2016-a Winter
邁向 Petascale 高速計算
Introduction to Parallel Programming for Multicore/Manycore Clusters
臺大數學科學中心 科學計算冬季學校
2016年2月15日星期一至2月18日星期四 (9:00-17:00)
國立臺灣大學 天文數學館 301 室
最新訊息
- [2016/02/02] 講義已開放下載：2/15 and 2/16: http://www.kata-lab.itc.u-tokyo.ac.jp/class-matrNTU2016.htm，2/17 and 2/18: http://nkl.cc.u-tokyo.ac.jp/NTU2016/
- [2016/01/10] 線上註冊網址：http://goo.gl/forms/HjDBmF0ixA (報名已經額滿截止。本課程不收費用，但名額有限，獲得錄取學員將另行通知。)
- [2015/12/20] 臺灣大學地圖：PDF 檔，Google 版地圖
課程簡介
快速累積的大量數據，以及高速發展的超級電腦，是當今與未來的重要趨勢。具備高速計算能力，才能解決更大更複雜的問題，也能大幅提升進行尖端研究與開發產業應用的競爭力。在這個四天的密集課程，我們將簡介 MPI 與 OpenMP 的平行計算環境，說明如何在此平行環境求解稠密矩陣的特徵值問題，並將有限體積法以及大型線性系統疊代法平行化，求解三維 Poisson 方程。
此課程的另一個特點是使用東京大學的超級電腦上機實作。這台名為「Oakleaf-FX」的電腦，是全球最先進的超級電腦之一，它採用 Fujitsu PRIMEHPC FX10 大規模平行電腦架構，包含 4,800 個計算節點，每個節點配有 16核心的 SPARC64 IXfx 處理器，節點間以六維 Torus Fusion 網路串接，全機包含高達 76,800 核心，150 TB 的主記憶體，以及 1.13 PFLOPS 的尖峰運算能力。Oakleaf-FX 是目前 (2015/11) 在 TOP500 排名全球第四的 K Computer 的改良機型。(註：Oakleaf-FX 帳號須經申請，並依相關規定審核。)
這個短期課程提供一個非常難得的機會，可以接觸到目前全世界最尖端的高速平行計算環境。歡迎教師、碩博士生、大學部同學報名參加。本課程以英語講授，課程內容概要如下：
- Fundamental (2/15-16):
- Training of fundamental MPI and OpenMP are provided with parallelizations of dense matrix-vector multiplications and power method for eigenvalue problems.
- Advanced (2/17-18):
- MPI and OpenMP/MPI are applied to 3D Poisson equation solver by finite-volume method (FVM) with preconditioned conjugate gradient iterative method (PCG).
授課教授
- Professor Kengo Nakajima (中島 研吾 教授，東京大學 情報基盤中心 超級計算研究部門)
- Professor Takahiro Katagiri (片桐 孝洋 教授，東京大學 情報基盤中心 超級計算研究部門)
主辦單位與聯絡人
主辦單位：臺大數學科學中心
協辦單位：東京大學情報基盤中心，臺灣大學數學系與應用數學科學研究所，台灣工業與應用數學學會 GPU與高效能計算活動學群
主持人：王偉仲教授 (臺灣大學數學系與應用數學科學研究所，wwang@ntu.edu.tw)
聯絡人：劉英貝小姐 (臺大數學科學中心，tassist4@tims.ntu.edu.tw，02-3366-9901)
1. Course Information
2016 TIMS Winter School for Scientific Computing
Introduction to Parallel Programming for Multicore/Manycore Clusters
Professor Kengo Nakajima and Professor Takahiro Katagiri
(Information Technology Center, The University of Tokyo, Tokyo, Japan)
Overview
In order to make full use of modern supercomputer systems with multicore/manycore architectures, hybrid parallel programming with message-passing and multithreading is essential. MPI for message-passing and OpenMP for multithreading are the most popular ways for parallel programming on multicore/manycore clusters.
This 4-day tutorial provides essential knowledge and experiences for parallel programming using MPI and OpenMP. Hands-on exercise by Fujitsu PRIMEHPC FX10 supercomputer at the University of Tokyo (Oakleaf-FX) is also given. (http://www.cc.u-tokyo.ac.jp/system/fx10/index-e.html).
