Index: include/llvm/Transforms/Utils/Evaluator.h
===================================================================
--- include/llvm/Transforms/Utils/Evaluator.h
+++ include/llvm/Transforms/Utils/Evaluator.h
@@ -73,6 +73,10 @@
     ValueStack.back()[V] = C;
   }
 
+  Function *getCallee(Value *V);
+
+  Constant *substCallExprArgument(Value *CallExpr, Constant *RV);
+
   const DenseMap<Constant*, Constant*> &getMutatedMemory() const {
     return MutatedMemory;
   }
Index: lib/Transforms/Utils/Evaluator.cpp
===================================================================
--- lib/Transforms/Utils/Evaluator.cpp
+++ lib/Transforms/Utils/Evaluator.cpp
@@ -217,6 +217,30 @@
   return nullptr;  // don't know how to evaluate.
 }
 
+Function *Evaluator::getCallee(Value *V) {
+  auto *CE = dyn_cast<ConstantExpr>(V);
+  if (!CE)
+    return dyn_cast<Function>(getVal(V));
+
+  Constant *C =
+      CE->getOpcode() == Instruction::BitCast
+          ? ConstantFoldLoadThroughBitcast(CE, CE->getOperand(0)->getType(), DL)
+          : CE;
+  return dyn_cast<Function>(C);
+}
+
+/// Subsitute function call in bitcast expression to evaluated return value
+Constant *Evaluator::substCallExprArgument(Value *CallExpr, Constant *RV) {
+  auto *CE = dyn_cast<ConstantExpr>(CallExpr);
+  Constant *NewRV = nullptr;
+
+  if (CE && CE->getOpcode() == Instruction::BitCast)
+    if (auto *FT =
+            dyn_cast<FunctionType>(CE->getType()->getPointerElementType()))
+      NewRV = ConstantFoldLoadThroughBitcast(RV, FT->getReturnType(), DL);
+
+  return NewRV ? NewRV : RV;
+}
 /// Evaluate all instructions in block BB, returning true if successful, false
 /// if we can't evaluate it.  NewBB returns the next BB that control flows into,
 /// or null upon return.
@@ -458,7 +482,7 @@
       }
 
       // Resolve function pointers.
-      Function *Callee = dyn_cast<Function>(getVal(CS.getCalledValue()));
+      Function *Callee = getCallee(CS.getCalledValue());
       if (!Callee || Callee->isInterposable()) {
         DEBUG(dbgs() << "Can not resolve function pointer.\n");
         return false;  // Cannot resolve.
@@ -471,7 +495,7 @@
       if (Callee->isDeclaration()) {
         // If this is a function we can constant fold, do it.
         if (Constant *C = ConstantFoldCall(CS, Callee, Formals, TLI)) {
-          InstResult = C;
+          InstResult = substCallExprArgument(CS.getCalledValue(), C);
           DEBUG(dbgs() << "Constant folded function call. Result: " <<
                 *InstResult << "\n");
         } else {
@@ -492,7 +516,7 @@
           return false;
         }
         ValueStack.pop_back();
-        InstResult = RetVal;
+        InstResult = substCallExprArgument(CS.getCalledValue(), RetVal);
 
         if (InstResult) {
           DEBUG(dbgs() << "Successfully evaluated function. Result: "
Index: test/Transforms/GlobalOpt/evaluate-call.ll
===================================================================
--- test/Transforms/GlobalOpt/evaluate-call.ll
+++ test/Transforms/GlobalOpt/evaluate-call.ll
@@ -0,0 +1,47 @@
+; RUN: opt -globalopt -instcombine %s -S -o - | FileCheck %s
+
+; CHECK:      i32 @main()
+; CHECK-NEXT:   %1 = load i64, i64* inttoptr (i64 32 to i64*), align 32
+; CHECK-NEXT:   %2 = trunc i64 %1 to i32
+; CHECK-NEXT:   ret i32 %2
+; CHECK-NOT: _GLOBAL__sub_I_main
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-pc-linux-gnu"
+
+%struct.Bar = type { i64, i64 }
+%struct.S = type { %struct.Foo* }
+%struct.Foo = type { i64, i64 }
+%struct.Baz = type { i64, i64, %struct.Bar }
+
+@instance = internal local_unnamed_addr global %struct.S zeroinitializer, align 8
+@llvm.global_ctors = appending global [1 x { i32, void ()*, i8* }] [{ i32, void ()*, i8* } { i32 65535, void ()* @_GLOBAL__sub_I_main.cpp, i8* null }]
+@gBaz = available_externally dso_local local_unnamed_addr global %struct.Baz* null, align 8
+@gFoo = available_externally dso_local local_unnamed_addr global %struct.Bar* null, align 8
+
+; Function Attrs: norecurse nounwind readonly uwtable
+define dso_local i32 @main() local_unnamed_addr {
+  %1 = load %struct.Foo*, %struct.Foo** getelementptr inbounds (%struct.S, %struct.S* @instance, i64 0, i32 0), align 8
+  %2 = getelementptr inbounds %struct.Foo, %struct.Foo* %1, i64 0, i32 0
+  %3 = load i64, i64* %2, align 8
+  %4 = trunc i64 %3 to i32
+  ret i32 %4
+}
+
+; Function Attrs: uwtable
+define internal void @_GLOBAL__sub_I_main.cpp() section ".text.startup" {
+  %1 = tail call %struct.Foo* bitcast (%struct.Bar* ()* @_Z6getFoov to %struct.Foo* ()*)()
+  %2 = getelementptr inbounds %struct.Foo, %struct.Foo* %1, i64 1
+  store %struct.Foo* %2, %struct.Foo** getelementptr inbounds (%struct.S, %struct.S* @instance, i64 0, i32 0), align 8
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly uwtable
+define available_externally dso_local %struct.Bar* @_Z6getFoov() local_unnamed_addr {
+  %1 = load %struct.Bar*, %struct.Bar** @gFoo, align 8
+  %2 = icmp eq %struct.Bar* %1, null
+  %3 = load %struct.Baz*, %struct.Baz** @gBaz, align 8
+  %4 = getelementptr inbounds %struct.Baz, %struct.Baz* %3, i64 0, i32 2
+  %5 = select i1 %2, %struct.Bar* %4, %struct.Bar* %1
+  ret %struct.Bar* %5
+}