JStar: A Declarative Language for a Parallel World
Speaker: Mark Utting, University of Waikato
When: 2009-04-07 10:00:00
Venue: 78-420
Host: Ian Hayes
Abstract:2005 marked the end of a era in computing. Computers are no
longer getting faster but they continue to grow exponentially in the
number of transistors. This poses a problem because it is now necessary
to write programs that can do many things at the same time (that is,
execute in a massively parallel fashion) in order to utilise these
additional transistors. The widely-used programming languages of today
were not designed for parallel computers, so parallel programming is
currently difficult and costly.
JStar is a new style of declarative parallel programming that aims to
make it easy to write high-performance parallel programs that can be
retargeted to a wide variety of computer architectures, including
cluster computers, many-core CPUs, GPUs, and potentially even FPGAs.
JStar separates the program logic from the parallelism details, avoids
premature commitment to data structures, and allows the compiler to
transform abstract programs into architecture-specific parallel
programs. Our long-term goal is to change the way the world programs,
by raising the abstraction level of parallel programs!
Over the last 10 years, we have developed the semantics of this
programming style, implemented several prototype languages, and
demonstrated the automatic selection of efficient data structures with
performance within a factor of two of hand-optimized programs in most
cases. This talk will give an overview of the programming style and the
underlying semantics, then focus on how the implicit parallelism in
JStar programs can be transformed into explicit parallelism for various
architectures.
Biography:Dr Mark Utting is an Associate Professor at the University of Waikato.
He is co-leader of the JStar project with Prof. John Cleary, and is also
the leader of the Community Z Tools (CZT) and the ModelJUnit open source
projects. He wrote the book "Practical Model-Based Testing: A Tools
Approach" in 2007 and continues to work with industry on developing
tools for model-based testing. In previous jobs, he worked in industry
as an analyst programmer in Sydney and as a post-doctoral researcher at
the Software Verification Research Centre at the University of
Queensland.
Type:
Contact: