Tag Line

PolyBoRi was created by man. There are two developers. And they have a plan.“

Saturday, July 20, 2013

GSOC 2013 project progress for weeks 0-4

I am Ioana Tamas, and this is a summary of the weekly updates of my GSOC 2013 project for Polybori.
Hopefully, the next updates will be posted in separate blog posts. My repository is: http://sourceforge.net/p/libzdd/code.

Weeks 0 &1:

I have been allocating a lot of time to learning how to work with big code bases etc., so the progress may not be very outstanding. Also, I figured that I will not be able to follow the proposed timeline, one example being the fact that updating the reference and memory allocation methods can't be done without updating the main classes and structures at the same time (which appear as later milestones).

Now, a summary of the code I have written until now:

1) I began the ZDD class, which will contain all the members and methods of the Cudd ZDD class, plus the ones inherited from Cudd classes that are of no use for Polybori (they were implemented in Cudd for handling more types of decision diagrams).

2) I started writing the structures needed for the completeness of the class definition, though I have not decided yet if I should rather have classes instead of structures. I also removed the structure members that are not useful for our purpose.

3) I have set up testing files, but I didn't include meaningful unit tests yet.

4) I added the Cudd code base in my directory, and I will gradually delete the files that are not useful until there is no Cudd left.

5) I made a namespace "libzdd" to avoid name coincidences. However, the way I did this may cause problems when I need Cudd files.

To finalize and summarize, the things I still have issues with would be:
- constructing the library using SCons, so that it accepts the boost unit test framework and takes into consideration the Cudd files
- correctly using the namespace
- choosing meaningful unit tests


During the past week, I started transforming the Cudd's structs into classes and further working on my main source and header file. After revising the timeline a little bit, I decided that I should take a break from what I started (so I temporarily commented a great part of my code) in order to work more on actually providing something that can be tested, using some temporary definitions. This part is still in progress.


The past week I temporarily kept only my Zero-Suppressed Decision Diagram class definition, using typedefs for the other relevant classes at the moment. This way, I was able to set up some boost tests and (more difficult than I expected) passing them. They test the things implemented until now, which are the constructors and some operators.

I also changed the structure of the repository, by making nice and clean header and source files for everything that I use currently use.

I also deleted a few things that I had introduced too early, but they will come back at the right time (they are saved somewhere in my computer for now).


This week, I made some significant progress in making the zdd_ZDD class (my project's main class) avoid using raw pointers, by wrapping the access to the Diagram Manager for all the constructors and operators.

In order to achieve this, I made use of boost shared pointers, and I introduced a new class.

I also introduced a unique table, with its own class (that will soon totally replace Cudd's), using boost tuples and std maps.