The Custom Computing Research Group in Department of Computing conducts research in various areas of specialised computing platforms focusing on efficiency, performance and productivity, particularly:
Theory and practice of reconfigurable computing, including:
- compilation and optimisation
including
- meta-programming,
- in-circuit assertions and exceptions,
- domain-specific architecture diagnosis and tuning - run-time reconfigurable design
- reconfigurable systems and platforms, including heterogeneous cloud computing
Models, architectures, development methods and tools for:
Applications of custom-designed systems in areas such as:
- accelerating machine learning and machine learning for accelerator design
- bioinformatics
- climate and weather modelling
- computational finance, database analytics and social media, including the first paper on reconfigurable acceleration of financial simulation
- high energy physics
- media processing and graphics
- security and design validation
Our research also covers other applications such as sparse kernels, medical imaging, brain data processing, software-defined radio, seismic computation, database search, and ecological modelling.
We are responsible for several surveys on deep neural network approximation, reconfigurable computing, Gaussian random number generators, and low-power techniques for FPGAs, and a textbook on system-on-chip.
Our research has led to a Research Excellence Award from Imperial College, as well as many awards at international conferences, including ASAP (2008, 2013, 2016), ARC (2012), FPL (2004, 2007, 2008, 2010), ERSA (2004), FCCM (2012), FPT (2005, 2008), RCAR (2017), SAMOS (2008), and SPL (2008, 2009); and candidates for awards at ASAP (2019), FPGA (2020), FCCM (2020), and FPT (2018, 2019).