THE UNIVERSITY of EDINBURGH

DEGREE REGULATIONS & PROGRAMMES OF STUDY 2024/2025

Timetable information in the Course Catalogue may be subject to change.

University Homepage
DRPS Homepage
DRPS Search
DRPS Contact
DRPS : Course Catalogue : School of Informatics : EPCC on-campus

Postgraduate Course: Accelerated Systems: Principles and Practice (EPCC11020)

Course Outline
SchoolSchool of Informatics CollegeCollege of Science and Engineering
Credit level (Normal year taken)SCQF Level 11 (Postgraduate) AvailabilityAvailable to all students
SCQF Credits10 ECTS Credits5
SummaryAccelerators (such as GPUs, FPGAs, matrix coprocessors, etc) have established themselves as a key technology in high-performance computing, particularly in the most powerful current systems, and machine learning training. This course will introduce the general architecture of these devices and how they are built into HPC systems. We will go in depth on the current main techniques for writing high-performance programs using widely deployed accelerators, including debugging, profiling and performance analysis from a single device up to an interconnected cluster. We will discuss techniques for writing and maintaining a program that offers portable performance as well as for adapting an unaccelerated application. We will survey emerging accelerator technologies and discuss the types of problems they may be most effective at tackling and provide some hands-on experience.
Course description Accelerators appear to be here to stay for HPC and AI workloads. This course is mainly a pragmatic course designed to give you a thorough grounding in their architecture and systems design, before progressing to in depth information on how to program these efficiently. We will concentrate on the most widely deployed architecture (GPUs) and programming models (for example CUDA, OpenMP, SYCL), looking at single device parallelism, dealing with shared memory and synchronisation with millions of threads. We will then introduce multi-device parallelism, particularly how to integrate with MPI but mentioning other vendor-specific techniques. We will discuss paradigms for performance portability and practical techniques for adapting an existing application to use accelerated systems. We will introduce less widely deployed but emerging accelerator technologies such as FPGA, matrix coprocessors and ML accelerators. A key aim for the course is for students to be able to make appropriate choices as to which technology and techniques they should adopt for problems they will encounter in the future.
Entry Requirements (not applicable to Visiting Students)
Pre-requisites Co-requisites
Prohibited Combinations Other requirements None
Information for Visiting Students
Pre-requisitesAbility to program in C++ and to make use of OpenMP and MPI, equivalent with material from EPCC11002 and EPCC11003.
Course Delivery Information
Academic year 2024/25, Available to all students (SV1) Quota:  80
Course Start Semester 2
Timetable Timetable
Learning and Teaching activities (Further Info) Total Hours: 100 ( Lecture Hours 20, Supervised Practical/Workshop/Studio Hours 10, Feedback/Feedforward Hours 1, Programme Level Learning and Teaching Hours 2, Directed Learning and Independent Learning Hours 67 )
Assessment (Further Info) Written Exam 0 %, Coursework 100 %, Practical Exam 0 %
Additional Information (Assessment) Coursework (100%): Coursework will consist of two class tests, one short coursework and one larger submission.
Feedback Provided via practical classes and on assessed work.
No Exam Information
Learning Outcomes
On completion of this course, the student will be able to:
  1. Describe the architecture and programming models of current and emerging accelerated systems used in high-performance computing.
  2. Analyse algorithms and data structures for suitability for accelerators.
  3. Implement specifications in robust programs which can efficiently use a variety of accelerated systems.
  4. Use tools, such as profilers and debuggers, to understand characteristics of accelerated programs and formulate improvements.
  5. Evaluate programming models for suitability for specific applications.
Reading List
None
Additional Information
Graduate Attributes and Skills Not entered
KeywordsGPUs,Accelerators,HPC,MPI,OpenMP
Contacts
Course organiserDr Rupert Nash
Tel: (0131) 6 517 112
Email: r.nash@epcc.ed.ac.uk
Course secretaryMr James Richards
Tel: 90131 6)51 3578
Email: J.Richards@epcc.ed.ac.uk
Navigation
Help & Information
Home
Introduction
Glossary
Search DPTs and Courses
Regulations
Regulations
Degree Programmes
Introduction
Browse DPTs
Courses
Introduction
Humanities and Social Science
Science and Engineering
Medicine and Veterinary Medicine
Other Information
Combined Course Timetable
Prospectuses
Important Information