open64 parser integrated in PDT for F90/95 Barbara Chapman, University of Houston Generate full PDB 2.0 specification (stage 2) [Q2 ‘04] PDT 3.0 release at SC2003

TAU Performance Measurement TAU supports profiling and tracing measurement Robust timing and hardware performance support Support for online performance monitoring Profile and trace performance data export to file system Selective exporting Extension of TAU measurement for multiple counters Creation of user-defined TAU counters Access to system-level metrics Support for callpath measurement Integration with system-level performance data Linux MAGNET/MUSE (Wu Feng, LANL)

TAU Measurement with Multiple Counters Extend event measurement to capture multiple metrics Begin/end (interval) events User-defined (atomic) events Multiple performance data sources can be queried Associate counter function list to event Defined statically or dynamically Different counter sources Timers and hardware counters User-defined counters (application specified) System-level counters Monotonically increasing required for begin/end events Extend user-defined counters to system-level counter

TAU Measurement Performance information Organization Performance events High-resolution timer library (real-time / virtual clocks) General software counter library (user-defined events) Hardware performance counters PCL (Performance Counter Library) (ZAM, Germany) PAPI (Performance API) (UTK, Ptools Consortium) consistent, portable API Organization Node, context, thread levels Profile groups for collective events (runtime selective) Performance data mapping between software levels

TAU Measurement Options Parallel profiling Function-level, block-level, statement-level Supports user-defined events TAU parallel profile data stored during execution Hardware counts values Support for multiple counters Support for callgraph and callpath profiling Tracing All profile-level events Inter-process communication events Trace merging and format conversion

Grouping Performance Data in TAU Profile Groups A group of related routines forms a profile group Statically defined TAU_DEFAULT, TAU_USER[1-5], TAU_MESSAGE, TAU_IO, … Dynamically defined group name based on string, such as “adlib” or “particles” runtime lookup in a map to get unique group identifier uses tau_instrumentor to instrument Ability to change group names at runtime Group-based instrumentation and measurement control

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) Trace format conversion (ALOG, SDDF, VTF, Paraver) Trace visualization using Vampir (Pallas)

Pprof Output (NAS Parallel Benchmark – LU) Intel Quad PIII Xeon F90 + MPICH Profile - Node - Context - Thread Events - code - MPI

ParaProf (NAS Parallel Benchmark – LU) Routine profile across all nodes node,context, thread Global profiles Event legend Individual profile

TAU + PAPI (NAS Parallel Benchmark – LU ) Floating point operations Re-link to alternate library Can use multiple counter support

TAU + Vampir (NAS Parallel Benchmark – LU) Callgraph display Timeline display Parallelism display Communications display

TAU Performance 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, 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

Selected Applications of TAU Center for Simulation of Accidental Fires and Explosion University of Utah, ASCI ASAP Center, C-SAFE Uintah Computational Framework (UCF) (C++) Center for Simulation of Dynamic Response of Materials California Institute of Technology, ASCI ASAP Center Virtual Testshock Facility (VTF) (Python, Fortran 90) Los Alamos National Lab Monte Carlo transport (MCNP) (Susan Post) Full code automatic instrumentation and profiling ASCI Q validation and scaling SAIC’s Adaptive Grid Eulerian (SAGE) (Jack Horner) Fortran 90 automatic instrumentation and profiling

Selected Applications of TAU (continued) Lawrence Livermore National Lab Radiation diffusion (KULL) C++ automatic instrumentation, callpath profiling Sandia National Lab DOE CCTTSS SciDAC project Common component architecture (CCA) integration Combustion code (C++, Fortran 90) Flash Center University of Chicago / Argonne, ASCI ASAP Center FLASH code (C, Fortran 90)

Performance Analysis and Visualization Analysis of parallel profile and trace measurement Parallel profile analysis ParaProf ParaVis Profile generation from trace data Performance database framework (PerfDBF) Parallel trace analysis Translation to VTF 3.0 and EPILOG Integration with VNG (Technical University of Dresden) Online parallel analysis and visualization

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 AMR toolkit (SAMRAI++), LLNL

