Index: include/clang/Driver/Driver.h
===================================================================
--- include/clang/Driver/Driver.h
+++ include/clang/Driver/Driver.h
@@ -380,15 +380,14 @@
                        phases::ID Phase, std::shared_ptr<Action> Input) const;
 
   /// BuildJobsForAction - Construct the jobs to perform for the
-  /// action \p A.
-  void BuildJobsForAction(Compilation &C,
-                          const Action *A,
-                          const ToolChain *TC,
-                          const char *BoundArch,
-                          bool AtTopLevel,
-                          bool MultipleArchs,
-                          const char *LinkingOutput,
-                          InputInfo &Result) const;
+  /// action \p A.  Will only construct jobs for a given (Action,
+  /// ToolChain) pair once.
+  void BuildJobsForAction(
+      Compilation &C, const Action *A, const ToolChain *TC,
+      const char *BoundArch, bool AtTopLevel, bool MultipleArchs,
+      const char *LinkingOutput,
+      std::set<std::pair<const Action *, std::string>> &VisitedActions,
+      InputInfo &Result) const;
 
   /// Returns the default name for linked images (e.g., "a.out").
   const char *getDefaultImageName() const;
Index: lib/Driver/Driver.cpp
===================================================================
--- lib/Driver/Driver.cpp
+++ lib/Driver/Driver.cpp
@@ -1644,6 +1644,8 @@
       if (A->getOption().matches(options::OPT_arch))
         ArchNames.insert(A->getValue());
 
+  // Set of (Action, canonical ToolChain triple) pairs we've built jobs for.
+  std::set<std::pair<const Action *, std::string>> VisitedActions;
   for (const std::shared_ptr<Action>& A : C.getActions()) {
     // If we are linking an image for multiple archs then the linker wants
     // -arch_multiple and -final_output <final image name>. Unfortunately, this
@@ -1664,7 +1666,7 @@
                        /*BoundArch*/ nullptr,
                        /*AtTopLevel*/ true,
                        /*MultipleArchs*/ ArchNames.size() > 1,
-                       /*LinkingOutput*/ LinkingOutput, II);
+                       /*LinkingOutput*/ LinkingOutput, VisitedActions, II);
   }
 
   // If the user passed -Qunused-arguments or there were errors, don't warn
@@ -1792,20 +1794,27 @@
   return ToolForJob;
 }
 
-void Driver::BuildJobsForAction(Compilation &C, const Action *A,
-                                const ToolChain *TC, const char *BoundArch,
-                                bool AtTopLevel, bool MultipleArchs,
-                                const char *LinkingOutput,
-                                InputInfo &Result) const {
+void Driver::BuildJobsForAction(
+    Compilation &C, const Action *A, const ToolChain *TC, const char *BoundArch,
+    bool AtTopLevel, bool MultipleArchs, const char *LinkingOutput,
+    std::set<std::pair<const Action *, std::string>> &VisitedActions,
+    InputInfo &Result) const {
   llvm::PrettyStackTraceString CrashInfo("Building compilation jobs");
 
+  std::pair<const Action *, std::string> ActionTC = {
+      A, TC->getTriple().normalize()};
+  if (VisitedActions.find(ActionTC) != VisitedActions.end())
+    return;
+  VisitedActions.insert(ActionTC);
+
   InputInfoList CudaDeviceInputInfos;
   if (const CudaHostAction *CHA = dyn_cast<CudaHostAction>(A)) {
     InputInfo II;
     // Append outputs of device jobs to the input list.
     for (const std::shared_ptr<Action> &DA : CHA->getDeviceActions()) {
       BuildJobsForAction(C, DA.get(), TC, nullptr, AtTopLevel,
-                         /*MultipleArchs*/ false, LinkingOutput, II);
+                         /*MultipleArchs*/ false, LinkingOutput, VisitedActions,
+                         II);
       CudaDeviceInputInfos.push_back(II);
     }
     // Override current action with a real host compile action and continue
@@ -1839,7 +1848,7 @@
       TC = &C.getDefaultToolChain();
 
     BuildJobsForAction(C, BAA->begin()->get(), TC, ArchName, AtTopLevel,
-                       MultipleArchs, LinkingOutput, Result);
+                       MultipleArchs, LinkingOutput, VisitedActions, Result);
     return;
   }
 
@@ -1849,7 +1858,8 @@
     assert(CDA->getGpuArchName() && "No GPU name in device action.");
     BuildJobsForAction(C, CDA->begin()->get(), C.getCudaDeviceToolChain(),
                        CDA->getGpuArchName(), CDA->isAtTopLevel(),
-                       /*MultipleArchs*/ true, LinkingOutput, Result);
+                       /*MultipleArchs*/ true, LinkingOutput, VisitedActions,
+                       Result);
     return;
   }
 
@@ -1868,7 +1878,8 @@
     InputInfo II;
     for (const std::shared_ptr<Action> &DA : CollapsedCHA->getDeviceActions()) {
       BuildJobsForAction(C, DA.get(), TC, "", AtTopLevel,
-                         /*MultipleArchs*/ false, LinkingOutput, II);
+                         /*MultipleArchs*/ false, LinkingOutput, VisitedActions,
+                         II);
       CudaDeviceInputInfos.push_back(II);
     }
   }
@@ -1885,7 +1896,7 @@
 
     InputInfo II;
     BuildJobsForAction(C, Input.get(), TC, BoundArch, SubJobAtTopLevel,
-                       MultipleArchs, LinkingOutput, II);
+                       MultipleArchs, LinkingOutput, VisitedActions, II);
     InputInfos.push_back(II);
   }