AtomicExpand: Don't crash on non-0 alloca

This now produces garbage on AMDGPU with a call to an nonexistent,
anonymous libcall but won't assert.

llvm-svn: 363022
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
index 64e710f..cdb406d 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
@@ -524,6 +524,7 @@
   setHasMultipleConditionRegisters(true);
 
   setMinCmpXchgSizeInBits(32);
+  setSupportsUnalignedAtomics(false);
 
   PredictableSelectIsExpensive = false;