Professional Documents
Culture Documents
IBM Yachay Workshop Plan 2017 Previous
IBM Yachay Workshop Plan 2017 Previous
IBM Yachay Workshop Plan 2017 Previous
Prerequisites
1. Computing experiences with networked workstations;
2. Have written programs using languages such as FORTRAN or C or C++ or Java;
3. A major in mathematics or in any of the physical sciences or engineering.
Goals
To gain extensive practical experience in design and analysis of parallel algorithms for
scientific, engineering and commerce applications.
Workshop Expectations
I. Demonstrate knowledge of supercomputer architectures:
1. Node architecture, central processing units, and accelerators;
2. Distributed- and shared-memory;
3. Various interconnection networks;
4. Network architectures and performance;
5. IO subsystem.
II. Demonstrate mastery of software architecture and MPI:
1. Communication patterns and protocols;
2. Message passing interface (MPI);
3. Process creation and management;
4. Single-sided communications;
5. Collective operations.
III. Demonstrate understanding of performance measurements:
1. Speedup;
2. Parallel efficiency;
3. Overhead;
4. Load balance;
5. Scalability, strong or weak;
6. Amdahls law.
IV. Demonstrate mastery of parallel algorithms:
1. Algorithm classification according to complexities;
2. Problem decomposition;
3. Communication handling;
V. Demonstrate mastery of applications of parallel computers:
1. Linear algebra: matrix multiplication, solution of AX=b, eigenvalues and eigenvectors;
2. Basic calculus including high-dimensional integration;
3. Solutions of ODEs and systems including molecular dynamics;
4. Solutions of a few PDES including wave, diffusion, and Poison equations;
5. Fast Fourier transforms;
6. Markov chain Monte Carlo methods.
Workshop Overview
1. Architectures