About the Project
Modern processors spend a substantial fraction of their execution time waiting for data to arrive from memory. As applications continue to grow in scale and complexity, from cloud services to artificial intelligence workloads, efficient management of the memory hierarchy has become one of the defining challenges of computer architecture. Cache prefetching and replacement mechanisms play a critical role in addressing this challenge, directly affecting both performance and energy consumption.
Many of the most effective memory-management techniques rely on online adaptation. These techniques observe the behaviour of a running program, learn patterns in its memory accesses, and dynamically adjust hardware policies to improve performance. While highly effective, such approaches often require significant hardware resources and communication overheads, which can limit their adoption in commercial processors.
This PhD project will investigate a fundamentally new approach: using large language models (LLMs) to provide hardware with compact optimisation hints derived from program code. Rather than learning everything during execution, the processor will be guided by information generated ahead of time, reducing the amount of runtime monitoring and adaptation required. The central hypothesis is that AI-generated knowledge can complement traditional hardware learning mechanisms, enabling more sophisticated optimisation strategies at a fraction of the hardware cost.
The project will explore how LLMs can identify patterns in source code that are predictive of future memory behaviour, allowing instructions with similar memory-access characteristics to be grouped together. This information will then be embedded into program binaries and exploited by novel cache prefetching and replacement mechanisms. The resulting hardware-software techniques will be evaluated using state-of-the-art architectural simulators and modern benchmark workloads, with the aim of improving performance and energy efficiency while reducing implementation cost.
The research sits at the intersection of computer architecture, machine learning, and compiler technology. It offers an opportunity to work on one of the most exciting emerging directions in systems research: applying generative AI techniques to the design and optimisation of future computing systems. The outcomes of the project are expected to include new machine-learning techniques for reasoning about program behaviour, novel hardware-software co-design mechanisms, and open-source research artefacts that can support future work in the area.
The successful candidate will receive interdisciplinary training spanning computer architecture, machine learning, compiler systems, simulation-based evaluation, and research software development. They will join a vibrant research community and collaborate with researchers working across systems, architecture, compilers, and AI. Opportunities will be available to attend international summer schools and leading conferences in computer architecture and systems research.
Beyond cache management, the ideas developed in this project could influence a broad range of runtime optimisation mechanisms, including memory allocation, branch prediction, and resource management. More broadly, the project will explore a new hardware-software co-design paradigm in which AI-generated knowledge helps future processors make better optimisation decisions, contributing to faster and more energy-efficient computing systems.
This project is expected to start in September 2027.
Before you apply: We strongly recommend that you contact the supervisors for this project before you apply.
How to apply:
To be considered for this project you must complete a formal application through our online application portal. If you already have an applicant account this link will directly open an application for PhD School of Engineering Scholarships. If you don’t already have an applicant account, please follow the instructions here.
When applying, please specify the full title and supervisor/s of the project, details of your previous study, and names and contact details of two referees. You must also upload a Supporting Statement describing the motivation to apply to the project, your CV and transcripts of awarded and in-progress university qualifications. Please note late or incomplete applications will not be considered.
Equality, diversity and inclusion are fundamental to the success of The University of Manchester and central to all our activities. A diverse research community strengthens creativity, productivity and quality, while increasing the societal and economic impact of our work. We welcome applicants from all career paths, backgrounds and sections of the community, regardless of age, disability, ethnicity, gender, gender expression, sexual orientation or transgender status.
We welcome applications from candidates returning to study after a career break or experience in other roles. Flexible study arrangements may be available, including part-time study at 50%, 60% or 80%, subject to the requirements of the project and funder.
Eligibility: The standard academic entry requirement for this PhD is an upper second-class (2:1) honours degree in a discipline directly relevant to the PhD Computer Science or Computer Engineering (or international equivalent) OR any upper-second class (2:1) honours degree and a Master’s degree at merit in a discipline directly relevant to the PhD Computer Science or Computer Engineering (or international equivalent).
This project will remain open until filled.
If your application is submitted by 1st November 2026, you can expect a decision by 18th December 2026.
If your application is submitted by 15th January 2027, you can expect a decision by 30th March 2027.
Self or externally funded students can also be considered for this project.
Funding Notes
Funding for this project covers tuition fees, UKRI minimum annual stipend (currently £21,805/annum) and up to a £2.5k/annum research training support grant for the full duration of the 3.5-year programme.