Skip to main navigation Skip to search Skip to main content

Accelerator-Centric Computing Ecosystems

Course

URL study guide

https://studiegids.vu.nl/en/courses/2026-2027/XM_0171

Course Objective

With the end of Moore's law, the need to continue performance scaling has led to the emergence of heterogeneous computing system infrastructure. Alongside traditional multicore CPUs, we now see GPUs (Graphic Processing Units), FPGAs (Field Programmable Gate Arrays), and other domain-specific architectures (such as DPUs and Machine Learning accelerators) being commonly offered as accelerators to speed up parts of applications. Each of these devices has different performance, power requirements, and parallelism opportunities that must be understood to leverage the diversity of modern and future ICT infrastructure. This course aims to equip students with an in-depth understanding of accelerated computing systems. This involves discussing diverse computing architectures, including alternatives to traditional ("von Neumann") computer organization, and analyzing their different programming models and optimization techniques, developing a comprehensive understanding of the intricacies and opportunities of each presented approach. After completing this course, you will: Understand diverse computing architectures (e.g., GPUs, FPGAs, and emerging paradigms) and relate them to classical computer organization [Knowledge and understanding]Acquire a comprehensive knowledge of the parallel programming models that characterize different compute accelerators and optimization techniques for diverse workloads. [Knowledge and understanding]Develop the ability to design, analyze, and implement parallel algorithms for specific computing architectures [Apply Knowledge]Evaluate and make informed decisions about the selection and optimization of algorithms for diverse computing architectures, considering factors such as performance, energy efficiency, and scalability [Apply Knowledge] [Making judgements]Effectively communicate technical concepts related to advanced computing systems [Communication]Develop learning skills to stay informed about and critically analyze emerging technologies in advanced computing systems [Lifelong learning skills]

Course Content

The course will consist of weekly lectures, lab sessions, and practical programming assignments. We will start by discussing the challenges faced by the computing community, such as the end of Dennard scaling and Moore's law, as well as the emergence of new computational-hungry domains – e.g., AI. We will discuss how these challenges have led to the development and diffusion of heterogeneous computing systems. We will introduce general-purpose graphics accelerators (GPGPUs) using, for example, NVidia's programming model CUDA. We will present dataflow architectures (particularly FPGAs) as an alternative to classical von Neumann architectures. We will discuss their programming challenges and how to develop efficient FPGA programs using High-Level Synthesis tools. Finally, we will conclude by discussing novel/research computing architectures and the current research directions in this field.

Teaching Methods

2 Lectures per week, Practical coding group assignments with 1 lab session per week for training/QA (deliverables: code and report).

Method of Assessment

There are two components in this course: Exam (50% of the final grade) and Practical Assignments (50% of the final grade). Resit is possible only for the exam. Exam and practical assignments must be passed separately to pass this course. Practical assignments are done in small teams. Students are graded on the completion of the assignment and their understanding of the proposed solution (based on the oral discussion/interview and report/code submitted) [FRAUD] The course is made fraud-resistant by having specialized group projects with interviews to assess the knowledge of the students. Furthermore, the exam is taken individually in-person on campus. Please note that your grade will be registered as follows:If you did not submit any practical assignment and did not participate in the exam, the result will be registered as No Show (NS).If you did not meet the minimum grade requirements for the practical or the exam, your grade is NVD.Otherwise, your final grade is computed from the exam and practical component, rounded to half points except for 5.5.There is no resit opportunity for the practical assignments. Results for practical assignments and for the exam can not be transferred from previous years.

Literature

The course does not follow a specific textbook. Slides and further online reading material/links will be provided via Canvas.

Target Audience

MSc Computer Science, MSc Computer Security

Additional Information

Use of generative AI is permitted in a limited way, with disclosure In this course, generative AI (GenAI), such as ChatGPT or similar tools, may be used in a limited way, for example for editorial support, rephrasing, or structure. Substantive contributions from GenAI may only be used within the limits set by the lecturer. Any use of GenAI must always be explicitly disclosed, in accordance with the lecturer's instructions. The student remains fully responsible for the content, quality, and accuracy of the submitted work. Incorrect use of GenAI is considered fraud. If established, the submitted work or assessment result will be declared invalid, and the Examination Board may impose further measures.

Recommended background knowledge

Computer organization (CPU caches, instructions, multicores), knowledge of C/C++, and parallel programming. For practical assignments: Linux development environment
- gcc, make, gdb, ssh.
Academic year1/09/26 → 31/08/27
Course level6.00 EC

Language of Tuition

  • English

Study type

  • Master