Commit Graph

157 Commits

Author SHA1 Message Date
Roman Levenstein
e6442dc9b2 [sil-devirtualizer] Improve devirtualization of class_method instructions
Use ClassHierarchyAnalysis to check if a given method is effectively final, i.e. cannot be overridden.
This replaces the old approach based on using  isKnownFinal, which was weaker because it didn't use the ClassHierarchyAnalysis.
2015-12-14 10:11:47 -08:00
Roman Levenstein
ed38ce409c [sil-devirtualizer] Add a helper function to compute all subclasses of a given class.
The new function uses the ClassHierarchyAnalysis to do its job.
2015-12-14 10:11:47 -08:00
Roman Levenstein
fd0b383570 [sil-devirtualizer] Handle devirtualization of generic classes 2015-12-14 10:11:47 -08:00
Roman Levenstein
c909ff4cf2 [sil-devirtualizer] Improve getInstanceWithExactDynamicType
It is now able to cover some new cases where the instance is a parameter of a function. If it can be proven that the static type of instance cannot be derived, then it's static type is also a dynamic type of the instance.

The new functionality is not directly triggered in the current setup, because isKnownFinalMethod is currently invoked before it and subsumes this cases. But getInstanceWithExactDynamicType got this functionality anyways to be self-contained, so that it can be used in other contexts, where it is not invoked after isKnownFinalMethod.
2015-12-14 10:11:47 -08:00
Roman Levenstein
e05a72a1cb [sil-devirtualizer] Add a ClassHierarchyAnalysis parameter to tryDevirtualizeApply.
Not functionality changes yet. Subsequent patches will make use of this parameter.
2015-12-14 10:11:47 -08:00
Roman Levenstein
cbcd008ac4 [sil-devirtualizer] Rename isClassWithUnboundGenericParameters and extend it to work with any nominal types 2015-12-14 10:11:46 -08:00
Andrew Trick
739b0e9c56 Reorganize SILOptimizer directories for better discoverability.
(libraries now)

It has been generally agreed that we need to do this reorg, and now
seems like the perfect time. Some major pass reorganization is in the
works.

This does not have to be the final word on the matter. The consensus
among those working on the code is that it's much better than what we
had and a better starting point for future bike shedding.

Note that the previous organization was designed to allow separate
analysis and optimization libraries. It turns out this is an
artificial distinction and not an important goal.
2015-12-11 15:14:23 -08:00