[cfe-dev] Enhancing the DataFlowSolver in clang

Simone Pellegrini spellegrini at dps.uibk.ac.at
Thu Feb 11 23:40:11 PST 2010

Hi all,

I am currently trying to implement few simple dataflow analysis, namely 
dominators and reaching definitions with the purpose to actually 
understand how the clang's DataFlowSolver works.

I got stuck with the dominator tree: 
The theory it's quite simple, but when I tried to implement it on the 
solver I started to have some problems.

The issue is that I need to set an initial state for each of the CFG 
blocks, and more important I need to specify a different initial state 
for each of the blocks (actually only the root node needs to have a 
different initialization)

For what I understand by reverse engineering the code, the method void 
setTopValue(DominatorGraph::ValTy& V) should be used to set the initial 
state of a dataflow problem.

However, in the dominators analysis I need to initialize the bitvectors 
for each of the blocks with different values and in order to be able to 
do that I need a different signature for the method which would be 
something like:

void SetTopValue(DominatorGraph::ValTy& V, clang::CFGBlock const& B){ }

This modification requires only small changes in the DataFlowSolver and 
implementation of dominators gets simpler.

Is there any other way to have this kind of behaviour or it is really 
missing? If you all agree with this change I can provide the patch!

regards, S. Pellegrini

More information about the cfe-dev mailing list