Innermost Fusion

Stratego -- Strategies for Program Transformation
An experimental fusion transformation for fusing the generic InnermostStrategy? with the rules it is instantiated with was implemented in StrategoRelease05. The transformation is described in the paper 'Fusing Logic and Control'.

The transformation is not as useful after StrategoRelease06, since InliningDefinitions? has been reduced. To get a better result out of the transformation, the inlining heuristics need to be improved. Probably it is a good idea to move the inliner from the front-end to the optimizer.

-- EelcoVisser - 09 Dec 2001

The fusion transformation is fully integrated in the compiler since StrategoRelease08 using its own inliner. The transformation is routinely used in the compiler itself and produces very fast simplifiers. The optimization works best for unconditional rules or rules with a condition not contributing to the right-hand side.

-- EelcoVisser - 17 Aug 2003


CategoryToDo? | ToDo | StrategoOptimization