Module COMP52315: Performance Modelling, Vectorisation and GPU Programming

Department: Computer Science

COMP52315: Performance Modelling, Vectorisation and GPU Programming

Prerequisites

• PHYS52015 Introduction to Scientific and High Performance Computing

Corequisites

• None

Excluded Combination of Modules

• n/a

Aims

• Provide advanced knowledge and critical understanding of paradigms, fundamental ideas, algorithms and methods of program performance analysis, modelling and engineering
• Provide advanced knowledge and critical understanding of low-level code optimisation paradigms, and how they may be applied in real-world scenarios

Content

• Performance modelling and engineering
• Vectorisation and low-level code optimisation
• GPU programming

Learning Outcomes

• Understanding and critical reflection of advanced ideas and techniques behind the performance analysis of scientific computing and data analysis codes
• Understanding of modern computer architectures, and how to design and implement codes to perform optimally on this hardware

• Basic familiarity with state-of-the-art algorithms to solve large-scale and data intense challenges
• Competent and educated selection and statements on potential performance of specific codes plus skills to improve existing solutions

• Familiarity with advanced paradigms and modern algorithms underlying scientific computing for discrete systems, and their analysis
• Familiarity with programming paradigms for modern wide-vector CPUs, and GPUs

Modes of Teaching, Learning and Assessment and how these contribute to the learning outcomes of the module

• Teaching will be by lectures and workshops
• The lectures provide the means to give a concise, focused presentation of the subject matter of the module
• When appropriate, the lectures will also be supported by the distribution of written material, or by information and relevant links on Ultra
• Regular problem exercises and workshops will give students the chance to develop their theoretical understanding and problem solving skills
• Students will be able to obtain further help in their studies by approaching their lecturers, either after lectures or at other mutually convenient times
• Student performance will be summatively assessed through coursework
• The formative exercises provide opportunities for feedback, for students to gauge their progress and for staff to monitor progress throughout the duration of the module

Teaching Methods and Learning Hours

Summative Assessment

Formative Assessment:

n/a

■ Attendance at all activities marked with this symbol will be monitored. Students who fail to attend these activities, or to complete the summative or formative assessment specified above, will be subject to the procedures defined in the University's General Regulation V, and may be required to leave the University