Performance Evaluation of General-Purpose Computing on Graphics Processing Units

  • Performance Evaluation of General-Purpose Computing on Graphics Processing Units
Internships Summer 2017/18

Proposed Project Title:

  • Performance Evaluation of General-Purpose Computing on Graphics Processing Units

Principal Supervisor(s):
  • Hans Vandierendonck

Project Description:

General-Purpose computing on Graphics Processing Units (GPGPU) concerns the use of GPUs for a variety of computing needs, other than their primary purpose which is rendering images on the screen. GPUs are massively parallel processors and can perform 1000s of independent computations concurrently, yielding over 20 TFLOPs peak throughput (20 trillion floating-point operations per second). The design of GPUs was inspired by the nature of video rendering, where each pixel on screen can be rendered independently from other pixels. Their characteristics, however, have proven useful for performing a variety of compute-intensive tasks.

This project provides a gently introduction to GPUs. The main task consists of evaluating the performance of a set of benchmark programs (in particular, Rodinia) on a variety of GPUs. The goal of this effort is to evaluate whether these programs are performance-portable, i.e., whether high performance on one GPU implies that the same code will have high performance on another GPU. This is an important concern for developers of GPU-accelerated codes as optimising programs for GPU usage is challenging (yet rewarding) and time-consuming. It is not cost-effective to optimise codes for a variety of different GPUs. Through evaluating the performance of a benchmark set on a variety of GPUs and through varying parameters of these workloads, this project will shed light on the scale of the performance portability issue. Depending on the preferences of the student, (s)he may subsequently develop GPU code and aim to achieve very good performance portability, or very bad performance portability. Both are equally useful from a research perspective.

  • To learn about GPGPU and GPU programming
  • To learn quantitative performance evaluation of computing systems
  • To play around with high-end GPUs
  • To gain experience working with Linux-based systems and build environments

Interested students are strongly encouraged to contact the supervisor prior to selecting the project.

Academic Requirements:

The scheme is open to all EEECS Undergraduates (apart from students on the BIT degree pathway and students who are due to graduate this summer)

General Information:

Each internship will last between 6-8 weeks and will pay a weekly stipend of £200.

Accommodation and travel costs are not provided under this scheme.

Start date: TBC

Duration:  6-8 (Weeks)

Location: Computer Science Building

Further information available at:

Contact details:

Supervisor Name: Hans Vandierendonck

Queens University of Belfast
School of EEECS,
Computer Science Building,
18 Malone Road,

Tel: +44 (0)28 9097 4654

For further information on Research Area click on link below: