[Soot-list] Any points-to analysis based on Shimple?
bodden at st.informatik.tu-darmstadt.de
Thu Sep 23 08:43:29 EDT 2010
> Q1: If simply re-implement existing algorithms in Jimple to Shimple,
> can we get flow-sensitive points-to sets(within a method) as Shimple is
> already in SSA form?
Yes, if you have a flow-sensitive pointer analysis over a
flow-sensitive representation like Shimple then you sort of get some
flow-sensitivity in the analysis for free. But of course,
reimplementing a whole pointer analysis on Shimple would be a lot of
> Q2: For your suggestions about 1) comparing pointers from different
> methods, 2) using local must alias analysis, can I regard them as refinement
> steps? The first refines intra-procedural points-to set to inter-procedural,
> and the second is to refine flow-insensitive points-to set to
> flow-sensitive. And regards local must alias as equivalent work to build SSA
> (variable numbering).
One can regard the LocalMustNotAliasAnalysis as a flow-sensitive,
intra-procedural refinement of the flow-insensitive inter-procedural
pointer analysis that Spark provides. The LocalMustAliasAnalysis
indeed uses value numbering and therefore functions quite similarly to
> I am interested in your "object representatives" approach, and I will
> read the paper and have a try.
More information about the Soot-list