Research Group
I'm lucky enough to work with a range of fantastic people here in the Computer Lab on a variety of research themes. My current research group members are:
Alexandra Chadwick (PostDoc)
Alex is working on the ParaSol project, characterising applications to identify opportunities to extract fine-grained parallelism. Her research interests include novel computer architecture and microarchitecture, and compiler and programming language support for targeting to them. She previously worked on hardware and compiler design for the Loki processor, a many-core academic test-chip that was taped out in 2019.
Peter Zhang (PhD student)
Peter is working on efficient soft-error detection techniques for legacy binaries, especially through binary modification. His work aims to take advantage of parallelism to keep performance overheads low.
Karl Mose (PhD student)
Karl is investigating predictors in high-performance microprocessors, looking at improving prediction accuracy and reducing resource requirements for a range of different tasks. Robert Mullins and I are Karl's joint supervisors.
Yuxin Guo (PhD student)
Yuxin works on hardware support for fine-grained task-level parallelism, including transparently extracting parallelism and efficient microarchitectures. He previously worked on projects related to PIM (processing in memory) and is also interested in microarchitectures that reduce the memory wall for general purpose workloads.
Ruben Ticehurst-James (PhD student)
Ruben is investigating the application of machine learning and mathematical optimisation techniques in compilers. In particular, he is looking at restructuring existing codebases to improve automatic vectorisation for novel ISA extensions.
Timi Adeniran (PhD student)
Timi works on performance modelling and analysis for neural processing units (NPUs), exploring techniques to generate fast and accurate results. More generally, he is interested in design-space exploration methods for processing on small, low-power devices.
Yash Shah (PhD student)
Yash is working on hardware-efficient machine learning models, finding optimisations to the models that enable better efficiency, considering new models and developing hardware to specialise compute for parts of the models.
Previous Group Members
I'm always sorry when someone leaves the group, but pleased for them that they can spread their wings and explore pastures new! Here's a list of my not-forgotten research group members and where they headed:
Was Senior PostDoc, moved within the CHERI group. Theo worked on the CAPcelerate and Chrompartments projects, leading parts of the research into CHERI capabilities for future accelerators.
Was PostDoc, moved within the CHERI group. Paul worked on the CAPcelerate and Chrompartments projects, looking at how we could incorportate CHERI capabilities into future accelerators, as well as how CPUs integrating CHERI could mix with accelerators without support for capabilities.
Was PostDoc, went to Arm. Utpal worked on the ParaSol project, exploring new static analyses and program transformations to extract fine-grained parallelism.
Was PostDoc, went to Leibniz-Rechenzentrum. Akshay worked on the ParaSol project, developing compiler transformations to aid speculative parallelisation.
Was PostDoc, went to Riken. Minli worked on reliability for multicore processors, implementing efficient hardware schemes for error detection and correction in commodity systems.
Was PhD student, then PostDoc, went to Huawei. Márton's research looked into hardware support for fine-grained parallelism, including microarchitectures and speculation.
Was PhD student, went to Huawei. Mahwish investigated techniques for modifying binaries to improve their security properties.
Matthew Naylor
Was Senior PostDoc, went to the University Information Services. Matt worked on SIMT microarchitectures and the CHERI instruction set extensions.
Was PostDoc, went to Edinburgh University as a Lecturer. Jianyi worked on the CAPcelerate project, looking at ways of incorporating CHERI capabilities into accelerators or fitting them around existing accelerators.
Was PhD student. Aida worked on processing-in-memory for machine-learning algorithms.
Was PostDoc, went to SouthEast University, China as academic faculty. Hugo's research considered run-time hardware support for processor security and reliability.
Was PhD student, went to Monzo. Joe's research looked at techniques to pass semantic information from programmer to compiler and then static analysis that used this to break data dependences.
Was PostDoc, went to Arm. Jasmin worked on Janus binary paralleliser and static binary analysis to ensure code is vector-length agnostic.
Nandor Licker
Was PhD student, went to SiFive. Nandor studied low-level compiler analysis and optimisation, in particular for creating optimised binaries based on two or more source languages.
Was PostDoc, went to Utrecht University as an Assistant Professor. Tobias studied optimisations in Python, particularly ways in which parallelism could be exploited.
Was PhD student, then PostDoc, went to the University of Edinburgh as a Lecturer. Sam worked on data prefetching through indirect memory accesses, reliability and security techniques in both hardware and software.
Peng (Tom) Sun
Was PhD student, went to Arm. Tom worked on speculative vectorisation hardware.
Ruoyu (Kevin) Zhou
Was PhD student, then PostDoc, went to Huawei. Kevin developed the Janus automatic binary parallelisation tool.
Was PostDoc, went to Sheffield Hallam University as a lecturer. Hsi-Ming brought a theoretical background to the group and worked on static analysis.
Xiaochun (Dennis) Zhang
Was PostDoc, went to Arm. Dennis explored the limits of automatic parallelisation within compilers.
Martijn Bakker
Was Research Assistant, went to a finance company in London. Martijn worked on the POETS project.
Jyothish Soman
Was PhD Student, went to the DTG. Jyothish investigated methods for continuing processor execution despite permanent errors.
Negar Miralaei
Was PhD Student, tragically died shortly before finishing her dissertation. Negar worked on characterising processor ageing and its effect on applications.
Was PostDoc, went to Intel. Konstantina's research explored soft error detection using the compiler.
Niall Murphy
Was PhD Student, co-advised with Robert Mullins, went to Arm. Niall was investigating the use of speculation for DoAcross parallelism.
Was PostDoc, went to Intel. Vasileios developed schemes to improve the coverage of SLP vectorisation.
Dan Jones
Was PostDoc, went to the University of Oxford. Dan was looking at accelerating graph computation.
Was PostDoc, went to EPFL, now Intel. Amitabha developed a scheme for optimised cache coherence.
Georgios Tournavitis
Was PostDoc, went to Intel. George investigated cache coherence optimisations.