Index: llvm/trunk/lib/IR/ValueSymbolTable.cpp
===================================================================
--- llvm/trunk/lib/IR/ValueSymbolTable.cpp
+++ llvm/trunk/lib/IR/ValueSymbolTable.cpp
@@ -13,7 +13,9 @@
 
 #include "llvm/IR/ValueSymbolTable.h"
 #include "llvm/ADT/SmallString.h"
+#include "llvm/ADT/Triple.h"
 #include "llvm/IR/GlobalValue.h"
+#include "llvm/IR/Module.h"
 #include "llvm/IR/Type.h"
 #include "llvm/IR/Value.h"
 #include "llvm/Support/Casting.h"
@@ -45,8 +47,17 @@
     // Trim any suffix off and append the next number.
     UniqueName.resize(BaseSize);
     raw_svector_ostream S(UniqueName);
-    if (isa<GlobalValue>(V))
-      S << ".";
+    if (auto *GV = dyn_cast<GlobalValue>(V)) {
+      // A dot is appended to mark it as clone during ABI demangling so that
+      // for example "_Z1fv" and "_Z1fv.1" both demangle to "f()", the second
+      // one being a clone.
+      // On NVPTX we cannot use a dot because PTX only allows [A-Za-z0-9_$] for
+      // identifiers. This breaks ABI demangling but at least ptxas accepts and
+      // compiles the program.
+      const Module *M = GV->getParent();
+      if (!(M && Triple(M->getTargetTriple()).isNVPTX()))
+        S << ".";
+    }
     S << ++LastUnique;
 
     // Try insert the vmap entry with this suffix.
Index: llvm/trunk/lib/Target/NVPTX/NVPTXAssignValidGlobalNames.cpp
===================================================================
--- llvm/trunk/lib/Target/NVPTX/NVPTXAssignValidGlobalNames.cpp
+++ llvm/trunk/lib/Target/NVPTX/NVPTXAssignValidGlobalNames.cpp
@@ -18,6 +18,7 @@
 //===----------------------------------------------------------------------===//
 
 #include "NVPTX.h"
+#include "llvm/IR/Function.h"
 #include "llvm/IR/GlobalVariable.h"
 #include "llvm/IR/LegacyPassManager.h"
 #include "llvm/IR/Module.h"
@@ -61,6 +62,11 @@
     }
   }
 
+  // Do the same for local functions.
+  for (Function &F : M.functions())
+    if (F.hasLocalLinkage())
+      F.setName(cleanUpName(F.getName()));
+
   return true;
 }
 
Index: llvm/trunk/test/CodeGen/NVPTX/symbol-naming.ll
===================================================================
--- llvm/trunk/test/CodeGen/NVPTX/symbol-naming.ll
+++ llvm/trunk/test/CodeGen/NVPTX/symbol-naming.ll
@@ -1,17 +1,17 @@
-; RUN: llc < %s -march=nvptx -mcpu=sm_20 | FileCheck %s --check-prefix=PTX32
-; RUN: llc < %s -march=nvptx64 -mcpu=sm_20 | FileCheck %s --check-prefix=PTX64
+; RUN: llc < %s -march=nvptx -mcpu=sm_20 | FileCheck %s
+; RUN: llc < %s -march=nvptx64 -mcpu=sm_20 | FileCheck %s
 
 ; Verify that the NVPTX target removes invalid symbol names prior to emitting
 ; PTX.
 
-; PTX32-NOT: .str
-; PTX64-NOT: .str
+; CHECK-NOT: .str
+; CHECK-NOT: .function.
 
-; PTX32-DAG: _$_str.1
-; PTX32-DAG: _$_str
+; CHECK-DAG: _$_str
+; CHECK-DAG: _$_str1
 
-; PTX64-DAG: _$_str.1
-; PTX64-DAG: _$_str
+; CHECK-DAG: _$_function_$_
+; CHECK-DAG: _$_function_$_2
 
 target datalayout = "e-i64:64-v16:16-v32:32-n16:32:64"
 target triple = "nvptx64-unknown-unknown"
@@ -22,10 +22,25 @@
 
 
 ; Function Attrs: nounwind
-define void @foo(i32 %a, float %b, i8 signext %c, i32 %e) {
+define internal void @.function.() {
 entry:
   %call = call i32 (i8*, ...) @printf(i8* getelementptr inbounds ([13 x i8], [13 x i8]* @.str, i32 0, i32 0))
   ret void
 }
 
+; Function Attrs: nounwind
+define internal void @_$_function_$_() {
+entry:
+  %call = call i32 (i8*, ...) @printf(i8* getelementptr inbounds ([13 x i8], [13 x i8]* @_$_str, i32 0, i32 0))
+  ret void
+}
+
+; Function Attrs: nounwind
+define void @global_function() {
+entry:
+  call void @.function.()
+  call void @_$_function_$_()
+  ret void
+}
+
 declare i32 @printf(i8*, ...)