Informações principais
Sobre o conteúdo
In this computer science course, you will learn advanced concepts underpinning the design of today’s multicore-based computers. Additionally, you will learn how design decisions affect energy efficiency and performance.
Overall, topics include fundamentals on exploiting parallelism among instructions such as out-of-order execution, branch prediction, exception handling and advanced concepts of memory systems including prefetching and cache coherency. These concepts are fundamental for future computer systems to maximize compute efficiency.
You will also engage with a community of learners with similar interests to share knowledge.
The course is derived from Chalmers’s advanced graduate course in computer architecture. Prospective students should have a foundation in basic computer design, as offered by, for example, in “Computer System Design - Improving Energy Efficiency and Performance.“
- The anatomy of a multicore-based computer
- Advanced design concepts used in future multicore-based computers
- How computer design affects how energy-efficiently and fast a program can execute on future computer systems
- How taught design concepts are used in modern multicore-based computer systems
Pré-requisito
This course addresses students/professionals with an undergraduate degree from a computer science/engineering program with basic knowledge in computer organization including familiarity with basic concepts of computer design, as well as pipelining and caches. Students lacking familiarity to these concepts can learn those from ChM007x.
Programa de estudos
Week 1: Instruction and thread-level parallelism and models for parallel and out-of-order execution (1 hour)
- Multicore systems: It is all about parallelism. Fundamentals.
- Microarchitecture fundamentals.
- Memory systems fundamentals.
- Performance fundamentals.
- Energy fundamentals.
- Putting it together.
Week 2: Speculative execution = OoO + branch prediction + Exception handling (1 hour)
- Baseline model and out-of-order completion.
- Tomasulo algorithm: step 1.
- Tomasulo algorithm: step 2.
- Branch prediction.
- Exception handling.
- Speculative execution.
Week 3: Advanced topics on cache design (1 hour)
- Fundamentals of memory design.
- The 3C model.
- Replacement algorithms.
- Prefetching algorithms.
- Memory systems (DRAM).
- Memory systems (Virtual memory and virtualization).
Week 4: Multicore design: multithreading and cache coherence (1 hour)
- Multicore design challenges.
- Multithreading fundamentals.
- Multithreading design alternatives (fine/course, SMT).
- Cache coherence concepts.
- Interconnection concepts.
- Putting it together.
Instrutores
Per Stenström
Professor, Computer Engineering
Chalmers University of Technology
Criador do conteúdo
Plataforma
EdX est une plateforme d'apprentissage en ligne (dite FLOT ou MOOC). Elle héberge et met gratuitement à disposition des cours en ligne de niveau universitaire à travers le monde entier. Elle mène également des recherches sur l'apprentissage en ligne et la façon dont les utilisateurs utilisent celle-ci. Elle est à but non lucratif et la plateforme utilise un logiciel open source.
EdX a été fondée par le Massachusetts Institute of Technology et par l'université Harvard en mai 2012. En 2014, environ 50 écoles, associations et organisations internationales offrent ou projettent d'offrir des cours sur EdX. En juillet 2014, elle avait plus de 2,5 millions d'utilisateurs suivant plus de 200 cours en ligne.
Les deux universités américaines qui financent la plateforme ont investi 60 millions USD dans son développement. La plateforme France Université Numérique utilise la technologie openedX, supportée par Google.
