diff --git a/llvm/include/llvm/Frontend/OpenMP/OMPKinds.def b/llvm/include/llvm/Frontend/OpenMP/OMPKinds.def
--- a/llvm/include/llvm/Frontend/OpenMP/OMPKinds.def
+++ b/llvm/include/llvm/Frontend/OpenMP/OMPKinds.def
@@ -523,6 +523,11 @@
                                    EnumAttr(NoCapture))
                     : AttributeSet())
 
+__OMP_ATTRS_SET(NoCaptureAttrs,
+                OptimisticAttributes
+                    ? AttributeSet(EnumAttr(NoCapture))
+                    : AttributeSet(EnumAttr(NoCapture)))
+
 #if 0
 __OMP_ATTRS_SET(WriteOnlyPtrAttrs,
                 OptimisticAttributes
@@ -835,6 +840,10 @@
 __OMP_RTL_ATTRS(__kmpc_doacross_fini, BarrierAttrs, AttributeSet(),
                 ParamAttrs(ReadOnlyPtrAttrs))
 
+__OMP_RTL_ATTRS(__kmpc_alloc_shared, DefaultAttrs, ReturnPtrAttrs, {})
+__OMP_RTL_ATTRS(__kmpc_free_shared, AllocAttrs, AttributeSet(),
+                ParamAttrs(NoCaptureAttrs))
+
 __OMP_RTL_ATTRS(__kmpc_alloc, DefaultAttrs, ReturnPtrAttrs, {})
 __OMP_RTL_ATTRS(__kmpc_free, AllocAttrs, AttributeSet(), {})
 
diff --git a/llvm/lib/Transforms/IPO/AttributorAttributes.cpp b/llvm/lib/Transforms/IPO/AttributorAttributes.cpp
--- a/llvm/lib/Transforms/IPO/AttributorAttributes.cpp
+++ b/llvm/lib/Transforms/IPO/AttributorAttributes.cpp
@@ -4478,6 +4478,13 @@
   AANoCaptureCallSiteReturned(const IRPosition &IRP, Attributor &A)
       : AANoCaptureImpl(IRP, A) {}
 
+  /// See AbstractAttribute::initialize(...).
+  void initialize(Attributor &A) override {
+    const Function *F = getAnchorScope();
+    // Check what state the associated function can actually capture.
+    determineFunctionCaptureCapabilities(getIRPosition(), *F, *this);
+  }
+
   /// See AbstractAttribute::trackStatistics()
   void trackStatistics() const override {
     STATS_DECLTRACK_CSRET_ATTR(nocapture)
diff --git a/llvm/lib/Transforms/IPO/OpenMPOpt.cpp b/llvm/lib/Transforms/IPO/OpenMPOpt.cpp
--- a/llvm/lib/Transforms/IPO/OpenMPOpt.cpp
+++ b/llvm/lib/Transforms/IPO/OpenMPOpt.cpp
@@ -517,6 +517,7 @@
     // Recollect uses, in case Attributor deleted any.
     OMPInfoCache.recollectUses();
 
+    Changed |= removeGlobalization();
     Changed |= deleteParallelRegions();
     if (HideMemoryTransferLatency)
       Changed |= hideMemTransfersLatency();
@@ -981,6 +982,44 @@
     return Changed;
   }
 
+  bool removeGlobalization() {
+    auto &RFI = OMPInfoCache.RFIs[OMPRTL___kmpc_alloc_shared];
+
+    auto removeAllocCalls = [&](Use &U, Function &F) {
+      auto &FreeCall = OMPInfoCache.RFIs[OMPRTL___kmpc_free_shared];
+      CallBase *CB =
+          dyn_cast<CallBase>(OpenMPOpt::getCallIfRegularCall(U, &RFI));
+      if (!CB)
+        return false;
+
+      IRPosition AllocPos = IRPosition::callsite_returned(*CB);
+      if (A.lookupAAFor<AANoCapture>(AllocPos)->isKnownNoCapture()) {
+        Constant *AllocSize = dyn_cast<Constant>(CB->getArgOperand(0));
+        if (!AllocSize)
+          return false;
+
+        CallBase *FC = dyn_cast<CallBase>(CB->user_back());
+        if (!FC || FC->getCalledFunction() != FreeCall.Declaration)
+          return false;
+
+        const DataLayout &DL = M.getDataLayout();
+        Type *Int8Ty = Type::getInt8Ty(M.getContext());
+        AllocaInst *NewAlloca =
+            new AllocaInst(Int8Ty, DL.getAllocaAddrSpace(), AllocSize,
+                           CB->getName(), &F.front().front());
+        NewAlloca->setDebugLoc(CB->getDebugLoc());
+        FC->eraseFromParent();
+        CB->replaceAllUsesWith(NewAlloca);
+        CB->eraseFromParent();
+      }
+
+      return false;
+    };
+    RFI.foreachUse(SCC, removeAllocCalls);
+
+    return false;
+  }
+
   /// Try to delete parallel regions if possible.
   bool deleteParallelRegions() {
     const unsigned CallbackCalleeOperand = 2;
@@ -1600,6 +1639,24 @@
 
       GetterRFI.foreachUse(SCC, CreateAA);
     }
+
+    RuntimeFunction GlobalizationRuntimeIDs[] = {OMPRTL___kmpc_free_shared,
+                                                 OMPRTL___kmpc_alloc_shared};
+    for (const auto GlobalizationCallID : GlobalizationRuntimeIDs) {
+      auto &RFI = OMPInfoCache.RFIs[GlobalizationCallID];
+      auto CreateAA = [&](Use &U, Function &Decl) {
+        CallInst *CI = OpenMPOpt::getCallIfRegularCall(U, &RFI);
+        if (!CI)
+          return false;
+
+        auto &CB = cast<CallBase>(*CI);
+
+        IRPosition CBPos = IRPosition::callsite_returned(CB);
+        A.getOrCreateAAFor<AANoCapture>(CBPos);
+        return false;
+      };
+      RFI.foreachUse(SCC, CreateAA);
+    }
   }
 };
 
diff --git a/llvm/test/Transforms/OpenMP/remove_globalization.ll b/llvm/test/Transforms/OpenMP/remove_globalization.ll
new file mode 100644
--- /dev/null
+++ b/llvm/test/Transforms/OpenMP/remove_globalization.ll
@@ -0,0 +1,51 @@
+; RUN: opt -S -attributor -openmpopt < %s | FileCheck %s
+; RUN: opt -S -passes='attributor,cgscc(openmpopt)'  < %s | FileCheck %s
+target datalayout = "e-i64:64-i128:128-v16:16-v32:32-n16:32:64"
+target triple = "nvptx64"
+
+@S = external local_unnamed_addr global i8*
+
+; CHECK: {{.*}} = alloca i8, i64 4
+; CHECK-NOT: {{.*}} = call i8* @__kmpc_alloc_shared({{.*}})
+; CHECK-NOT: call void @__kmpc_free_shared({{.*}})
+define void @foo() #0 {
+entry:
+  %0 = call i8* @__kmpc_alloc_shared(i64 4)
+  %x_on_stack = bitcast i8* %0 to i32*
+  %1 = bitcast i32* %x_on_stack to i8*
+  call void @use(i8* %1)
+  call void @__kmpc_free_shared(i8* %0)
+  ret void
+}
+
+; CHECK-NOT: {{.*}} = alloca i8, i64 4
+; CHECK: {{.*}} = call i8* @__kmpc_alloc_shared({{.*}})
+; CHECK: call void @__kmpc_free_shared({{.*}})
+define void @bar() #0 {
+entry:
+  %0 = call i8* @__kmpc_alloc_shared(i64 4)
+  %x_on_stack = bitcast i8* %0 to i32*
+  %1 = bitcast i32* %x_on_stack to i8*
+  call void @share(i8* %1)
+  call void @__kmpc_free_shared(i8* %0)
+  ret void
+}
+
+define void @use(i8* %x) {
+entry:
+  %.addr = alloca i8*
+  store i8* %x, i8** %.addr
+  ret void
+}
+
+define void @share(i8* %x) {
+entry:
+  store i8* %x, i8** @S
+  ret void
+}
+
+; CHECK: declare i8* @__kmpc_alloc_shared(i64)
+declare i8* @__kmpc_alloc_shared(i64)
+
+; CHECK: declare void @__kmpc_free_shared(i8* nocapture)
+declare void @__kmpc_free_shared(i8*)