First 2 days are focusing on training of fundamental MPI and OpenMP. The basic functions and usage of MPI and OpenMP are explained. Several trainings of parallelization by using sample programs based on fundamental numerical computations, such as matrix-matrix multiplication, are provided.
On the 3rd and 4th days, MPI and OpenMP/MPI are applied to 3D Poisson equation solver by finite-volume method (FVM) with preconditioned conjugate gradient iterative method (PCG). Detailed lectures on data structure for parallel FVM are also provided.
Prerequisites
- Experiences in Unix/Linux
- Experiences of programming (Fortran or C/C++)
- Fundamental numerical algorithms (Gaussian Elimination, LU Factorization, Jacobi/Gauss-Seidel/SOR Iterative Solvers)
- Experiences in SSH Public Key Authentication Method
Prerequisties
- Experiences in Unix/Linux
- Experiences in emacs or vi
- Experiences of programming (Fortran or C/C++)
- Fundamental numerical algorithms (Gaussian Elimination, LU Factorization, Jacobi/Gauss-Seidel/SOR Iterative Solvers)
- Experiences in SSH Public Key Authentication Method
Preparation
- Windows: Cygwin with gcc/gfortran, SSH Client (e.g. PUTTY), ParaView
- MacOS, UNIX/Linux: ParaView
- Cygwin: https://www.cygwin.com/
- PuTTY: http://www.chiark.greenend.org.uk/~sgtatham/putty/
- ParaView: http://www.paraview.org/
Course Materials
- http://nkl.cc.u-tokyo.ac.jp/NTU2016/ (Available soon)
- http://www.kata-lab.itc.u-tokyo.ac.jp/class-matrNTU2016.htm (Available soon)
Schedule
February 15, 2016 (Monday)
Part 1. Fundamental I.
February 16, 2016 (Tuesday)
Part 2. Fundamental II.
09:10-10:00 Introduction
10:10-11:00 Overview of OpenMP
11:10-12:00 Functions of OpenMP
13:10-14:00 Training of OpenMP
14:10-15:00 Overview of MPI
15:10-16:00 How to use the FX10
16:10-17:00 Trainings of FX10
Homework 1
09:10-10:00 Functions of MPI Non-blocking and Persistent Communication
10:10-11:00 Parallelization of dense Matrix-Vector Multiplications (1/2)
11:10-12:00 Parallelization of dense Matrix-VectorMultiplications (2/2)
Homework 2
13:10-14:00 Parallelization of dense Power Method for eigenvalue problem (1/2)
14:10-15:00 Parallelization of dense Power Method for eigenvalue problem (2/2)
Homework 3
15:10-16:00 Parallelization of Fully Distributed
dense Matrix-Matrix Multiplication (1/2)
Homework 4
16:10-17:00 Parallelization of Fully Distributed
dense Matrix-Matrix Multiplication (2/2)
Homework 5
February 17, 2016 (Wednesday)
Part 3. Advanced I.
February 18, 2016 (Thursday)
Part 4. Advanced II.
09:10-10:00 FVM code (1/4)
10:10-11:00 FVM code (2/4)
11:10-12:00 FVM code (3/4)
13:10-14:00 FVM code (4/4) and sparse linear system solver
14:10-15:00 Parallel Data Structure (1/2)
15:10-16:00 Parallel Data Structure (2/2)
16:10-17:00 Parallel FVM (1/4)
Quiz
09:10-10:00 Parallel FVM (2/4)
10:10-11:00 Parallel FVM (3/4)
11:10-12:00 Parallel FVM (4/4)
13:10-14:00 OpenMP/MPI Hybrid (1/4)
14:10-15:00 OpenMP/MPI Hybrid (2/4)
15:10-16:00 OpenMP/MPI Hybrid (3/4)
16:10-17:00 OpenMP/MPI Hybrid (4/4)
Report (topics including strong scaling, weak scaling,
preconditioning, performance comparison, etc.)
2. Registration
2016 臺大數學科學中心科學計算冬季學校，線上註冊網址：http://goo.gl/forms/HjDBmF0ixA。
(報名截止日期： 2016/01/22 。本課程不收費用，但名額有限，獲得錄取學員將另行通知。)
3. Poster
