Numbering Code	G-INF05 63541 LJ11 G-INF05 63541 LJ54	Year/Term	2022 ・ Second semester
Number of Credits	2	Course Type	Lecture
Target Year		Target Student
Language	Japanese	Day/Period	Tue.5
Instructor name	FUKAZAWA KEIICHIROU (Academic Center for Computing and Media Studies Associate Professor)
Outline and Purpose of the Course	This course will cover the functions and configuration methods of high-performance parallel systems such as supercomputer systems, as well as high-performance computing technology and parallel processing technology in scientific computing. We plan to use the supercomputer in the Academic Information Media Center. This course will be offered as a 5th period course so that graduate students belonging to all graduate schools conducting education and research related to computational science can easily take it.
Course Goals	You will acquire the following knowledge and abilities necessary for high-performance programming related to computational science: Knowledge of the architecture of the high-performance systems that executes the programs covered in this course, and important technical matters related to the design of high-performance program based on this knowledge. High-performance program implementation methods, concrete optimization / parallelization techniques, all based on actual examples.
Schedule and Contents	Regarding computer architecture from the viewpoint of computational science, we will cover concrete practical application examples in computational science and actual examples using the supercomputer in the Media Center. Furthermore, using the wave equation as an example, we will learn optimization, parallelization, use of GPU, etc. for implementing high-performance programs. Through these tasks, you will acquire competencies that will be of use when conducting research using supercomputers. ・ Introduction to parallel computing / high performance computing (1 lecture) ・ Introduction to applications for supercomputers (1 lecture) ・ Differentiation of wave equation (3 lectures) ・ Processor architecture and high performance computing (3 lectures) ・ Shared memory architecture and high performance computing (2 lectures) ・ Distributed memory architecture and high performance computing (2 lectures) ・ Implementation for accelerators such as GPU (2 lectures) ・ Cutting-edge research in high-performance computing (1 lecture)
Evaluation Methods and Policy	Your understanding of lecture content will be evaluated in one report assignment. The goal is for students to acquire the knowledge and skills necessary to carry out advanced parallelization simulations in their specialized fields. In particular, the purpose is to acquire the knowledge and experience necessary for applying advanced program tuning technology and various parallelization methods to specialized fields after fully understanding the structure of large-scale parallel computing systems.
Course Requirements	In order to acquire introductory knowledge of MPI and OpenMP in advance, we will assume that students have taken "Simulation Science" or "Computational Science Exercises (A or B)" or have taken the "Parallel Programming Course" held by the Academic Information Media Center.
Study outside of Class (preparation and review)	To prepare, review the contents of each prerequisite lecture/course, and reconfirm your acquisition of prerequisite knowledge. In addition, review the details of and concepts related to the technical concepts that will be introduced in this course by examining the prescribed reference texts.
Textbooks	Textbooks/References	In addition, we will distribute materials such as lecture slides (or download instructions from the web page).
References, etc.	"OpenMP Application Program Interface." http://www.openmp.org/mp-documents/spec30.pdf "MPI: A Message-Passing Interface Standard (Version 3.0)" http://www.mpi-forum.org/docs/mpi-3.0/mpi30-report.pdf "学術情報メディアセンター　スーパーコンピュータの利用手引．" http://web.kudpc.kyoto-u.ac.jp/hpc/tebiki