Performance Analyzer Tutorial

Performance Analyzer Tutorial ™Sun Studio 9 Sun Microsystems, Inc. www.sun.com 817-7624-10 May 2004 Submit comments about this document at: http://www.sun.com/hwdocs/feedbackCopyright © 2004 Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, California 95054, U.S.A. All rights reserved. U.S. Government Rights - Commercial software. Government users are subject to the Sun Microsystems, Inc. standard license agreement and applicable provisions of the FAR and its supplements. Use is subject to license terms. This distribution may include materials developed by third parties. Parts of the product may be derived from Berkeley BSD systems, licensed from the University of California. UNIX is a registered trademark in the U.S. and in other countries, exclusively licensed through X/Open Company, Ltd. Sun, Sun Microsystems, the Sun logo, Java, and JavaHelp are trademarks or registered trademarks of Sun Microsystems, Inc. in the U.S. and other countries.All SPARC trademarks are used under license and are trademarks or registered trademarks of SPARC International, Inc. in the U.S. and other countries. Products bearing SPARC trademarks are based upon architecture developed by Sun Microsystems, Inc. This product is covered and controlled by U.S. Export Control laws and may be subject to the export or import laws in other countries. Nuclear, missile, chemical biological weapons or nuclear maritime end uses or end users, whether direct or indirect, are strictly prohibited. Export or reexport to countries subject to U.S. embargo or to entities identified on U.S. export exclusion lists, including, but not limited to, the denied persons and specially designated nationals lists is strictly prohibited. DOCUMENTATION IS PROVIDED "AS IS" AND ALL EXPRESS OR IMPLIED CONDITIONS, REPRESENTATIONS AND WARRANTIES, INCLUDING ANY IMPLIED WARRANTY OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT, ARE DISCLAIMED, EXCEPT TO THE EXTENT THAT SUCH DISCLAIMERS ARE HELD TO BE LEGALLY INVALID. Copyright © 2004 Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, California 95054, Etats-Unis. Tous droits réservés. L’utilisation est soumise aux termes de la Licence. Cette distribution peut comprendre des composants développés par des tierces parties. Des parties de ce produit pourront être dérivées des systèmes Berkeley BSD licenciés par l’Université de Californie. UNIX est une marque déposée aux Etats-Unis et dans d’autres pays et licenciée exclusivement par X/Open Company, Ltd. Sun, Sun Microsystems, le logo Sun, Java, et JavaHelp sont des marques de fabrique ou des marques déposées de Sun Microsystems, Inc. aux Etats-Unis et dans d’autres pays.Toutes les marques SPARC sont utilisées sous licence et sont des marques de fabrique ou des marques déposées de SPARC International, Inc. aux Etats-Unis et dans d’autres pays. Les produits portant les marques SPARC sont basés sur une architecture développée par Sun Microsystems, Inc. Ce produit est soumis à la législation américaine en matière de contrôle des exportations et peut être soumis à la règlementation en vigueur dans d’autres pays dans le domaine des exportations et importations. Les utilisations, ou utilisateurs finaux, pour des armes nucléaires,des missiles, des armes biologiques et chimiques ou du nucléaire maritime, directement ou indirectement, sont strictement interdites. Les exportations ou réexportations vers les pays sous embargo américain, ou vers des entités figurant sur les listes d’exclusion d’exportation américaines, y compris, mais de manière non exhaustive, la liste de personnes qui font objet d’un ordre de ne pas participer, d’une façon directe ou indirecte, aux exportations des produits ou des services qui sont régis par la législation américaine en matière de contrôle des exportations et la liste de ressortissants spécifiquement désignés, sont rigoureusement interdites. LA DOCUMENTATION EST FOURNIE "EN L’ÉTAT" ET TOUTES AUTRES CONDITIONS, DECLARATIONS ET GARANTIES EXPRESSES OU TACITES SONT FORMELLEMENT EXCLUES, DANS LA MESURE AUTORISEE PAR LA LOI APPLICABLE, Y COMPRIS NOTAMMENT TOUTE GARANTIE IMPLICITE RELATIVE A LA QUALITE MARCHANDE, A L’APTITUDE A UNE UTILISATION PARTICULIERE OU A L’ABSENCE DE CONTREFAÇON.Contents Contents xiii Figures xvii Tables xxi Before You Begin iii How This Book Is Organized iii Typographic Conventions iv Shell Prompts v Accessing Sun Studio Software and Man Pages v Accessing Compilers and Tools Documentation viii Accessing Related Solaris Documentation xi Resources for Developers xii Contacting Sun Technical Support xii Sending Your Comments xiii 1. Learning to Use the Performance Tools 1 Setting Up the Examples for Execution 2 System Requirements 3 Choosing Alternative Compiler Options 3 xiiiBasic Features of the Performance Analyzer 4 2. Basic Performance Analysis 7 Collecting Data forsynprog 7 Simple Metric Analysis 