[Soot-list] Must alias analysis
Ondrej Lhotak
olhotak at sable.mcgill.ca
Wed Sep 21 12:54:40 EDT 2005
On Tue, Sep 20, 2005 at 06:40:15PM -0700, Manav Mital wrote:
> Hi
>
> I am looking for an implementation of must-alias
> analysis. I think earlier someone mentioned that
> he/she has one such implementation by Ondrej. How can
> I have access to that or some other such
> implementation?
Attached to this e-mail is the must-alias analysis that I wrote
way back when for Lin and Navin. It was intended to be an example,
a starting point.
The following are the instructions that I sent to them with it back
then:
> For each statement, it gives you a UnionFind object giving the
> must-alias relationship between locals before or after the statement.
> On this object, you can test whether two locals must be aliased using:
>
> if( unionFind.lookup( local1 ) == unionFind.lookup( local2 ) ) {
> // local1 and local2 must be aliased
> }
>
> Also, by calling unionFind.findSet( local ), you get a List of all the
> locals that must be aliased to local (including local itself).
>
> You can also call unionFind.allSets(), which returns a List of Lists,
> such that each local appears in exactly one of the lists, and within
> each list, all the locals must be aliased.
I don't remember much about it anymore, but I'll try to answer
any questions about it...
Ondrej
-------------- next part --------------
A non-text attachment was scrubbed...
Name: mustalias.tar.gz
Type: application/octet-stream
Size: 5666 bytes
Desc: not available
Url : http://mailman.CS.McGill.CA/pipermail/soot-list/attachments/20050921/3ad7c9ac/mustalias.tar.obj
More information about the Soot-list
mailing list