Analysis/Transf. of Seq. Programs [Pisa,UPM,ECRC]
WS Presentations Main Page
Next: Analysis of Concurrent
Up: Document: achievements
Previous: Parallelization [UPM]
- State of the art prior to project:
- Top-down, goal-dependent frameworks. Not full Prolog/CLP.
- Achievements:
- Analysis of full Prolog (/CLP) (including cut, var, etc.), with
varying degrees of precision. [Pisa, UPM, ECRC]
- Common source syntax, including annotations. [ECRC, UPM]
- Bottom-up, goal independent semantics. Control information
mapped to constraint system, using sequences of
constraints, each approximated by upward and downward
appr. [Pisa]
- Prolog implementation of the framework, with instantiation on
groundness analysis. [Pisa]
- Fully incremental analysis proposed and implemented. [UPM]
- Designed and implemented partial evaluator: [ECRC]
- can handle full parallel ECLPS
- correct (pure progs.)/fully automatic/always terminates
- transforms large programs quickly
- uses online transformation strategy with multiple
specialization based on unfold-fold rules
- compares well with other partial evaluators on benchmarks
- Future plans:
- More domains in bottom-up prototype; more concrete collecting semantics
(allow top-down/bottom-up comparison).
- Transfer to industry: global analysis
in commercial LP/CLP + investigate use as program
development tools.
Thu Jan 18 11:30:54 MET 1996 <herme@fi.upm.es>-<webmaster@clip.dia.fi.upm.es>