Counting summaries in IFDS

Moderator: Designing code analyses for large software systems

tobias.hamann
Windoof-User
Windoof-User
Beiträge: 39
Registriert: 4. Okt 2010 15:58

Counting summaries in IFDS

Beitrag von tobias.hamann »

Hello,

We encountered the following questions while working on exercise sheet 3: What are distinct summaries in IFDS and how are they counted?

Specifically, how many distinct summaries are there on slide 47 of lecture 8? We believe that there are 3 distinct summaries:
1.) {0, (v,0)} -> {0, (v,0), (r,1)}
2.) {0, (v,1)} -> {0, (v,1), (r,2)}
3.) {0, (v,2)} -> {0, (v,2), (r,3)}
Are those the correct summaries? Is {0} -> {0} a distinct summary as well?

Thanks in advance,
Tobias
Tobias Hamann, M.Sc.
Modeling and Analysis of Information Systems
Department of Computer Science, TU Darmstadt
http://www.mais.informatik.tu-darmstadt.de

Benutzeravatar
ericbodden
Sonntagsinformatiker
Sonntagsinformatiker
Beiträge: 243
Registriert: 5. Apr 2010 19:06

Re: Counting summaries in IFDS

Beitrag von ericbodden »

Hi.
tobias.hamann hat geschrieben: Specifically, how many distinct summaries are there on slide 47 of lecture 8? We believe that there are 3 distinct summaries:
1.) {0, (v,0)} -> {0, (v,0), (r,1)}
2.) {0, (v,1)} -> {0, (v,1), (r,2)}
3.) {0, (v,2)} -> {0, (v,2), (r,3)}
Are those the correct summaries? Is {0} -> {0} a distinct summary as well?
That is a good question. In general, as long as you explain in detail how you counted any sensible result is ok with me.

I personally think, though, that it makes most sense to really count the number of path edges that end up being turned into summaries, i.e., all path edges from the method start node to exit nodes. For the above example, those would be six individual edges, not counting the edge {0} -> {0}, which is often represented implicitly. Ok?
-- Eric

tobias.hamann
Windoof-User
Windoof-User
Beiträge: 39
Registriert: 4. Okt 2010 15:58

Re: Counting summaries in IFDS

Beitrag von tobias.hamann »

Hi,

makes sense, thanks for the fast answer!
Tobias Hamann, M.Sc.
Modeling and Analysis of Information Systems
Department of Computer Science, TU Darmstadt
http://www.mais.informatik.tu-darmstadt.de

Antworten

Zurück zu „Designing code analyses for large software systems (DECA)“