[Soot-list] FlowAnalysis on large graphs

[Karthik]

Have there been any efforts to parallelize such analyses? I did find a few
papers about parallelization of such analyses.
Does Soot have any of those features?

Also, in general, I've observed that many analyses trade off precision in
order to reduce the explosion of memory consumption and time.
I'm not sure, but it may be interesting to explore other approaches such as
BDDBDDB, which may have different memory and time
characteristics, depending on the nature and complexity of the analysis.

Karthik.



On Thu, Mar 29, 2012 at 12:27 PM, Eric Bodden <eric.bodden at ec-spride.de>wrote:

> > I don't know about tomcat. But I believe that there's a problem with
> > "eclipse" as we usually analyze it in the context of DaCapo: my
> > understanding is that the default way of analyzing it only gets the
> > Eclipse loader, not the bulk of the code, which is inside jar files and
> > hence unanalyzed.
>
> That's true. In general, one should use TamiFlex when analyzing the
> DaCapo suite. That should take care of this.
>
> Karthik, analyzing large programs up to several hundreds of thousands
> of lines of code should certainly work given a few gigabytes of
> memory. If it does not, this probably means that you programmed a
> particularly memory-hungry or inefficient analysis. Getting analyses
> to scale to such programs is not a simple task and often requires very
> careful fine-tuning.
>
> Cheers,
> Eric
> _______________________________________________
> Soot-list mailing list
> Soot-list at sable.mcgill.ca
> http://mailman.cs.mcgill.ca/mailman/listinfo/soot-list
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.cs.mcgill.ca/pipermail/soot-list/attachments/20120330/a0a9358f/attachment.html