AArch64: add atomic support parameters to TargetInfo

This allows Clang to detect and deal wih __atomic_* operations properly on
AArch64. Previously we produced an error when encountering them at high
optimisation levels.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@175438 91177308-0d34-0410-b5e6-96231b3b80d8
diff --git a/lib/Basic/Targets.cpp b/lib/Basic/Targets.cpp
index 6f8dfe9..eaf2e7d 100644
--- a/lib/Basic/Targets.cpp
+++ b/lib/Basic/Targets.cpp
@@ -3167,6 +3167,10 @@
     WCharType = UnsignedInt;
     LongDoubleFormat = &llvm::APFloat::IEEEquad;
 
+    // AArch64 backend supports 64-bit operations at the moment. In principle
+    // 128-bit is possible if register-pairs are used.
+    MaxAtomicPromoteWidth = MaxAtomicInlineWidth = 64;
+
     TheCXXABI.set(TargetCXXABI::GenericAArch64);
   }
   virtual void getTargetDefines(const LangOptions &Opts,