[PATCH] [SCCP] Calculate Null Pointer Comparisons

[Nicholas White]

Thanks all for the feedback. I've split the refactoring of the
PostDominatorPass into its own review - http://reviews.llvm.org/D9809
- as that part of the change is independent (and useful in general, I
think).

> You haven't updated any of the existing transforms to keep the post dominance info up to date.  As a result, the code as currently structured is really, really wrong.
I'll look into how dominance information invalidation works and revise
my patch accordingly.

> As a result, you'd need to justify the cost of the post-dom-tree computation.
> If you then want to propose using post-dominance information more widely, I'm open to that discussion.
That definitely makes sense. What about making the PostDominatorTree
pass an optional dependency of the SimplifyCFG pass (and so the only
runs when the user has decided to opt into that analysis. Maybe when a
critical mass of optimisations have optional dependencies on PDT it'd
be time to talk about adding it to the default O(3?) options.

Thanks -

Nick