Intro -- Preface -- Organization -- Keynote Speech and Invited Talks -- BenchCouncil Achievement Award -- Benchmarking Supercomputers in the Post-Moore Era -- Benchmarks and Middleware for Designing Convergent HPC, Big Data and Deep Learning Software Stacks for Exascale Systems -- InfiniBand In-Network Computing Technology for Scalable HPC/AI -- Benchmarking Perspectives on Emerging HPC Workloads -- Lightweight Requirements Engineering for Exascale Co-design -- FloraBench: An End-to-End Application Benchmark Suite for Datacenter -- Towards Benchmarking AIOT Device Based on MCU
Harmonizing High-Level Abstraction and High Performance for Graph Mining -- Deep Learning on HPC: Performance Factors and Lessons Learned -- Towards a Methodology for Benchmarking Edge Processing Frameworks -- Contents -- Best Paper Session -- Performance Analysis of GPU Programming Models Using the Roofline Scaling Trajectories -- 1 Introduction -- 2 Motivation -- 3 Performance on GPU Architecture -- 4 GPU Programming Models in This Study -- 5 GPU Roofline Scaling Trajectories -- 6 Experimental Setup -- 6.1 Benchmark Suite -- 6.2 System Setup
7 Scaling Trajectory Analysis for Computational Kernels -- 7.1 GPU Cache Impact on Scaling Trajectories -- 8 Performance Tuning -- 9 Related Work -- 10 Conclusions -- References -- GraphBench: A Benchmark Suite for Graph Computing Systems -- 1 Introduction -- 2 Related Work -- 3 The Methodology of GraphBench -- 4 Basic Operations of Graph Computing -- 5 The Implementations of GraphBench -- 5.1 Component Benchmarks -- 5.2 Micro Benchmarks and Fine-Grained Analysis -- 5.3 Data Sets -- 5.4 Software Stacks -- 5.5 Summary -- 6 Evaluations -- 6.1 Experimental Configurations and Methodology
6.2 Experiment Results and Observations -- 7 Conclusion -- References -- Early Experience in Benchmarking Edge AI Processors with Object Detection Workloads -- 1 Introduction -- 2 Background -- 2.1 Inference in Deep Learning -- 2.2 Object Detection and YOLO-Based Systems -- 3 Overview of Edge AI Processors -- 3.1 Edge TPU -- 3.2 NVIDIA Xavier -- 3.3 NovuTensor -- 4 Benchmarking Methodology -- 4.1 Workload Selection -- 4.2 Platform Selection -- 4.3 Metrics and Dimensions -- 4.4 Experimental Methodology -- 5 Experiments -- 5.1 Hardware Configuration -- 5.2 Setup -- 5.3 Results
5.4 Observations and Summary -- 6 Related Work -- 6.1 Modern AI Processors -- 6.2 Edge AI Benchmarking -- 7 Conclusion and Future Work -- References -- AI Challenges on Cambricon Using AIBench -- XDN: Towards Efficient Inference of Residual Neural Networks on Cambricon Chips -- 1 Introduction -- 2 Our Approach -- 2.1 Pruning Optimizer and Trainer -- 2.2 Fusion Optimizer -- 2.3 Auto-tuner -- 2.4 Executor and Evaluator -- 3 Experimental Evaluation -- 3.1 Environmental Setting -- 3.2 Overall Performance -- 3.3 Analysis and Discussion -- 4 Conclusion -- References
Summary
This book constitutes the refereed proceedings of the Second International Symposium on Benchmarking, Measuring, and Optimization, Bench 2019, held in Denver, CO, USA, in November 2019. The 20 full papers and 11 short papers presented were carefully reviewed and selected from 79 submissions. The papers are organized in topical sections named: Best Paper Session; AI Challenges on Cambircon using AIBenc; AI Challenges on RISC-V using AIBench; AI Challenges on X86 using AIBench; AI Challenges on 3D Face Recognition using AIBench; Benchmark; AI and Edge; Big Data; Datacenter; Performance Analysis; Scientific Computing.-- Provided by publisher
