[PATCH] [Polly] Model scalar dependences to avoid trivial statements

[Johannes Doerfert]

I addressed your comments and commited it with a extended test case in r228847.


================
Comment at: lib/Analysis/TempScopInfo.cpp:150
@@ -147,1 +149,3 @@
+  bool canSynthesizeInst = canSynthesize(Inst, LI, SE, R);
+  if (canSynthesizeInst && !Functions.empty())
     return false;
----------------
grosser wrote:
> jdoerfert wrote:
> > grosser wrote:
> > > I don't really understand why scalar dependences to uses outside of the scop region only need to be modelled for otherwise trivial basic blocks. Should we not model them in general?
> > Maybe,... I'll think about it.
> > 
> > [Later I wanted to model the write only for the last iteration of the domain if there is no other scalar dependence for that instruction]
> OK, I would be interested in the outcome of your thoughts. ;-)
> 
> [Why would you want to only model the last write? What would be the benefit?]
Your right, we __need__ to model all accesses as we would allow reordering otherwise.

http://reviews.llvm.org/D7473

EMAIL PREFERENCES
  http://reviews.llvm.org/settings/panel/emailpreferences/