General-Purpose Computing on GPU

Course guarantor:

Lecturer:

Tutor:

Supervisor:

Department of Computer Graphics and Interaction

Synopsis:

The course aims to introduce students to basic principles of General-Purpose Computing on Graphics Processing Units (GPGPU). The course gives an overview of the architecture and capabilities of modern graphics processing units (GPUs) and covers elementary concepts in parallel programming on GPUs. Students will gain programming skills with the CUDA (or OpenCL) technology and become familiar with basic parallel algorithms (e.g., parallel prefix scan/reduction) that are building blocks for designing and implementing efficient parallel algorithms.

Requirements:

Programming in C/C++, linear algebra.

Syllabus of lectures:

1. Introduction to general-purpose computing on GPU (architectures, languages, GPU versus CPU).

2. Introduction to CUDA architecture and basics of its programming.

3. CUDA - working with threads and memories, memory hierarchy.

4. CUDA - implementation of basic parallel patterns (reduction, prefix sum).

5. CUDA - programming for performance - optimizations.

6. CUDA - extensions - dynamic parallelism, cooperative groups, unified memory, graph.

7. Other high-level languages for GPGPU programming I - OpenCL (Open Computing Language)

8. Other high-level languages for GPGPU programming II - OpenGL Compute Shaders

Syllabus of tutorials:

1. Introduction - requirements, evaluation.

2. CUDA - compilation workflow, debugging, code examples. Specification of the individual student projects.

3. CUDA - solving simple examples.

4. CUDA - solution of given tasks I.

5. CUDA - solution of given tasks II.

6. CUDA - solution of given tasks III.

7. Individual work on projects - consultations I.

8. OpenCL - compilation workflow, debugging, code examples.

9. OpenCL - solution of given tasks IV.

10. Individual work on projects - consultations II.

11. Individual work on projects - consultations II.

12. Submitting of individual student projects.

13. Presentation of projects. Assessment.

Study Objective:

Study materials:

1. Wen-mei W. Hwu, David B. Kirk, Izzat El Hajj: Programming Massively Parallel Processors: A Hands-on Approach. 4th ed., Elsevier, 2022, https://doi.org/10.1016/C2020-0-02969-5.

2. Jason Sanders, Edward Kandrot: CUDA by Example: An Introduction to General-Purpose GPU Programming. 1st ed., Addison-Wesley Professional, 2010, ISBN 978-0131387683.

3. Aaftab Munshi, Benedict Gaster, Timothy G. Mattson, James Fung, Dan Ginsburg: OpenCL Programming Guide. Addison-Wesley Professional, 2011, ISBN 978-0321749642.

4. Gerassimos Barlas: Multicore and GPU Programming: An Integrated Approach. Morgan Kaufmann, 2014, ISBN 978-0124171374.

Note:

For more detailed information about the course, including the requirements for receiving credit, please visit the course web page: http://cent.felk.cvut.cz/courses/GPU.

Further information:

http://cent.felk.cvut.cz/courses/GPU

No time-table has been prepared for this course

The course is a part of the following study plans: