diff --git a/llvm/lib/Analysis/AliasAnalysis.cpp b/llvm/lib/Analysis/AliasAnalysis.cpp
--- a/llvm/lib/Analysis/AliasAnalysis.cpp
+++ b/llvm/lib/Analysis/AliasAnalysis.cpp
@@ -42,6 +42,7 @@
 #include "llvm/IR/BasicBlock.h"
 #include "llvm/IR/Instruction.h"
 #include "llvm/IR/Instructions.h"
+#include "llvm/IR/IntrinsicInst.h"
 #include "llvm/IR/Module.h"
 #include "llvm/IR/Type.h"
 #include "llvm/IR/Value.h"
@@ -226,6 +227,18 @@
                                     AAQueryInfo &AAQI) {
   ModRefInfo Result = ModRefInfo::ModRef;
 
+  // Unless convergent intrinsic carries explicit memory access attributes,
+  // assume that it may touch anything.
+  // TODO: This is overly conservative. We may need to query TTI whether
+  // particular intrinsic really affects the given memory location.
+  // E.g. some intrinsics may affect only specific address spaces.
+  if (Call->isConvergent() && isa<IntrinsicInst>(Call)) {
+    Function *F = Call->getCalledFunction();
+    if (!(F->onlyReadsMemory() || F->doesNotAccessMemory() ||
+          F->onlyAccessesArgMemory()))
+      return ModRefInfo::ModRef;
+  }
+
   for (const auto &AA : AAs) {
     Result = intersectModRef(Result, AA->getModRefInfo(Call, Loc, AAQI));
 
diff --git a/llvm/test/Analysis/GlobalsModRef/intrinsic_convergent.ll b/llvm/test/Analysis/GlobalsModRef/intrinsic_convergent.ll
new file mode 100644
--- /dev/null
+++ b/llvm/test/Analysis/GlobalsModRef/intrinsic_convergent.ll
@@ -0,0 +1,33 @@
+; RUN: opt -globals-aa -gvn -S < %s | FileCheck %s
+target datalayout = "e-i64:64-i128:128-v16:16-v32:32-n16:32:64"
+target triple = "nvptx64-nvidia-cuda"
+
+@s = internal local_unnamed_addr addrspace(3) global i32 undef, align 4
+
+; CHECK-LABEL: @bar_sync
+; CHECK: store
+; CHECK: tail call void @llvm.nvvm.bar.sync(i32 0)
+; CHECK: load
+define dso_local i32 @bar_sync(i32 %0) local_unnamed_addr {
+  store i32 %0, i32* addrspacecast (i32 addrspace(3)* @s to i32*), align 4
+  tail call void @llvm.nvvm.bar.sync(i32 0)
+  %2 = load i32, i32* addrspacecast (i32 addrspace(3)* @s to i32*), align 4
+  ret i32 %2
+}
+
+declare void @llvm.nvvm.bar.sync(i32) #0
+
+; CHECK-LABEL: @barrier0
+; CHECK: store
+; CHECK: tail call void @llvm.nvvm.barrier0()
+; CHECK: load
+define dso_local i32 @barrier0(i32 %0) local_unnamed_addr  {
+  store i32 %0, i32* addrspacecast (i32 addrspace(3)* @s to i32*), align 4
+  tail call void @llvm.nvvm.barrier0()
+  %2 = load i32, i32* addrspacecast (i32 addrspace(3)* @s to i32*), align 4
+  ret i32 %2
+}
+
+declare void @llvm.nvvm.barrier0() #0
+
+attributes #0 = { convergent nounwind }