[Soot-list] Soot the classpath library
Philippe Laporte
philippe.laporte at gatespacetelematics.com
Mon Jun 19 03:18:10 EDT 2006
On Fri, 2006-06-16 at 09:29 -0400, Eric Bodden wrote:
> > but why the gnu.* package could shrink ? Is there really no way to
> > shiink the java.*
> > javax.* and org.* ?
>
> Well, sometimes it can be that interprocedural optimizations can achive
> something. The Java bytecode verifier only rejects bytecode which are
> "obviously dead" (such bytecode is in fact invalid). Soot might
> sometimes be able to do a better job with identifying dead code but
> without looking across method boundaries this will almost never be the
> case IMHO.
>
> So my guess is just that the java.* etc. packages simply do not contain
> much (or maybe no at all) dead code. Any of the contained code *could
> potentially* used by people using the APIs. For *each fixed* given
> program you can eliminate all unused parts by an interprocedural dead
> code elimination.
>From what you write it doesn't sounds like SOOT includes a peep-hole
pattern optimizer.
Do you know of a good peep-hole optimizer, or any ideas in this area?
Thanks a lot,
--
Philippe Laporte
More information about the Soot-list
mailing list