Programming and Performance Visualization Tools : International Workshops, ESPT 2017 and VPA 2017, Denver, CO, USA, November 12 and 17, 2017, and ESPT 2018 and VPA 2018, Dallas, TX, USA, November 16 and 11, 2018, Revised Selected Papers

This book contains the revised selected papers of 4 workshops held in conjunction with the International Conference on High Performance Computing, Networking, Storage and Analysis (SC) in November 2017 in Denver, CO, USA, and in November 2018 in Dallas, TX, USA: the 6th and 7th International Workshop on Extreme-Scale Programming Tools, ESPT 2017 and ESPT 2018, and the 4th and 5th International Workshop on Visual Performance Analysis, VPA 2017 and VPA 2018. The 11 full papers of ESPT 2017 and ESPT 2018 and the 6 full papers of VPA 2017 and VPA 2018 were carefully reviewed and selected for inclusion in this book. The papers discuss the requirements for exascale-enabled tools as well as new approaches of applying visualization and visual analytic techniques to large-scale applications. Topics of interest include: programming tools; methodologies for performance engineering; tool technologies for extreme-scale challenges (e.g., scalability, resilience, power); tool support for accelerated architectures and large-scale multi-cores; tool infrastructures and environments; evolving/future application requirements for programming tools and technologies; application developer experiences with programming and performance tools; scalable displays of performance data; case studies demonstrating the use of performance visualization in practice; data models to enable scalable visualization; graph representation of unstructured performance data; presentation of high-dimensional data; visual correlations between multiple data sources; human-computer interfaces for exploring performance data; and multi-scale representations of performance data for visual exploration. -- Provided by publisher Preface 6 ESPT 2017 6 Organizing Committee 7 Program Committee 7 ESPT 2018 7 Organizing Committee 8 Program Committee 9 VPA 2017 9 Workshop Chairs 10 Steering Committee 10 Program Committee 10 VPA 2018 11 Workshop Chairs 12 Steering Committee 12 Program Committee 12 Contents 13 ESPT 2017 15 Enhancing PAPI with Low-Overhead rdpmc Reads 16 1 Introduction 16 2 Background 18 2.1 Performance Counter Hardware 18 2.2 Linux perf_event Interface 18 2.3 PAPI Library 19 2.4 Linux rdpmc Support 19 2.5 PAPI rdpmc Code 19 2.6 Linux rdpmc Bugs Found 21 3 Related Work 22 3.1 Lower-Level Interface Overhead 22 3.2 PAPI Overhead 23 3.3 Other Performance Counter Tools 23 4 Experimental Setup 24 5 Results 25 5.1 Outliers 28 5.2 Historical Comparison 30 6 Conclusion and Future Work 31 References 31 Generic Library Interception for Improved Performance Measurement and Insight 34 1 Introduction 34 2 Related Work 36 3 Methodology 37 3.1 Library Call Interception 37 3.2 Workflow 38 3.3 Implementation Details 44 4 Case Study 44 4.1 GROMACS 45 4.2 PERMON 47 5 Conclusions 48 6 Future Work 48 References 49 Improved Accuracy for Automated Communication Pattern Characterization Using Communication Graphs and Aggressive Search Space Pruning 51 1 Introduction 52 2 Characterizing Application Communication 53 2.1 Augmented Communication Graphs 53 2.2 Non-greedy Volume Attribution 57 2.3 Search Space Pruning 58 3 Implementation 59 4 Evaluation 60 4.1 Augmented Communication Graphs 60 4.2 Aggressive Pruning 62 5 Case Study: Xolotl 64 6 Related Work 66 7 Summary and Future Work 67 References 67 Moya—A JIT Compiler for HPC 69 1 Introduction 69 2 Motivation 70 2.1 Programmer Annotations 71 2.2 Compile-Time JIT-Aware Static Analysis 71 2.3 Dynamic JIT-Time Optimizations 71 3 Moya 72 4 Programmer Annotations 72 5 Compile-Time Analysis 73 5.1 Identification of Dynamic Constants 73 5.2 Mutability Analysis 74 5.3 Library Models 