Sustained Simulation Performance 2021 : proceedings of the 32nd Joint Workshop on Sustained Simulation Performance, University of Stuttgart (HLRS) and Tohoku University, 2021

This book presents the state of the art in High-Performance Computing on modern supercomputer architectures. It addresses trends in hardware and software development in general. The contributions cover a broad range of topics, from performance evaluations in context with power efficiency to Computational Fluid Dynamics and High-Performance Data Analytics. In addition, they explore new topics like the use of High-Performance Computers in the field of Artificial Intelligence and Machine Learning. All contributions are based on selected papers presented in 2021 at the 31st Workshop on Sustained Simulation Performance, WSSP31, held at HLRS in Stuttgart, Germany, and WSSP32, held at Tohoku University in Sendai, Japan. Preface 5 Contents 6 Supercomputer for Quest to Unsolved Interdisciplinary Datascience (SQUID) and its Five Challenges 7 1 Introduction 7 2 Hardware configuration of SQUID 8 3 Five challenges behind SQUID 14 3.1 Tailor-made computing 14 3.2 HPC and HPDA integration 15 3.3 Cloud-interlinked and -synergized 17 3.4 Secure computing environment 18 3.5 Data aggregation environment 19 4 Future issues towards next-generation infrastructure 21 4.1 Tradeoff between user experience and performance for practical cloud bursting 22 4.2 Software and tools to support integrated way of HPC and HPDA 23 4.3 Evaluation of secure computing environment for an operational point of view 23 5 Summary 24 References 24 Simulating Molecular Docking on the SX-Aurora TSUBASA Vector Engine 26 1 Introduction 26 2 Background 27 2.1 Molecular docking 28 2.2 SX-Aurora TSUBASA Vector Engine 30 3 Methodology 31 3.1 Porting 32 3.2 Optimization 32 3.2.1 Main Optimizations 32 3.2.2 Additional Optimizations 34 4 Evaluation 34 4.1 Execution profiling 36 4.2 Performance comparison against CPUs and GPUs 37 5 Conclusions 38 References 39 Simulation of Field-induced Chiral Phenomena in Inhomogeneous Superconductivity 41 1 Field-induced chiral phenomena in inhomogeneous superconductivity 42 1.1 Chiral state 42 1.2 Field-induced chiral phenomena 42 2 Field-induced chiral phenomena in a eutectic superconductor 43 2.1 Inhomogeneous interface superconductivity 43 2.2 Chiral transition represented with order parameter 45 2.3 Simulation of field-induced chiral transition 46 2.4 Paramagnetic chiral supercurrent 51 3 Computation of field-induced chiral phenomena 51 4 Summary 52 References 54 Exploiting Hybrid Parallelism in the LBM Implementation Musubi on Hawk 57 1 The Lattice Boltzmann method 57 2 The Musubi implementation 59 2.1 OpenMP in Musubi 60 3 Hybrid parallelization 60 4 Performance assessment on Hawk 61 4.1 The Hawk computing system 61 4.2 The Musubi setup 61 4.3 Results 62 5 Conclusion 67 References 68 MPI Continuations And How To Invoke Them 70 1 Background 70 2 Current state 71 2.1 Continuations 72 2.2 Continuation Requests (CR) 72 2.3 Current API design 73 3 Callback interface requirements 75 3.1 Efficiency 75 3.2 Safety 76 3.3 Usability 76 4 Callback interface variations 76 4.1 Passing requests and user data 79 4.1.1 MPI-provided Request Buffer 80 4.1.2 User-provided Request Buffer 81 4.2 Passing only user data 82 5 C++ lambda capture 83 6 Conclusions 85 References 85 Xevolver for Performance Tuning of C Programs 87 1 Introduction 87 2 Related work 88 3 Xevolver for C 89 4 Evaluation and discussions 92 5 Conclusions 94 References 95 Scalability Evaluation of the CFD Solver CODA on the AMD Naples Architecture 96 1 Introduction 96 2 The CFD solver CODA 98 3 The test case for external aerodynamics 99 4 The CARA HPC system 100 5 Evaluation 101 5.1 Measurement setup 101 5.2 Evaluation of different hybrid-parallel setups 102 5.3 Evaluation of different mesh sizes and node placement 103 5.4 Comparison of AMD Naples and Intel Cascade Lake architectures 104 6 Conclusion 105 References 106