Compiler based parallelisation and optimisation
2 PhD studentships are available to study
The studentships will be held under the supervision of
Prof. Michael O'Boyle and
Dr Bjoern Franke
Institute for Computing Systems Architecture, at the
School of Informatics, University of
Edinburgh, to begin in 2017, start date flexible. Both these studentships
are in collaboaration with ARM.
GPUs were initially designed as dedicated hardware accelerators for a
graphics pipeline. Since then, they have expanded their usage to a
much wider class of application. GPUs are widely recognised as having
the potential to deliver power efficient high performance. However,
achieving this potential is difficult due to rapidly evolving
architecture and the increasing diversity of applications.
This project will investigate compiler optimisation to improve the
performance portability of GPUs. It will explore the impact
optimisation has on existing OpenCL and graphics shader workloads as
well as emerging applications from Computer Vision.
The aim of this applied research is to investigate how to improve
parallelisation of sequential legacy applications using combined
static and dynamic analyses and parallel patterns (also known as
algorithmic skeletons). Existing parallelising compilers are built on
the same, fundamentally flawed principle: Reliance on static analysis
and focus on a single type of parallelism only.
Unfortunately, this does not work in practice and state-of-the-art
auto-parallelisers fail to detect parallelism or, even worse, result
in performance degradation. This project proposes to explore a
radically new approach to parallelisation, where we (a) exploit
well-known parallel patterns exposing all levels and shapes of
parallelism, and (b) use compiler-directed dynamic information to
overcome static analyis limitations.
The exact details of each project are flexible depending on the
candidate's interests and background.
This Studentship will cover all tuition fees and
provide a tax-free stipend at the EPSRC rate. Students receive funding
for a full EPSRC studentship for 3.5 years (currently around £68,648) plus
an additional 3.5k per annum and funded internship.
Suitable candidates will have a strong first degree in Computer
Science or Mathemeatics and a strong interest in parallel programming,
design space exploration, optimizing compilers or machine learning.
The exact topic of the PhD is flexible depending on the candidate's
interests. We are looking for the brightest minds to pursue research
in a cutting-edge arena. Candidate with Masters degrees or significant
insdustrial experience are also strongly encouraged to apply.
The successful candidate will benefit from
both academic and industrial research, have the opportunity to gain
real working experience in a world leading processor design centre at
ARM Ltd., Cambridge.
The anticipated start date is Sept 2015 but this is flexible
Institute for Computing Systems Architecture (ICSA)
The Institute for Computing Systems Architecture (ICSA) is one of
seven research institutes in the School of Informatics at the
University of Edinburgh. It was founded in 1998, following the
creation of Informatics. ICSA is primarily concerned with the
architecture and engineering of future computing systems. Within its
five research groups, ICSA covers topics which include: performance
and scalability, innovative algorithms, architectures, compilers,
languages, and protocols.
The CaRD group at Edinburgh is internationally
leading in for compiler and architecture
co-design and optimisation - this will form the backbone to this project.
Industrial Partner: ARM Ltd.
ARM Holdings is the world's leading semiconductor intellectual
property (IP) supplier and as such is at the heart of the development
of digital electronic products. Headquartered in Cambridge, UK, and
employing over 2,000 people, ARM has offices around the world,
including design centres in Taiwan, France, India, Sweden, and the US.
The ARM business model involves the designing and licensing of IP
rather than the manufacturing and selling of actual semiconductor
chips. ARM licenses IP to a network of Partners, which includes the
world's leading semiconductor and systems companies. These Partners
utilise ARM IP designs to create and manufacture system-on-chip
designs, paying ARM a license fee for the original IP and a royalty on
every chip or wafer produced. In addition to processor IP, ARM
provides a range of tools, physical and systems IP to enable optimised
system-on-chip designs. With the diversity of ARM IP and the broad
ecosystem of supporting silicon and software for ARM based solutions,
the world's leading Original Equipment Manufacturers (OEMs) use ARM
technology in a wide variety of applications ranging from mobile
handsets and digital set top boxes to car braking systems and network
routers. Today ARM technology is in use in 95% of smart phones, 80% of
digital cameras, and 35% of all electronic devices.
Applying for the Studentship Candidates are encouraged to
contact Michael O'Boyle
or Bjoern Franke to
informally discuss the project further. Formal application will be
through the School's normal PhD