Algorithms and Data Structures for HPC

Garant předmětu:

Lecturer:

Ivan Šimeček

Tutor:

Ivan Šimeček

Supervisor:

Department of Computer Systems

Synopsis:

The most computation intensive tasks (or memory complex tasks) are solved by large HPC systems.

Seven so called „dwarfs“ were identified as typical tasks computed by HPC systems.

In this subject these „dwarfs“ are described (including their variants).

Also typical algortihms and advanced data structures for solution will be discussed.

Requirements:

Baic knowledge of principles of parallel systems

Syllabus of lectures:

1) Overview of nowadays technologies I (OpenMP and MPI)

2) Overview of nowadays technologies II (OpenACC and CUDA)

3) Single-node optimization

4) Dense linear algebra: data stuctures, operations and optimizations

5) Sparse linear algebra: data stuctures, operations and optimizations

6) 7 dwarfs: Spectral methods, N-body methods, Monte Carlo methods

7) 7 dwarfs: Structured and unstructured grids

8) More dwarfs: Graph Traversal, etc.

9) HPC debugging and profiling tools

10) Performance limits given by architecture

11) Survey of scientific libraries for HPC

12) HPC benchmarks: High Performance LINPACK vs HPCG

13) Summary

Syllabus of tutorials:

Study Objective:

Study materials:

Krste Asanovic, Ras Bodik, Bryan Christopher Catanzaro,Joseph James Gebis, Parry Husbands, Kurt Keutzer, David A. Patterson, William Lester Plishker, John Shalf,Samuel Webb Williams, Katherine A. Yelick The Landscape of Parallel Computing Research: A View from Berkeley

Prof. Dr. Michael Bader: HPC - Algorithms and Applications

Note:

Further information:

No time-table has been prepared for this course

The course is a part of the following study plans:

• Informatics (doctoral) (compulsory elective course)
• Informatics (compulsory elective course)