Adaptive Parallel Optimising Compilation using Machine Learning
A PhD studentship is available to study Adaptive Parallel
Optimising Compilation Optimisation for Parallel Programs using
Machine Learning. The studentship will be held under the
supervision of
Prof. Michael O'Boyle, within the
Institute for Computing Systems Architecture, at the
School of Informatics, University of
Edinburgh, to begin in 2009, start date flexible. The studentship is funded under the umbrella of the
Centre for Numerical Algorithms and Intelligent Software.
Background
Multi-core processors are the most viable means to
delivering sustainable performance. However, this potential cannot be
realised unless the application has been well
parallelised. Unfortunately, efficient parallelisation of a sequential
program is a challenging and error-prone task. It is generally agreed
that manual code parallelisation by expert programmers results in the
most streamlined parallel implementation, but at the same time this is
the most costly and time-consuming approach. Parallelising compiler
technology, on the other hand, has the potential to greatly reduce
cost and time-to-market while ensuring formal correctness of the
resulting parallel code. Given that the underlying processor architecture
will change many times throughout the lifetime of the code, we would
like parallel programs that are performance future proof too.
Project
The project student will investigate new compiler directed approaches
to delivering performance in a multi-core environment where
the data input, concurrent workload and underlying architecture are
evolving. Probabilistic analysis can be used to determine program parallelism
which can then be mapped to available resources.
Central to this work will be the use of machine learning as a
technique to learn and adapt the parallel code to this changing
parallel landscape. The CaRD group at Edinburgh is internationally
leading in the use of machine learning for compiler and architecture
co-design and optimisation - this will form the backbone to this project.
Candidate Profile
Suitable candidates will have a strong first degree in Computer Science and a
strong interest in parallel programming, optimizing compilers or
machine learning. We are looking for the brightest minds to pursue
research in a cutting-edge arena.
Applying for the Studentship
Candidates are encouraged to
contact Michael O'Boyle to
informally discuss the project further. Formal application will be
through the School's normal PhD
application process.