On Wed, Aug 16, 2017 at 8:16 AM, Dibyendu Majumdar <mobile@xxxxxxxxxxxxxxx> wrote: > > Is it the problem that you are constructing SSA while linearizing - so > that you do not have the gotos resolved? Would it be a better design That is the paper is trying to construct SSA skipping the CFG stage. When you have goto, you can't really skip CFG any more. That is why it is fundamental hard to resolve for this paper. > to finish linear stream, and then run SSA conversion? The advantage of Yes, that is just the traditional way, the way that LLVM was doing. The paper is trying to claim having a simpler and efficient solution than LLVM. It is not true any more with considering of "goto" and irreducible graph input. > this would also be that you can reuse the SSA conversion phase later > on if some transformation of the linear stream introduced non-SSA > constructs. That is exactly the traditional way. You start to reason against the method using in this paper. :-) Chris -- To unsubscribe from this list: send the line "unsubscribe linux-sparse" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
