[CUDA] Reject values for --cuda-gpu-arch that are not of the form /sm_\d+/.

[echristo]

LGTM.

-eric

[jlebar]

Add missing change to DiagnosticDriverKinds.td.

[jlebar]

Oops, I added the new error to the generated file in my objdir, rather than in the source tree. It worked fine until I rebuilt...

Eric, please have a look at the change to DiagnosticDriverKinds.td.

[echristo]

Yep it's fine. Thanks!

-eric

include/clang/Driver/Action.h

 #ifndef LLVM_CLANG_DRIVER_ACTION_H
 #define LLVM_CLANG_DRIVER_ACTION_H
 
 #include "clang/Driver/Types.h"
 #include "clang/Driver/Util.h"
 #include "llvm/ADT/SmallVector.h"
 
 namespace llvm {
+
+class StringRef;
+
 namespace opt {
   class Arg;
 }
 }
 
 namespace clang {
 namespace driver {
 
public:
 
   static bool classof(const Action *A) {
     return A->getKind() == BindArchClass;
   }
 };
 
 class CudaDeviceAction : public Action {
   virtual void anchor();
-  /// GPU architecture to bind -- e.g 'sm_35'.
+  /// GPU architecture to bind.  Always of the form /sm_\d+/.
   const char *GpuArchName;
   /// True when action results are not consumed by the host action (e.g when
   /// -fsyntax-only or --cuda-device-only options are used).
   bool AtTopLevel;
 
 public:
   CudaDeviceAction(Action *Input, const char *ArchName, bool AtTopLevel);
 
   const char *getGpuArchName() const { return GpuArchName; }
   bool isAtTopLevel() const { return AtTopLevel; }
 
+  static bool IsValidGpuArchName(llvm::StringRef ArchName);
+
   static bool classof(const Action *A) {
     return A->getKind() == CudaDeviceClass;
   }
 };
 
 class CudaHostAction : public Action {
   virtual void anchor();
   ActionList DeviceActions;

lib/Driver/Action.cpp

 //===--- Action.cpp - Abstract compilation steps --------------------------===//
 //
 //                     The LLVM Compiler Infrastructure
 //
 // This file is distributed under the University of Illinois Open Source
 // License. See LICENSE.TXT for details.
 //
 //===----------------------------------------------------------------------===//
 
 #include "clang/Driver/Action.h"
 #include "llvm/Support/ErrorHandling.h"
+#include "llvm/Support/Regex.h"
 #include <cassert>
 using namespace clang::driver;
 using namespace llvm::opt;
 
 Action::~Action() {}
 
 const char *Action::getClassName(ActionClass AC) {
   switch (AC) {

 BindArchAction::BindArchAction(Action *Input, const char *_ArchName)
     : Action(BindArchClass, Input), ArchName(_ArchName) {}
 
 void CudaDeviceAction::anchor() {}
 
 CudaDeviceAction::CudaDeviceAction(Action *Input, const char *ArchName,
                                    bool AtTopLevel)
     : Action(CudaDeviceClass, Input), GpuArchName(ArchName),
-      AtTopLevel(AtTopLevel) {}
+      AtTopLevel(AtTopLevel) {
+  assert(IsValidGpuArchName(GpuArchName));
+}
+
+bool CudaDeviceAction::IsValidGpuArchName(llvm::StringRef ArchName) {
+  static llvm::Regex RE("^sm_[0-9]+$");
+  return RE.match(ArchName);
+}
 
 void CudaHostAction::anchor() {}
 
 CudaHostAction::CudaHostAction(Action *Input, const ActionList &DeviceActions)
     : Action(CudaHostClass, Input), DeviceActions(DeviceActions) {}
 
 void JobAction::anchor() {}
 

lib/Driver/Driver.cpp

static Action *buildCudaActions(Compilation &C, DerivedArgList &Args,
 
   // Collect all cuda_gpu_arch parameters, removing duplicates.
   SmallVector<const char *, 4> GpuArchList;
   llvm::StringSet<> GpuArchNames;
   for (Arg *A : Args) {
     if (!A->getOption().matches(options::OPT_cuda_gpu_arch_EQ))
       continue;
     A->claim();
-    if (GpuArchNames.insert(A->getValue()).second)
-      GpuArchList.push_back(A->getValue());
+
+    const auto& Arch = A->getValue();
+    if (!CudaDeviceAction::IsValidGpuArchName(Arch))
+      C.getDriver().Diag(clang::diag::err_drv_cuda_bad_gpu_arch) << Arch;
+    else if (GpuArchNames.insert(Arch).second)
+      GpuArchList.push_back(Arch);
   }
 
   // Default to sm_20 which is the lowest common denominator for supported GPUs.
   // sm_20 code should work correctly, if suboptimally, on all newer GPUs.
   if (GpuArchList.empty())
     GpuArchList.push_back("sm_20");
 
   // Replicate inputs for each GPU architecture.

test/Driver/cuda-bad-arch.cu

+// Checks errors generated by passing a bad value for --cuda-gpu-arch.
+// REQUIRES: clang-driver
+// REQUIRES: x86-registered-target
+// REQUIRES: nvptx-registered-target
+
+// RUN: %clang -### -target x86_64-linux-gnu --cuda-gpu-arch=compute_20 -c %s 2>&1 \
+// RUN: | FileCheck -check-prefix BAD %s
+// RUN: %clang -### -target x86_64-linux-gnu --cuda-gpu-arch=foo_20 -c %s 2>&1 \
+// RUN: | FileCheck -check-prefix BAD %s
+// RUN: %clang -### -target x86_64-linux-gnu --cuda-gpu-arch=sm20 -c %s 2>&1 \
+// RUN: | FileCheck -check-prefix BAD %s
+// RUN: %clang -### -target x86_64-linux-gnu --cuda-gpu-arch=sm_abc -c %s 2>&1 \
+// RUN: | FileCheck -check-prefix BAD %s
+// RUN: %clang -### -target x86_64-linux-gnu --cuda-gpu-arch=sm_20a -c %s 2>&1 \
+// RUN: | FileCheck -check-prefix BAD %s
+// RUN: %clang -### -target x86_64-linux-gnu --cuda-gpu-arch=sm_a20 -c %s 2>&1 \
+// RUN: | FileCheck -check-prefix BAD %s
+// RUN: %clang -### -target x86_64-linux-gnu --cuda-gpu-arch=ssm_20 -c %s 2>&1 \
+// RUN: | FileCheck -check-prefix BAD %s
+// RUN: %clang -### -target x86_64-linux-gnu --cuda-gpu-arch=sm_ -c %s 2>&1 \
+// RUN: | FileCheck -check-prefix BAD %s
+
+// BAD: error: Unsupported CUDA gpu architecture
+
+// RUN: %clang -### -target x86_64-linux-gnu --cuda-gpu-arch=sm_2 -c %s 2>&1 \
+// RUN: | FileCheck -check-prefix OK %s
+// RUN: %clang -### -target x86_64-linux-gnu --cuda-gpu-arch=sm_20 -c %s 2>&1 \
+// RUN: | FileCheck -check-prefix OK %s
+// RUN: %clang -### -target x86_64-linux-gnu --cuda-gpu-arch=sm_999 -c %s 2>&1 \
+// RUN: | FileCheck -check-prefix OK %s
+// RUN: %clang -### -target x86_64-linux-gnu -c %s 2>&1 \
+// RUN: | FileCheck -check-prefix OK %s
+
+// OK-NOT: error: Unsupported CUDA gpu architecture