Arm Centre of Excellence

The SpiNNaker project has been 20 years in conception and 15 years in development. It has delivered the world’s largest neuromorphic computing platform incorporating over a million Arm processors that is used to support an open neuromorphic computing service under the auspices of the EU Flagship Human Brain Project. In addition to the million-core machine at Manchester, around 100 smaller SpiNNaker machines have been loaned and sold to research labs around the world.

The primary goal of the SpiNNaker project is to support brain science, with real time models of various brain regions already developed, but interest is n potential commercial applications are growing. The second generation SpiNNaker2 system, where each chip incorporates 153 ARM Cortex M4F cores, has been co-developed with TU Dresden, and SpiNNcloud Systems GmbH has been set up to commercialise the technology.

• Read our case study on SpiNNaker2: next level thinking

Nikolaos Kyparissas (PhD candidate) is investigating near-data acceleration of machine learning under the supervision of Professors Mikel Luján and Gavin Brown. In contrast to traditional computer systems, this project aims to exploit the increased capabilities of memory systems and understand the opportunities that near-memory processing offers for machine learning workloads. The Department of Computer Science Kilburn Award and Arm, as part of the Arm / RAEng Research Chair award, support Nick’s work.

Filippos Ventirozos (PhD candidate) and his supervisory team, Riza Batista-Navarro and Sarah Clinch, are researching textual data for end-user programming of IoT kitchen devices. This project brings together natural language processing, human-computer interaction and software engineering / modelling. Arm has been funding this project and has given Filippos an internship opportunity to work on machine learning applications running on the next-generation Corstone IoT chip. The Corstone silicon designs prioritise privacy and security -- critical aspects of IoT applications in domotic environments. This collaboration has enabled us to broaden our horizons on the future uses of IoT and embedded smart systems use cases.

Konstantinos Iordanou (PhD candidate) investigates the interaction of specialised compute kernels with processors in modern systems and their overall impact on system performance. In particular, he explores and understands the design process of hardware kernels for domain-specific applications that can co-exist efficiently with processors in a heterogeneous system.

Konstantinos is creating the design tools and methodologies, which will assist in simulating from the architectural and microarchitectural perspectives. During his internships at Arm Research in Cambridge, he focused on exploiting parallelism at loop-level mainly by studying the effects of various parallelisation strategies, constraints and data dependencies across loops. This work is funded by Arm and EPSRC.

One surprising discovery of the software era is the phenomenon of software rot. Over time, software becomes progressively more costly to change. This is especially true of software that is well used. Constant changes to the features offered by the software gradually degrade the design and add complexity until it becomes what is called “legacy software”: software that is too valuable to lose and too expensive to keep.

Periodically, such systems need to be rebuilt from scratch. But throwing away the old code means also throwing away much of the organisational expertise about the system. In this fellowship, Suzanne Embury will build tools to extract this expertise from legacy systems in the form of reusable test artefacts that will transfer to the new implementation. To achieve this, she will use novel information sources, including code change history and issue trackers, to reduce the costs and risks of system rebuild.