77 6 JIT - Time Optimizations 77 6.1 Function Argument Specialization 77 6.2 Dynamic Constant Propagation (DCP) 78 6.3 Invariant Load Detection 78 7 Results 79 7.1 Compile-Time Static Analysis 80 7.2 PlasComCM 80 7.3 NAS Parallel Benchmarks 81 8 Related Work 82 9 Future Work 83 10 Conclusion 83 References 84 Polyhedral Optimization of TensorFlow Computation Graphs 87 1 Introduction 87 2 Design 89 2.1 Overview 89 2.2 Subgraph Selection 89 2.3 Operator Code Generators 91 2.4 Subgraph Code Generator 91 2.5 R-Stream Optimization 91 2.6 TensorFlow Operator 94 2.7 Leveraging Broadcast 94 3 Experiments 95 4 Enabled Experiments/Work 99 5 Related Work 99 6 Conclusion 100 References 101 CAASCADE: A System for Static Analysis of HPC Software Application Portfolios 103 1 Introduction 103 2 Background and Related Work 105 3 Design and Methods 106 3.1 GNU Compiler Plugin Implementation 107 3.2 Database Infrastructure 109 4 Results 111 5 Conclusions and Future Work 113 References 115 Visual Comparison of Trace Files in Vampir 118 1 Introduction 118 2 Related Work 120 3 Methodology 121 3.1 Comparing Application Characteristics Using Charts 121 3.2 Aligning Traces Manually 122 3.3 Aligning Traces Automatically Using Predefined Markers 123 3.4 Aligning Traces Automatically Using Call Invocation Profiles 124 4 Case Study 125 4.1 LSMS – Comparing Performance Between Different Hardware 126 4.2 CloverLeaf – Comparing Performance Between Programming Models 127 4.3 Trinity RNA-Seq Assembler – Comparing Performance Between Different Process Numbers 129 5 Conclusions 132 References 132 ESPT 2018 135 Understanding the Scalability of Molecular Simulation Using Empirical Performance Modeling 136 1 Introduction 136 2 ms2 Application 138 3 Methodology 139 3.1 Simulation Parameters 139 3.2 Benchmarking 141 3.3 Empirical Modeling 143 4 Experimental Setup 145 4.1 Parameter Values 145 4.2 Measurements Variability 146 5 Result Analysis 147 6 Related Work 151 7 Conclusion 152 References 153 Advanced Event-Sampling Support for PAPI 155 1 Introduction 155 1.1 Hardware Performance Counters 156 1.2 Advanced Sampling 156 1.3 Software Interfaces 157 2 Hardware Sampling Interfaces 157 2.1 Intel x86_64 158 2.2 AMD x86_64 160 2.3 Other Processors 161 3 Software Interface for Sampling 161 3.1 Linux perf_event Interface 161 3.2 PAPI Library Interface 164 4 Related Work 165 4.1 Existing Profiling Tools 165 4.2 NUMA Profiling 166 4.3 GPU Profiling 166 4.4 Other Tools with Sampling Interfaces 166 4.5 Other Proposed PAPI Sampling Interfaces 167 5 Proposed Advanced PAPI Sampling API 167 5.1 Abstracted Interface 167 5.2 Direct perf_event Interface 168 6 Preliminary Results 170 7 Conclusion and Future Work 171 References 171 ParLoT: Efficient Whole-Program Call Tracing for HPC Applications 173 1 Introduction 173 2 Background and Related Work 175 2.1 Binary Instrumentation 176 2.2 Efficient Tracing 177 3 Design of ParLoT 177 3.1 Tracing Operation 178 3.2 Incremental Compression 179 3.3 Compression Algorithm 179 3.4 PIN and Call-Stack Correction 180 4 Evaluation Methodology 181 4.1 Benchmarks and System 181 4.2 Metrics 181 4.3 Tracing Tools 182 5 Results 184 5.1 Tracing Overhead 184 5.2 Required Bandwidth 186 5.3 Compression Ratio 187 5.4 Overheads 188 5.5 Compression Impact 190 6 Discussion and Conclusion 191 References 193 Gotcha: An Function-Wrapping Interface for HPC Tools 196 1 Introduction 196 2 Background and Related Work 197 2.1 Background 197 2.2 Related Work 198 3 Gotcha Abstractions and Implementation 200 3.1 Gotcha Wrapping Abstraction 200 3.2 Multi-tool Support 201 3.3 Interface-Independent Wrapping 202 4 Use Cases 204 4.1 Caliper 204 4.2 Generic MPI Wrapper 204 5 Performance/Results 205 6 Future Work 207 7 Conclusions 207 References 208 VPA 2017 209 Projecting Performance Data over Simulation Geometry Using SOSflow and ALPINE 210 