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
@@ -58,6 +58,28 @@
 static constexpr auto TAG = "[" DEBUG_TYPE "]";
 #endif
 
+/// Apply \p CB to all uses of \p F. If \p LookThroughConstantExprUses is
+/// true, constant expression users are not given to \p CB but their uses are
+/// traversed transitively.
+template <typename CBTy>
+static void foreachUse(Function &F, CBTy CB,
+                       bool LookThroughConstantExprUses = true) {
+  SmallVector<Use *, 8> Worklist(make_pointer_range(F.uses()));
+
+  for (unsigned idx = 0; idx < Worklist.size(); ++idx) {
+    Use &U = *Worklist[idx];
+
+    // Allow use in constant bitcasts and simply look through them.
+    if (LookThroughConstantExprUses && isa<ConstantExpr>(U.getUser())) {
+      for (Use &CEU : cast<ConstantExpr>(U.getUser())->uses())
+        Worklist.push_back(&CEU);
+      continue;
+    }
+
+    CB(U);
+  }
+}
+
 /// Helper struct to store tracked ICV values at specif instructions.
 struct ICVValue {
   Instruction *Inst;
@@ -103,11 +125,12 @@
 /// Attributor runs.
 struct OMPInformationCache : public InformationCache {
   OMPInformationCache(Module &M, AnalysisGetter &AG,
-                      BumpPtrAllocator &Allocator, SetVector<Function *> *CGSCC,
-                      SmallPtrSetImpl<Function *> &ModuleSlice,
+                      BumpPtrAllocator &Allocator, SetVector<Function *> &CGSCC,
                       SmallPtrSetImpl<Kernel> &Kernels)
-      : InformationCache(M, AG, Allocator, CGSCC), ModuleSlice(ModuleSlice),
-        OMPBuilder(M), Kernels(Kernels) {
+      : InformationCache(M, AG, Allocator, &CGSCC), OMPBuilder(M),
+        Kernels(Kernels) {
+    initializeModuleSlice(CGSCC);
+
     OMPBuilder.initialize();
     initializeRuntimeFunctions();
     initializeInternalControlVars();
@@ -227,8 +250,45 @@
     DenseMap<Function *, std::shared_ptr<UseVector>> UsesMap;
   };
 
+  /// Initialize the ModuleSlice member based on \p SCC. ModuleSlices contains
+  /// (a subset of) all functions that we can look at during this SCC traversal.
+  /// This includes functions (transitively) called from the SCC and the
+  /// (transitive) callers of SCC functions. We also can look at a function if
+  /// there is a "reference edge", i.a., if the function somehow uses (!=calls)
+  /// a function in the SCC or a caller of a function in the SCC.
+  void initializeModuleSlice(SetVector<Function *> &SCC) {
+    ModuleSlice.insert(SCC.begin(), SCC.end());
+
+    SmallPtrSet<Function *, 16> Seen;
+    SmallVector<Function *, 16> Worklist(SCC.begin(), SCC.end());
+    while (!Worklist.empty()) {
+      Function *F = Worklist.pop_back_val();
+      ModuleSlice.insert(F);
+
+      for (Instruction &I : instructions(*F))
+        if (auto *CB = dyn_cast<CallBase>(&I))
+          if (Function *Callee = CB->getCalledFunction())
+            if (Seen.insert(Callee).second)
+              Worklist.push_back(Callee);
+    }
+
+    Seen.clear();
+    Worklist.append(SCC.begin(), SCC.end());
+    while (!Worklist.empty()) {
+      Function *F = Worklist.pop_back_val();
+      ModuleSlice.insert(F);
+
+      // Traverse all transitive uses.
+      foreachUse(*F, [&](Use &U) {
+        if (auto *UsrI = dyn_cast<Instruction>(U.getUser()))
+          if (Seen.insert(UsrI->getFunction()).second)
+            Worklist.push_back(UsrI->getFunction());
+      });
+    }
+  }
+
   /// The slice of the module we are allowed to look at.
-  SmallPtrSetImpl<Function *> &ModuleSlice;
+  SmallPtrSet<Function *, 8> ModuleSlice;
 
   /// An OpenMP-IR-Builder instance
   OpenMPIRBuilder OMPBuilder;
@@ -1001,12 +1061,9 @@
   if (DisableOpenMPOptimizations)
     return PreservedAnalyses::all();
 
-  SmallPtrSet<Function *, 16> ModuleSlice;
   SmallVector<Function *, 16> SCC;
-  for (LazyCallGraph::Node &N : C) {
+  for (LazyCallGraph::Node &N : C)
     SCC.push_back(&N.getFunction());
-    ModuleSlice.insert(SCC.back());
-  }
 
   if (SCC.empty())
     return PreservedAnalyses::all();
@@ -1026,8 +1083,7 @@
   SetVector<Function *> Functions(SCC.begin(), SCC.end());
   BumpPtrAllocator Allocator;
   OMPInformationCache InfoCache(*(Functions.back()->getParent()), AG, Allocator,
-                                /*CGSCC*/ &Functions, ModuleSlice,
-                                OMPInModule.getKernels());
+                                /*CGSCC*/ Functions, OMPInModule.getKernels());
 
   Attributor A(Functions, InfoCache, CGUpdater);
 
@@ -1065,14 +1121,11 @@
     if (DisableOpenMPOptimizations || skipSCC(CGSCC))
       return false;
 
-    SmallPtrSet<Function *, 16> ModuleSlice;
     SmallVector<Function *, 16> SCC;
     for (CallGraphNode *CGN : CGSCC)
       if (Function *Fn = CGN->getFunction())
-        if (!Fn->isDeclaration()) {
+        if (!Fn->isDeclaration())
           SCC.push_back(Fn);
-          ModuleSlice.insert(Fn);
-        }
 
     if (SCC.empty())
       return false;
@@ -1094,7 +1147,7 @@
     BumpPtrAllocator Allocator;
     OMPInformationCache InfoCache(
         *(Functions.back()->getParent()), AG, Allocator,
-        /*CGSCC*/ &Functions, ModuleSlice, OMPInModule.getKernels());
+        /*CGSCC*/ Functions, OMPInModule.getKernels());
 
     Attributor A(Functions, InfoCache, CGUpdater);