[Soot-list] Performance problem (and fix?)

[Ismail Badawi]

That code was brought over from abc (
https://github.com/Sable/soot/commit/d10e3bc00bca1d59e90aa4d7ccc572f86befc529).
I had a quick look through the abc history and the commit for that revision
is just:

r574 | ganesh | 2004-06-21 10:56:43 -0400 (Mon, 21 Jun 2004) | 2 lines
Changed paths:
   M /trunk/aop/abc/src/abc/soot/util/Validate.java
   M /trunk/aop/abc/src/abc/weaving/matching/MethodCallShadowMatch.java

work around soot bug

Ismail

On Thu, Mar 14, 2013 at 9:47 AM, Michael Pradel <michael at binaervarianz.de>wrote:

> Hi,
>
> I found a performance problem and a maybe-fix for it, which speeds up my
> analysis by 5x. The problem is that the active hierarchy is repeatedly
> computed when calling Scene.getActiveHierarchy(). The Scene caches the
> hierarchy to avoid this redundant computation, but unfortunately Body
> contains the following statement (line 701):
>
>         // FIXME: Work around for bug in soot
>         Scene.v().releaseActiveHierarchy();
>
> I.e., the cached hierarchy is deleted for each Body that is validated
> (I'm using the -validate option). Is this really necessary? The
> hierarchy shouldn't change while validating method bodies. As far as I
> can tell from running my own analyses, removing the statement doesn't
> change the behavior (except for avoiding the redundant computation).
>
> Does anyone remember the reason for the FIXME (it's from October 2004)?
> Can it be removed safely? Getting a 5x speedup with such a simple change
> would be great.
>
> Thanks,
> Michael
> _______________________________________________
> 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/20130314/87ffbb73/attachment.html