diff --git a/include/llvm/LTO/LTOCodeGenerator.h b/include/llvm/LTO/LTOCodeGenerator.h
index 4f78d27b10d..91b034cd905 100644
--- a/include/llvm/LTO/LTOCodeGenerator.h
+++ b/include/llvm/LTO/LTOCodeGenerator.h
@@ -201,6 +201,7 @@ private:
 
   LLVMContext &Context;
   std::unique_ptr<Module> MergedModule;
+  bool MainModuleHasSummary = false;
   std::unique_ptr<Linker> TheLinker;
   std::unique_ptr<TargetMachine> TargetMach;
   bool EmitDwarfDebugInfo = false;
diff --git a/lib/LTO/LTOCodeGenerator.cpp b/lib/LTO/LTOCodeGenerator.cpp
index 01304641598..8497114f64f 100644
--- a/lib/LTO/LTOCodeGenerator.cpp
+++ b/lib/LTO/LTOCodeGenerator.cpp
@@ -18,6 +18,7 @@
 #include "llvm/ADT/Statistic.h"
 #include "llvm/ADT/StringExtras.h"
 #include "llvm/Analysis/Passes.h"
+#include "llvm/Analysis/ModuleSummaryAnalysis.h"
 #include "llvm/Analysis/TargetLibraryInfo.h"
 #include "llvm/Analysis/TargetTransformInfo.h"
 #include "llvm/Bitcode/ReaderWriter.h"
@@ -145,6 +146,7 @@ void LTOCodeGenerator::setModule(std::unique_ptr<LTOModule> Mod) {
 
   MergedModule = Mod->takeModule();
   TheLinker = make_unique<Linker>(*MergedModule);
+  MainModuleHasSummary = Mod->isThinLTO();
 
   const std::vector<const char*> &Undefs = Mod->getAsmUndefinedRefs();
   for (int I = 0, E = Undefs.size(); I != E; ++I)
@@ -209,8 +211,13 @@ bool LTOCodeGenerator::writeMergedModules(const char *Path) {
     return false;
   }
 
+  std::unique_ptr<ModuleSummaryIndex> Index;
+  if (MainModuleHasSummary)
+    Index = ModuleSummaryIndexBuilder(MergedModule.get()).takeIndex();
+
   // write bitcode to it
-  WriteBitcodeToFile(MergedModule.get(), Out.os(), ShouldEmbedUselists);
+  WriteBitcodeToFile(MergedModule.get(), Out.os(), ShouldEmbedUselists,
+                     Index.get());
   Out.os().close();
 
   if (Out.os().has_error()) {
diff --git a/test/LTO/X86/set-merged-thin.ll b/test/LTO/X86/set-merged-thin.ll
new file mode 100644
index 00000000000..be2f97ab4d9
--- /dev/null
+++ b/test/LTO/X86/set-merged-thin.ll
@@ -0,0 +1,37 @@
+; RUN: opt  -module-summary  < %s > %t1
+; RUN: llvm-lto -save-merged-module -exported-symbol=_main -set-merged-module -o %t2 %t1
+; RUN: llvm-bcanalyzer -dump %t2.merged.bc | FileCheck %s
+; Verify that the module includes the ThinLTO summary
+; CHECK: <PERMODULE
+
+
+target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx10.10.0"
+
+define i32 @_Z3fooi(i32 %a) {
+entry:
+  %a.addr = alloca i32, align 4
+  store i32 %a, i32* %a.addr, align 4
+  %0 = load i32, i32* %a.addr, align 4
+  %1 = load i32, i32* %a.addr, align 4
+  %call = call i32 @_Z4bar2i(i32 %1)
+  %add = add nsw i32 %0, %call
+  ret i32 %add
+}
+
+define i32 @_Z4bar2i(i32 %a) {
+entry:
+  %a.addr = alloca i32, align 4
+  store i32 %a, i32* %a.addr, align 4
+  %0 = load i32, i32* %a.addr, align 4
+  %mul = mul nsw i32 2, %0
+  ret i32 %mul
+}
+
+define i32 @main() {
+entry:
+  %retval = alloca i32, align 4
+  store i32 0, i32* %retval
+  %call = call i32 @_Z3fooi(i32 44)
+  ret i32 %call
+}