HOME 大学院 科学技術計算Ⅰ

# 科学技術計算Ⅰ

OpenMP is the most widely-used way for parallelization on each compute node with multiple cores because multi-threading can be done easily by just inserting "directives". In this class, lectures and exercises for parallelization by multi-threading of the target application (Poisson’s equation solver by FVM (finite-volume method)) on multicore processors using OpenMP are provided, which covers numerical algorithms, and programming methods. Moreover, lectures on recent research topics on parallel preconditioning methods will be also provided. The "Supercomputer System (Wisteria/BDEC-01(Odyssey))" are available for hands-on exercises.

コース名

4820-1027
GIF-MA6213L3

S1 S2

2

NO

・有限体積法 ・クリロフ部分空間法 ・前処理 ・OpenMP入門 ・リオ－ダリング／色づけ法 ・OpenMPによる並列化 ・並列反復法に関する研究動向 ・Wisteria/BDEC-01(Odyssey)（情報基盤センター）による実習 - Finite-Volume Method (FVM) - Krylov Iterative Method - Preconditioning - Introduction to OpenMP - Reordering/Coloring Method - Parallel FVM Code using OpenMP - Recent Research Topics on Parallel Preconditioning Methods - Exercise using Wisteria/BDEC-01(Odyssey) of ITC

Materials are provided by the Instructor

Information about reference will be given in lectures

【前提とする知識・経験】 ・大学教養程度の物理，数学，基礎的数値解析（ガウスの消去法，SOR法等）の知識 ・UNIXの基本的な知識と利用経験（viまたはemacsを最低限使えること） ・CまたはFortranの基本的な知識と利用経験 【Prerequisites】 - Knowledge and experiences in fundamental methods for numerical analysis (e.g. Gaussian elimination, SOR) - Knowledge and experiences in UNIX (experiences in vi or emacs are minimum requirement) - Experiences in programming using Fortran or C
その他