LFU:online

By completing this course, students will gain an understanding of GPU Computing, including both knowledge of GPU hardware and its performance characteristics as well as programming methodologies for GPUs. They will be able to identify whether a given algorithm is suitable for GPU-computing, and know how to build larger high-performance applications from individual parallel building blocks.

The 3 walls of hardware development
Task parallelism vs. data parallelism
Flynn 's Taxonomy
Key concepts of OpenCL
Modern accelerator hardware design
Understanding GPU memory
GPU threads, predication and occupancy
Matrix multiplication and matrix transpose
Parallel reduction and parallel prefix sum
Sorting networks and bitonic sorting
Wavefront parallelism

Basic concepts will be presented in a normal lecture format and immediately examined on practical code examples and their performance on real parallel hardware.

Projects and a written exam.

Solid C Programming Knowledge
Basic Parallel Programming