Programming graphics cards with CUDA

In recent years, graphics cards have not only become increasingly powerful (currently >10 times faster than CPUs) but also easier to program. In this project, the participants learn to use NVIDIA's CUDA graphics cards to efficiently solve well-known problems in computer science.

Tasks
In the first weeks of the project, an introduction to CUDA is offered, at the end of which each group of participants must solve a simple task (e.g. matrix-vector multiplication) with the help of CUDA. Afterwards, each group will be given a more complex task that must be solved by the end of the semester. Possible task areas include:

• Sorting algorithms

• search algorithms

• Sparse matrix operations

• Traveling Salesman Problem

Further topics can be suggested by email.

Usefullinks
CUDA CMake Script
NVIDIA SDK Samples
NVIDIA CUDA Forums

Requirements
Basic knowledge of parallel programming and C/C++;

Experience with CUDA or graphics cards is not required.

Participants
Students of the Diploma, Bachelor or Master program in Computer Science from the 5th semester and other interested parties.

Project manager
Dipl. Inf. Jens Breitbart

Dates
By arrangement. Please get in touch with Dipl. Inf . Jens Breitbart.

Scope

• Bachelor 4 SWS (6 CP) or 8 SWS (12 CP)
• Master 4 SWS (8 CP)

Miscellaneous
For further information, please contact Prof. Dr. Claudia Fohry and Dipl. Inf . Jens Breitbart.

Other links
Project Bachelor in the HISLSF online course catalog
Project Master in the HISLSF online course catalog