1 Introduction 210 1.1 Research Contributions 211 2 Related Work 212 3 SOSflow 212 3.1 SOSflow Daemons 213 3.2 SOSflow Client Library 216 3.3 SOSflow Data 217 4 ALPINE Ascent 218 5 Experiments 220 5.1 Evaluation Platform 220 5.2 Experiment Setup 220 5.3 Overview of Processing Steps 221 5.4 Evaluation of Geometry Extraction 222 5.5 Evaluation of Overhead 222 6 Results 222 6.1 Geometry Extraction and Performance Data Projection 223 6.2 Overhead 223 7 Conclusion 225 7.1 Future Work 225 References 226 Visualizing, Measuring, and Tuning Adaptive MPI Parameters 228 1 Introduction 228 2 Visualizing AMPI with Projections 230 2.1 Implementation 230 2.2 Visualizations 231 3 Application Case Studies 233 3.1 LULESH 233 3.2 Particle-in-cell 235 4 Related Work 237 5 Conclusions 237 References 238 VPA 2018 240 Visual Analytics Challenges in Analyzing Calling Context Trees 241 1 Introduction 241 2 HPC Domain Data 242 3 State of the Art 243 4 Data/Visual Analytics Operations on a CCT 245 5 Prototype of a Flow-Based Visualization Framework 246 5.1 Flow-Based Navigation 246 5.2 A First Flow Example 248 5.3 Alternative Function View 250 5.4 Data and Visualization Challenges 251 6 Conclusion 254 References 255 PaScal Viewer: A Tool for the Visualization of Parallel Scalability Trends 258 1 Introduction 258 2 The PaScal Viewer 260 2.1 The Color Diagrams 260 2.2 Input File Format 263 3 Case Studies 264 3.1 The Blackscholes Application 264 3.2 The Bodytrack Application 266 3.3 The Freqmine Application 268 4 Related Works 270 5 Conclusion and Future Works 270 References 271 Using Deep Learning for Automated Communication Pattern Characterization: Little Steps and Big Challenges 273 1 Introduction 274 2 Integrating Deep Learning into AChax 276 2.1 Training 276 2.2 Recognition and Parameterization 277 3 Early Experiments 278 4 Summary 279 References 280 Visualizing Multidimensional Health Status of Data Centers 281 1 Introduction 281 2 Existing Approaches 282 3 Visualization Components 283 3.1 HPC System Spatial Layout 285 3.2 Multidimensional Analysis of Health Status 286 4 Discussion and Future Work 288 5 Conclusion 290 References 290 Author Index 292 Front Matter ....Pages i-xiv Front Matter ....Pages 1-1 Enhancing PAPI with Low-Overhead rdpmc Reads (Yan Liu, Vincent M. Weaver)....Pages 3-20 Generic Library Interception for Improved Performance Measurement and Insight (Ronny Brendel, Bert Wesarg, Ronny Tschüter, Matthias Weber, Thomas Ilsche, Sebastian Oeste)....Pages 21-37 Improved Accuracy for Automated Communication Pattern Characterization Using Communication Graphs and Aggressive Search Space Pruning (Philip C. Roth)....Pages 38-55 Moya—A JIT Compiler for HPC (Tarun Prabhu, William Gropp)....Pages 56-73 Polyhedral Optimization of TensorFlow Computation Graphs (Benoît Pradelle, Benoît Meister, Muthu Baskaran, Jonathan Springer, Richard Lethin)....Pages 74-89 CAASCADE: A System for Static Analysis of HPC Software Application Portfolios (M. Graham Lopez, Oscar Hernandez, Reuben D. Budiardja, Jack C. Wells)....Pages 90-104 Visual Comparison of Trace Files in Vampir (Matthias Weber, Ronny Brendel, Michael Wagner, Robert Dietrich, Ronny Tschüter, Holger Brunst)....Pages 105-121 Front Matter ....Pages 123-123 Understanding the Scalability of Molecular Simulation Using Empirical Performance Modeling (Sergei Shudler, Jadran Vrabec, Felix Wolf)....Pages 125-143 Advanced Event-Sampling Support for PAPI (Forrest Smith, Vincent M. Weaver)....Pages 144-161 ParLoT: Efficient Whole-Program Call Tracing for HPC Applications (Saeed Taheri, Sindhu Devale, Ganesh Gopalakrishnan, Martin Burtscher)....Pages 162-184 Gotcha: An Function-Wrapping Interface for HPC Tools (David Poliakoff, Matt LeGendre)....Pages 185-197 Front Matter ....Pages 199-199 Projecting Performance Data over Simulation Geometry Using SOSflow and ALPINE (Chad Wood, Matthew Larsen, Alfredo Gimenez, Kevin Huck, Cyrus Harrison, Todd Gamblin et al.)