PNaCl: Remove and disallow llvm.invariant.start/end intrinsics

These markers work in a similar way to llvm.lifetime.start/end, so we
should remove them for similar reasons: it's not very well defined how
one marker cancels out the effects of the other.

Arguably, invariant.start/end are less useful than lifetime.start/end.
They are ignored by the backend.  They are generated in fewer places:
invariant.start is generated by Clang (at -O1 or higher) when a const
global is initialised with a non-POD initialiser.  invariant.end is
apparently not generated at all.

Do the stripping in ReplacePtrsWithInts for consistency with the
existing lifetime.start/end stripping.

BUG=https://code.google.com/p/nativeclient/issues/detail?id=3443
TEST=PNaCl toolchain trybots

Review URL: https://codereview.chromium.org/15995004

1 files changed, 7 insertions, 2 deletions

diff --git a/lib/Transforms/NaCl/ReplacePtrsWithInts.cpp b/lib/Transforms/NaCl/ReplacePtrsWithInts.cpp
index 00d374dc7b..9be459d986 100644
--- a/lib/Transforms/NaCl/ReplacePtrsWithInts.cpp
+++ b/lib/Transforms/NaCl/ReplacePtrsWithInts.cpp
@@ -38,7 +38,8 @@
 //  * a bitcast of an InherentPtr.
 //
 // This pass currently strips out lifetime markers (that is, calls to
-// the llvm.lifetime.start/end intrinsics).
+// the llvm.lifetime.start/end intrinsics) and invariant markers
+// (calls to llvm.invariant.start/end).
 //
 //===----------------------------------------------------------------------===//
 
@@ -407,7 +408,9 @@ static void ConvertInstruction(DataLayout *DL, Type *IntPtrType,
   } else if (CallInst *Call = dyn_cast<CallInst>(Inst)) {
     if (IntrinsicInst *ICall = dyn_cast<IntrinsicInst>(Inst)) {
       if (ICall->getIntrinsicID() == Intrinsic::lifetime_start ||
-          ICall->getIntrinsicID() == Intrinsic::lifetime_end) {
+          ICall->getIntrinsicID() == Intrinsic::lifetime_end ||
+          ICall->getIntrinsicID() == Intrinsic::invariant_start ||
+          ICall->getIntrinsicID() == Intrinsic::invariant_end) {
         // Remove alloca lifetime markers for now.  This is because
         // the GVN pass can introduce lifetime markers taking PHI
         // nodes as arguments.  If ReplacePtrsWithInts converts the
@@ -416,6 +419,8 @@ static void ConvertInstruction(DataLayout *DL, Type *IntPtrType,
         // not safe in general.  So, until LLVM better defines the
         // semantics of lifetime markers, we drop them all.  See:
         // https://code.google.com/p/nativeclient/issues/detail?id=3443
+        // We do the same for invariant.start/end because they work in
+        // a similar way.
         Inst->eraseFromParent();
       } else {
         FC->convertInPlace(Inst);