Re: Limiting whole-program analysis

[Ondrej Lhotak <olhotak@sable.mcgill.ca>]

On Thu, Jun 03, 2004 at 01:20:28PM -0500, Venkatesh Prasad Ranganath wrote:
> > Few, 5 to be exact, in the attached test program program: A.main,
> > X.<init>, java.lang.Object.<init>, java.lang.Object.finalize, X.toString
> >
> Shouldn't the class initializers be included?

They would be if those classes had any class initializers. If I add
statics so that those classes have class initializers, they do get
included in the call graph.

> So, Ondrej, if you have some time may be we can try to churn out some 
> numbers to compare SPARK and OFA?

What kinds of numbers are you interested in?

> >> It would useful to see which specific methods in implicit()
> >> cause your call graph to become unbearably large.
> >
> >
> > Perhaps, it's really quite easy to try. Maybe you could get more/better
> > quality info from just uncommenting those lines yourself and running
> > soot with "-f grimp -app A -w -p cg verbose:true;context:1cfa -p
> > cg.spark on,dump-solution:true,add-tags:true -p jap.pat on -p bb.lp
> > unsplit-original-locals:true" as I do.

BTW if you use -p cg context:1cfa, your analysis will become much slower
again. But you probably already know that.

Ondrej