Full Profile Window (Exclusive Time) 512 processes

Node / Context / Thread Profile Window

Derived Metrics

Full Profile Window (Metric-specific) 512 processes

ParaProf Enhancements Readers completely separated from the GUI Access to performance profile database Profile translators mpiP, papiprof, dynaprof Callgraph display prof/gprof style with hyperlinks Integration of 3D performance plotting library Scalable profile analysis Statistical histograms, cluster analysis, … Generalized programmable analysis engine Cross-experiment analysis

ParaVis Scalable parallel profile analysis Scalable performance displays 3D graphics Analysis across profile samples Allow for runtime use Animated / interactive visualization Initially develop with SCIRun Computational environment Performance graphics toolkit Portable plotting library OpenGL Performance Visualizer Performance Analyzer Performance Data Reader

Performance Visualization in SCIRun SCIRun program EVH1, IBM EVH1, Linux IA-32

“Terrain” Visualization (Full profile) Uintah

“Scatterplot” Visualization Each point coordinate determined by three values: MPI_Reduce MPI_Recv MPI_Waitsome Min/Max value range Effective for cluster analysis Uintah

“Bargraph” Visualization (MPI routines) Uintah, 512 processes, ASCI Blue Pacific

Empirical-Based Performance Optimization Experiment Schemas Experiment Trials management Process characterization Performance Tuning Performance Diagnosis Experimentation Observation hypotheses properties ? observability requirements

TAU Performance Database Framework Performance analysis programs Performance data description Raw performance data Other tools PerfDML translators Performance analysis and query toolkit . . . ORDB PostgreSQL PerfDB profile data only XML representation project / experiment / trial

PerfDBF Components Performance Data Meta Language (PerfDML) Common performance data representation Performance meta-data description PerfDML translators to common data representation Performance DataBase (PerfDB) Standard database technology (SQL) Free, robust database software (PostgresSQL, MySQL) Commonly available APIs Performance DataBase Toolkit (PerfDBT) Commonly used modules for query and analysis PerfDB API to facilitate analysis tool development

PerfDBF Browser

PerfDBF Cross-Trial Analysis

TAU Application (Selected) SAMRAI (LLNL) Overture (LLNL) C-SAFE (ASCI ASAP, University of Utah) VTF (ASCI ASAP, Caltech) SAGE (ASCI LANL) POOMA, POOMA-II (LANL, Code Sourcery) PETSc (ANL) CCA (DOE SciDAC) GrACE (Rutgers University) DOE ACTS toolkit Aurora / SCALEA (University of Vienna)

Work in Progress Trace visualization Event traces with counters (Vampir 3.0 will visualize) EPILOG trace conversion Runtime performance monitoring and analysis Online performance data access Performance analysis and visualization in SCIRun Performance Database Framework XML parallel profile representation of TAU profiles PostgresSQL performance database Next-generation PDT Performance analysis for component software (CCA)

Concluding Remarks Complex software and parallel computing systems pose challenging performance 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 consistently and coherently in software and system environments TAU performance system offers robust performance technology that can be broadly integrated … so USE IT!

Acknowledgements Department of Energy (DOE) MICS office DOE 2000 ACTS contract “Performance Technology for Tera-class Parallel Computer Systems: Evolution of the TAU Performance System” PERC SciDAC project affiliate 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

Case Study: SAMRAI (LLNL) Structured Adaptive Mesh Refinement Application Infrastructure (SAMRAI) Programming C++ and MPI SPMD Instrumentation PDT for automatic instrumentation of routines MPI interposition wrappers SAMRAI timers for interesting code segments timers classified in groups (apps, mesh, …) timer groups are managed by TAU groups

SAMRAI (Profile) Euler (2D) routine name return type

SAMRAI Euler (Profile)

SAMRAI Euler (Trace)

Case Study: EVH1 Enhanced Virginia Hydrodynamics #1 (EVH1) "TeraScale Simulations of Neutrino-Driven Supernovae and Their Nucleosynthesis" SciDAC project Configured to run a simulation of the Sedov-Taylor blast wave solution in 2D spherical geometry Performance study found EVH1 communication bound for more than 64 processors Predominant routine (>50% of execution time) at this scale is MPI_ALLTOALL Used in matrix transpose-like operations

EVH1 Execution Profile

EVH1 Execution Trace MPI_Alltoall is an execution bottleneck