[PATCH] PlaceSafepoints: use IRBuilder helpers

[Ramkumar Ramachandra]

In http://reviews.llvm.org/D7518#120876, @reames wrote:

> The relocate part of this change LGTM w/one change:  The old code took the name of the replaced call for the result.  This is useful for keeping the IR readable.  Please add a test for this as well.


It's not what you'd naively expect, because the `replaceAllUsesWith` is called after the new instruction is inserted into the tree: therefore, you get `%foo1` where `%foo` was the name of the original call. Would you like me to retain (and test) this bogus expectation, or should I retain the empty string, and attempt the name splicing in another patch (doing it the way I did it in the early CGP patch)?


http://reviews.llvm.org/D7518

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