TOAST
Project Members
Tom Crick,
Martin Brain,
Marina De Vos
and John Fitch.
Overview
TOAST is a practical, extensible
superoptimising system that uses the declarative
logic programming language AnsProlog and answer set programming
(ASP) as a modelling and computational framework. Superoptimisation is a different approach to optimising code, first proposed by
Henry Massalin, that seeks to generate optimal sequences from the outset, rather than
via incremental improvements.
We are currently able to superoptimise short acyclic code sequences for MIPS (R2000) and
SPARC (V7 and V8) architectures; we are targetting new architectures, including
ARM. We intend to release the
TOAST codebase under the
GNU General Public License (GPL) in the near future.
Benchmarks
NEW: Test programs from the paper accepted for LPNMR'09.
Timings and graphs for searching for and verifying code sequences
for a range of answer set solvers (July 2008).
Benchmarks from the ICLP 2006 paper.
Publications
-
Tom Crick, Martin Brain, Marina De Vos and John Fitch.
Generating Optimal Code using Answer Set Programming.
accepted for the 10th International Conference on Logic Programming and Nonmonotonic Reasoning (LPNMR'09), to appear in volume 5753 of LNCS (Springer), 2009.
-
Martin Brain, Tom Crick, Marina De Vos and John Fitch.
TOAST: Applying Answer Set Programming to Superoptimisation.
In Proceedings of the 22nd International Conference on Logic Programming (ICLP 2006), pages 270--284, volume 4079 of LNCS (Springer), 2006.
-
Martin Brain, Tom Crick, Marina De Vos and John Fitch.
An Application of Answer Set Programming: Superoptimisation - A Preliminary Report.
In Proceedings of the 11th International Workshop on Nonmonotonic Reasoning (NMR'06), pages 258--266.
IFI Technical Report Series (IFI-06-4), 2006.
