[Soot-list] Seeking proposals for Google Summer Of Code projects

[Richard Xiao]

Hi, Eric:

Sorry that I missed your reply.

If I were right, Andromeda resolves aliases first and performs IDE
analysis. Basically, Of course, we can feed a non-distributed problem to a
distributed solver, the only problem is how precise it is. Therefore, a
non-distributive summary database is useful that everybody who owns big
servers can shares her precise semi-product to other Soot users. It's no
doubt that everybody wants precise result but the only problem is the high
price to pay. Sharing can alleviate the pain significantly. If we can build
such an open infrastructure, I think that's very attractive to all Java
researchers.

Personally, I wrote a points-to analysis persistence tool for Soot. It's
also a pain to re-run Spark or other points-to engines again and again, so
I make a persistent and compact disk index for the points-to result.
Therefore, we only run Spark once and query it anytime and anywhere. This
saves a lot of time for recomputing and a lot of memory for querying
because of the index is quite small. This is not fully integrated with Soot
now, I will submit as a patch in future. But of course, I believe the
summary database is more useful than my tool.

Regards,
Xiao


On Mon, Feb 18, 2013 at 4:10 PM, Eric Bodden <eric.bodden at ec-spride.de>wrote:

> Hi Richard.
>
> > Routev's IDE summary is only applied to distributive data flow problems,
> > non-distributive data flow problems such as points-to analysis, heap
> > analysis, etc, cannot benefit from that summary. However, in my
> experience,
> > the most expensive part is always the interprocedural non-distributive
> data
> > flow analysis. Maybe adding a proposal for non-distributive summary is
> also
> > a good idea.
>
> I have worked quite a bit with IFDS and IDE now and so far I have yet
> to find a good use case for an analysis that would largely benefit
> from non-distributive flow functions. Even alias analysis seems to
> work quite well in IDE if you look at approaches such as Andromeda. Do
> you know any good examples of analyses that could benefit from such a
> more expressive framework?
>
> Eric
> _______________________________________________
> Soot-list mailing list
> Soot-list at sable.mcgill.ca
> http://mailman.cs.mcgill.ca/mailman/listinfo/soot-list
>



-- 
Richard Xiao Xiao
PhD Student @ CSE @ Hong Kong University of Science and Technology
www.cse.ust.hk/~richardxx
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.cs.mcgill.ca/pipermail/soot-list/attachments/20130222/d3941bbc/attachment.html