RE: Incremental Callgraph built

[Patrick LAM <plam@sable.mcgill.ca>]

This is not exactly what you want, but it's close: they compute partial
call graphs, but they don't change the program during analysis.

Gagan Agrawal, Jinqian Li, and Qi Su.  Evaluating a Demand Driven
Technique for Call Graph Construction.

Says here that this was published in CC 2000 and 2002.
http://www.cis.ohio-state.edu/~agrawal/Projects/prog_analysis.html

I also have to point out this PLDI 2001 paper:

Frederic Vivien and Martin Rinard.  Incrementalized Pointer and Escape
Analysis.

http://www.cag.lcs.mit.edu/~rinard/paper/pldi01.pdf

pat

On Mon, 1 Dec 2003, Eric Bodden wrote:

> Hi again! Has really nobody done any work on that field before? I would be
> very grateful for any hints...
>
>
> > I am currently investigating the possibility of developing a
> > static approximation that helps to optimize weaving of a
> > cflow statement in AspectJ. For this purpose one would need a
> > partial call graph giving the transitive closure of all
> > outgoing calls of a certain method. So far no problem - apart
> > from efficiency. As I know the call graph analysis is
> > typically very slow, especially when it comes to VTA or even
> > more. So is there anybody who ever considered or even tried
> > to implement kind of an incremental approach for building
> > call graphs, meaning that e.g. if only one class is changed,
> > only the edges touching this very class are recalculated?
> > Or are there even problems that I did not think about which
> > might make that impossible?
>
>
> Cheers,
> Eric
>
>