Index: clang/include/clang/Basic/Cuda.h
===================================================================
--- clang/include/clang/Basic/Cuda.h
+++ clang/include/clang/Basic/Cuda.h
@@ -25,7 +25,8 @@
   CUDA_91,
   CUDA_92,
   CUDA_100,
-  LATEST = CUDA_100,
+  CUDA_101,
+  LATEST = CUDA_101,
 };
 const char *CudaVersionToString(CudaVersion V);
 // Input is "Major.Minor"
@@ -107,6 +108,8 @@
 enum class CudaFeature {
   // CUDA-9.2+ uses a new API for launching kernels.
   CUDA_USES_NEW_LAUNCH,
+  // CUDA-10.1+ needs explicit end of GPU binary registration.
+  CUDA_USES_FATBIN_REGISTER_END,
 };
 
 bool CudaFeatureEnabled(llvm::VersionTuple, CudaFeature);
Index: clang/lib/Basic/Cuda.cpp
===================================================================
--- clang/lib/Basic/Cuda.cpp
+++ clang/lib/Basic/Cuda.cpp
@@ -25,6 +25,8 @@
     return "9.2";
   case CudaVersion::CUDA_100:
     return "10.0";
+  case CudaVersion::CUDA_101:
+    return "10.1";
   }
   llvm_unreachable("invalid enum");
 }
@@ -37,7 +39,8 @@
       .Case("9.0", CudaVersion::CUDA_90)
       .Case("9.1", CudaVersion::CUDA_91)
       .Case("9.2", CudaVersion::CUDA_92)
-      .Case("10.0", CudaVersion::CUDA_100);
+      .Case("10.0", CudaVersion::CUDA_100)
+      .Case("10.1", CudaVersion::CUDA_101);
 }
 
 const char *CudaArchToString(CudaArch A) {
@@ -352,6 +355,8 @@
     return CudaVersion::CUDA_92;
   case 100:
     return CudaVersion::CUDA_100;
+  case 101:
+    return CudaVersion::CUDA_101;
   default:
     return CudaVersion::UNKNOWN;
   }
@@ -365,6 +370,8 @@
   switch (Feature) {
   case CudaFeature::CUDA_USES_NEW_LAUNCH:
     return Version >= CudaVersion::CUDA_92;
+  case CudaFeature::CUDA_USES_FATBIN_REGISTER_END:
+    return Version >= CudaVersion::CUDA_101;
   }
   llvm_unreachable("Unknown CUDA feature.");
 }
Index: clang/lib/CodeGen/CGCUDANV.cpp
===================================================================
--- clang/lib/CodeGen/CGCUDANV.cpp
+++ clang/lib/CodeGen/CGCUDANV.cpp
@@ -445,6 +445,10 @@
   llvm::FunctionCallee RegisterFatbinFunc = CGM.CreateRuntimeFunction(
       llvm::FunctionType::get(VoidPtrPtrTy, VoidPtrTy, false),
       addUnderscoredPrefixToName("RegisterFatBinary"));
+  // void __cudaRegisterFatBinaryEnd(void **);
+  llvm::FunctionCallee RegisterFatbinEndFunc = CGM.CreateRuntimeFunction(
+      llvm::FunctionType::get(VoidTy, VoidPtrPtrTy, false),
+      "__cudaRegisterFatBinaryEnd");
   // struct { int magic, int version, void * gpu_binary, void * dont_care };
   llvm::StructType *FatbinWrapperTy =
       llvm::StructType::get(IntTy, IntTy, VoidPtrTy, VoidPtrTy);
@@ -616,6 +620,11 @@
     // Call __cuda_register_globals(GpuBinaryHandle);
     if (RegisterGlobalsFunc)
       CtorBuilder.CreateCall(RegisterGlobalsFunc, RegisterFatbinCall);
+
+    // CUDA version requires calling __cudaRegisterFatBinaryEnd(Handle);
+    if (CudaFeatureEnabled(CGM.getTarget().getSDKVersion(),
+                           CudaFeature::CUDA_USES_FATBIN_REGISTER_END))
+      CtorBuilder.CreateCall(RegisterFatbinEndFunc, RegisterFatbinCall);
   } else {
     // Generate a unique module ID.
     SmallString<64> ModuleID;
Index: clang/lib/Driver/ToolChains/Cuda.cpp
===================================================================
--- clang/lib/Driver/ToolChains/Cuda.cpp
+++ clang/lib/Driver/ToolChains/Cuda.cpp
@@ -60,6 +60,8 @@
     return CudaVersion::CUDA_92;
   if (Major == 10 && Minor == 0)
     return CudaVersion::CUDA_100;
+  if (Major == 10 && Minor == 1)
+    return CudaVersion::CUDA_101;
   return CudaVersion::UNKNOWN;
 }
 
Index: clang/lib/Headers/__clang_cuda_runtime_wrapper.h
===================================================================
--- clang/lib/Headers/__clang_cuda_runtime_wrapper.h
+++ clang/lib/Headers/__clang_cuda_runtime_wrapper.h
@@ -62,7 +62,7 @@
 #include "cuda.h"
 #if !defined(CUDA_VERSION)
 #error "cuda.h did not define CUDA_VERSION"
-#elif CUDA_VERSION < 7000 || CUDA_VERSION > 10000
+#elif CUDA_VERSION < 7000 || CUDA_VERSION > 10010
 #error "Unsupported CUDA version!"
 #endif