....Pages 201-218 Visualizing, Measuring, and Tuning Adaptive MPI Parameters (Matthias Diener, Sam White, Laxmikant V. Kale)....Pages 219-230 Front Matter ....Pages 231-231 Visual Analytics Challenges in Analyzing Calling Context Trees (Alexandre Bergel, Abhinav Bhatele, David Boehme, Patrick Gralka, Kevin Griffin, Marc-André Hermanns et al.)....Pages 233-249 PaScal Viewer: A Tool for the Visualization of Parallel Scalability Trends (Anderson B. N. da Silva, Daniel A. M. Cunha, Vitor R. G. Silva, Alex F. de A. Furtunato, Samuel Xavier-de-Souza)....Pages 250-264 Using Deep Learning for Automated Communication Pattern Characterization: Little Steps and Big Challenges (Philip C. Roth, Kevin Huck, Ganesh Gopalakrishnan, Felix Wolf)....Pages 265-272 Visualizing Multidimensional Health Status of Data Centers (Tommy Dang)....Pages 273-283 Back Matter ....Pages 285-285 The two volume set LNCS 8033 and 8034 constitutes the refereed proceedings of the 9th International Symposium on Visual Computing, ISVC 2013, held in Rethymnon, Crete, Greece, in July 2013. The 63 revised full papers and 35 poster papers presented together with 32 special track papers were carefully reviewed and selected from more than 220 submissions. The papers are organized in topical sections: Part I (LNCS 8033) comprises computational bioimaging; computer graphics; motion, tracking, and recognition; segmentation; visualization; 3D mapping, modeling and surface reconstruction; feature extraction, matching, and recognition; sparse methods for computer vision, graphics, and medical imaging; and face processing and recognition. Part II (LNCS 8034) comprises topics such as visualization; visual computing with multimodal data streams; visual computing in digital cultural heritage; intelligent environments: algorithms and applications; applications; and virtual reality The two volume sets LNCS 8033 and 8034 constitutes the refereed proceedings of the 9th International Symposium on Visual Computing, ISVC 2013, held in Rethymnon, Crete, Greece, in July 2013. The 63 revised full papers and 35 poster papers presented together with 32 special track papers were carefully reviewed and selected from more than 220 submissions. The papers are organized in topical sections: Part I (LNCS 8033) comprises computational bioimaging; computer graphics; motion, tracking and recognition; segmentation; visualization; 3D mapping, modeling and surface reconstruction; feature extraction, matching and recognition; sparse methods for computer vision, graphics and medical imaging; and face processing and recognition. Part II (LNCS 8034) comprises topics such as visualization; visual computing with multimodal data streams; visual computing in digital cultural heritage; intelligent environments: algorithms and applications; applications and virtual reality. The two volume set LNCS 8887 and 8888 constitutes the refereed proceedings of the 10th International Symposium on Visual Computing, ISVC 2014, held in Las Vegas, NV, USA. The 74 revised full papers and 55 poster papers presented together with 39 special track papers were carefully reviewed and selected from more than 280 submissions. The papers are organized in topical sections: Part I (LNCS 8887) comprises computational bioimaging, computer graphics; motion, tracking, feature extraction and matching, segmentation, visualization, mapping, modeling and surface reconstruction, unmanned autonomous systems, medical imaging, tracking for human activity monitoring, intelligent transportation systems, visual perception and robotic systems. Part II (LNCS 8888) comprises topics such as computational bioimaging , recognition, computer vision, applications, face processing and recognition, virtual reality, and the poster sessions.