Course Overview
High performance computing is the basic technology needed to support today’s large scale scientific simulations. It covers a wide variety of issues on hardware and software for high-end computing such as high speed computation, high speed networking, large scale memory and disk storage, high speed numerical algorithms, programming schemes and the system softwares to support them. Current advanced supercomputer systems are based on large scale parallel processing systems. Nowadays, even application users are required to understand these technologies to a certain level for their effective utilization. In this class, we focus on the basic technology of high-end computing systems, programming, algorithm and performance tuning for application users who aim to use these systems for their practical simulation and computing.
Lecture Day and Location
| Lecture Day: | February 20 (Mon), 21 (Tue), 2023 |
|---|---|
| Location: | Hybrid form |
| Onsite: International Workshop Room, Center for Computational Sciences (Access) | |
| Online: Zoom (Zoom link will be sent by email.) |
Schedule
| Feb. 20 (Mon) | Feb. 21 (Tue) | |
| 09:00 – 10:30 | Fundamentals of HPC and Parallel Processing | Parallel Programming 2: OpenMP |
| 10:45 – 12:15 | Parallel Processing Systems | Parallel Numerical Algorithm 1 |
| 13:30 – 15:00 | Parallel Programming 1: MPI | Parallel Numerical Algorithm 2 |
| 15:15 – 16:45 | Optimization 1: Communication Optimization | Optimization 2: Computation Optimization |
Contents
| Lecture name | Contents | Instructor | |
|---|---|---|---|
| 1 | Fundamentals of HPC and Parallel Processing | Amdahl’s law, Parallelization methods (EP, Data parallelism, Pipeline parallelism), Communication, Synchronization, Parallelization efficiency, Load balance. | Taisuke Boku |
| 2 | Parallel Processing Systems | Parallel processing systems (SMP, NUMA, Cluster, Grid, etc.), Memory hierarchy, Memory bandwidth, Network, Communication bandwidth, Delay. | Ryohei Kobayashi |
| 3 | Parallel Programming 1: MPI | Parallel programming language MPI. | Norihisa Fujita |
| 4 | Optimization 1: Communication Optimization | Optimization techniques and performance evaluation of parallel programming on parallel processing systems. | Osamu Tatebe |
| 5 | Parallel Programming 2: OpenMP | Parallel programming model, parallel programming language OpenMP. | Akira Nukada |
| 6 | Parallel Numerical Algorithm 1 | Krylov subspace iterative methods and their parallelization methods. | Hiroto Tadano |
| 7 | Parallel Numerical Algorithm 2 | Fast Fourier Transformation (FFT) and its parallelization methods. | Daisuke Takahashi |
| 8 | Optimization 2: Computation Optimization | Program optimization techniques (Register blocking, Cache blocking, Memory allocation, etc.) and performance evaluation on a compute node of parallel processing systems. | Daisuke Takahashi |
*本セミナーは，筑波大学理工情報生命学術院共通専門基盤科目「High Performance Parallel Computing Technology for Computational Sciences」(0AH0209)と共通です． 本セミナーを授業として受講する方は，TWINS で履修登録して下さい．