[PATCH] Semantic tests for memory invalidation at statepoints

[Sanjoy Das]

LGTM with the comment corrected.

================
Comment at: test/CodeGen/X86/statepoint-forward.ll:67
@@ +66,3 @@
+; highlight the fact that *any* gc pointer can be updated at a safepoint,
+; not just those live in the current frame
+define i1 @test_load_forward_nongc_heap(i32 addrspace(1)** %p) {
----------------
This comment is misleading -- AFAICT what we're checking here is that we can have pointers to GC-ed objects on non GC heaps.

http://reviews.llvm.org/D6491