Entry requirements

120 credits. High Performance and Parallel Computing 7.5 credits or High Performance Programming 10 credits. Proficiency in English equivalent to the Swedish upper secondary course English 6.

Learning outcomes

On completion of the course the student shall be able to:

• write efficient code directly adapted to a specific accelerator type to solve specific problems;
• use frameworks that offer abstractions of accelerator technology;
• identify and motivate strengths and weaknesses of different accelerator architectures for given problems;
• reason about data locality in heterogeneous systems;
• measure and improve the performance of self-written and framework-based accelerator code;

   

Content

Orientation about characteristics for, among other things, the architecture types CPU, GPU, TPU, wide vector instructions. Memory architectures and the relationship between the host CPU and other devices. CUDA programming and explicit vector operations. Use of TensorFlow to solve optimization problems and calculation of major mathematical expressions. Solutions in traditionally compiled and interpreted languages ​​to use accelerators. Profiling of the relevant solutions.

Instruction

Lectures, computer labs, assignments and projects.

Assessment

Oral presentation of laboratory work and projects. Written assignments and project report. Laboratory work and assignments (5 credits) and projects (2.5 credits).

If there are special reasons, the examiner may make an exception from the specified examination method and allow an individual student to be examined in another way. Special reasons may, for example, be information about special pedagogical support from the university's coordinator for students with disabilities.