8 Extension Exercise for Simple Metric Analysis 12 Metric Attribution and thegprof Fallacy 13 The Effects of Recursion 16 Loading Dynamically Linked Shared Objects 20 Descendant Processes 21 Extension Exercise for Descendant Processes 27 3. Analyzing the Performance of a Mixed Java/C++ Application 29 jsynprog Program Structure and Control Flow 29 Collecting Data forjsynprog 30 Analyzingjsynprog Program Data 31 4. OpenMP Parallelization Strategies 41 Collecting Data foromptest 41 Comparing Parallel Sections and Parallel Do Strategies 43 Comparing Critical Section and Reduction Strategies 45 5. Locking Strategies in Multithreaded Programs 47 Collecting Data formttest 47 How Locking Strategies Affect Wait Time 48 How Data Management Affects Cache Performance 53 Extension Exercises formttest 56 6. Cache Behavior and Optimization 57 Collecting Data forcachetest 57 Execution Speed 58 xiv Performance Analyzer • March 2004Program Structure and Cache Behavior 60 Program Optimization and Performance 63 A. Profiling Programs Withprof,gprof, andtcov 67 Usingprof to Generate a Program Profile 68 Usinggprof to Generate a Call Graph Profile 70 Usingtcov for Statement-Level Analysis 73 Creatingtcov Profiled Shared Libraries 76 Locking Files 77 Errors Reported bytcov Runtime Functions 77 Usingtcov Enhanced for Statement-Level Analysis 79 Creating Profiled Shared Libraries fortcov Enhanced 80 Locking Files 80 tcov Directories and Environment Variables 81 Index 83 Contents xvxvi Performance Analyzer • March 2004Figures FIGURE 2-1 Source Tab Showing Annotated Source Code for Functioncputime 9 FIGURE 2-2icputime 10 FIGURE 2-3 Disassembly Tab Showing Instructions for the Line in Whichx Is Incremented in Function cputime 11 FIGURE 2-4x icputime 12 FIGURE 2-5 Callers-Callees Tab Withgpf_work as the Selected Function 14 FIGURE 2-6 Source Tab Showing Annotated Source Code for Functionsgpf_a andgpf_b 15 FIGURE 2-7 Source Tab Showing Annotated Source Code for Functiongpf_work 16 FIGURE 2-8 Callers-Callees Tab Withreal_recurse as the Selected Function 17 FIGURE 2-9bounce_a 18 FIGURE 2-10bounce_b 19 FIGURE 2-11 Functions Tab Showing Functionsso_burncpu andsx_burncpu 21 FIGURE 2-12 Timeline Tab Showing the Seven Experiments Recorded for the Parent Process and its Descendant Processes 22 FIGURE 2-13 Timeline Tab at High Zoom Showing Event Markers and Gaps Between Them 24 FIGURE 2-14 Experiments Tab Showing Seven Experiments, Three of Which Are Marked as “Bad” 25 FIGURE 2-15 Timeline Tab at High Zoom, Showing Short Sample for Experiment test.2.er/_f2_f1.er 26 FIGURE 2-16 Event Tab Showing Very Short Duration Sample 27 FIGURE 3-1 The Functions Tab, Showing Severaljsynprog Experiment Methods 32 FIGURE 3-2 The Summary Panel 33 Figures xviiFIGURE 3-3 The Metrics Tab of the Set Data Presentation Dialog 33 FIGURE 3-4 The Callers-Callees Tab, Withjsynprog.main Selected 34 FIGURE 3-5 Source Tab, Listingjsynprog.java 35 FIGURE 3-6 The Disassembly Tab, Showing Annotated Bytecode 36 FIGURE 3-7 The Timeline Tab in the Java Representation 37 FIGURE 3-8 Setting Java Mode Using the Formats Tab of the Set Data Presentation Dialog Box 38 FIGURE 3-9 The Timeline Tab in the Expert-Java Representation 38 FIGURE 3-10 The Functions Tab, Showing Interpreted and Dynamically-compiled Versions of Routine.sys_op 39 FIGURE 4-1 Summary Tabs for Functionpsec_ From the Four-CPU Run (Left) and the Two-CPU Run (Right) 43 FIGURE 4-2pdo_ (Right) 44 FIGURE 4-3 Functions Tab Showing Entries forcritsum_ andredsum_ 45 FIGURE 5-1 Functions Tab for the Four-CPU Experiment Showing Data forlock_local and lock_global 49 FIGURE 5-2 Source Tab for the Four-CPU Experiment for Functionlock_global 50 FIGURE 5-3lock_local 51 FIGURE 5-4 Functions Tab for the One-CPU Experiment Showing Data forlock_local and lock_global 52 FIGURE 5-5 Functions Tab for the One-CPU Experiment Showing Data for FunctionscomputeA and computeB 53 FIGURE 5-6 Functions Tab for the Four-CPU Experiment Showing Data for FunctionscomputeA and computeB 54 FIGURE 5-7 Source Tab for the Four-CPU Experiment Showing Annotated Source Code forcomputeA andcomputeB 55 FIGURE 5-8 Source tab for the Four-CPU Experiment Showing Annotated Source Code for cache_trash 55 FIGURE 6-1 Functions Tab Showing User CPU, FP Adds and FP Muls for the Six Variants ofdgemv 59 FIGURE 6-2 Functions Tab Showing User CPU Time, CPU Cycles, Instructions Executed and D- and E- Cache Stall Cycles for the Six Variants ofdgemv 60 FIGURE 6-3 Source Tab Showing Annotated Source Code fordgemv_g1 anddgemv_g2 62 FIGURE 6-4 Source Tab fordgemv_hi1 Showing Compiler Commentary That Includes Loop Interchange Messages 64 Figures xviiiFIGURE 6-5 Source Tab fordgemv_hi2 Showing Compiler Commentary 65 FIGURE 6-6 Disassembly Tab fordgemv_hi1 Showing Non-Sequential Line Numbering and Instruction Repetition Due to Loop Unrolling 66 Figures xixxx Performance Analyzer • March 2004