[CUDA] Add utility functions for dealing with CUDA versions / architectures.

[jlebar]

Summary

Currently our handling of CUDA architectures is scattered all around
clang. This patch centralizes it.

A key advantage of this centralization is that you can now write a C++
switch on e.g. CudaArch and get a compile error if you don't handle one
of the enum values.

[tra]

LGTM.

lib/Basic/Cuda.cpp

8–19 ↗

(On Diff #62299)

We seem to do a lot of enum->string and string->enum mapping in this file.
Is there something comparable to Boost.bimap in standard c++ library or in LLVM?

lib/Driver/Driver.cpp

1026–1028 ↗

(On Diff #62299)

I think this could be collapsed to just CudaArchToString(CDA->getGpuArch()).
"(multiple archs)" is as informative as (and indistinguishable from) "unknown" here.

[jlebar]

lib/Basic/Cuda.cpp

8–19 ↗

(On Diff #62299)

Not to my knowledge.

lib/Driver/Driver.cpp

1026–1028 ↗

(On Diff #62299)

I'm not crazy about "unknown", since it *is* actually known. How about we just not output anything?

[tra]

lib/Driver/Driver.cpp

1026–1028 ↗

(On Diff #62299)

It's a debugging output so it would be good to accurately reflect our internal state.
In this case if we for some reason end up with CudaArch::UNKNOWN, I'd want to know that.
If we really use UNKNOWN to represent multiple archs, perhaps it needs an enum for multiple-archs.

[jlebar]

lib/Driver/Driver.cpp

1026–1028 ↗

(On Diff #62299)

We really do use UNKNOWN here to represent multiple architectures. It is used for the architecture of the Action corresponding to the call to fatbin.

I think adding an enum value for multiple-archs is going to be more harmful than useful, because it means that everywhere that we switch() on arch, we're going to have to handle (and assert) MULTIPLE_ARCHs.

[tra]

lib/Driver/Driver.cpp

1026–1028 ↗

(On Diff #62299)

OK. No output is fine with me.

[jlebar]

Address Art's review.

cfe/trunk/include/clang/Basic/Cuda.h

+//===--- Cuda.h - Utilities for compiling CUDA code  ------------*- C++ -*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_CLANG_BASIC_CUDA_H
+#define LLVM_CLANG_BASIC_CUDA_H
+
+namespace llvm {
+class StringRef;
+} // namespace llvm
+
+namespace clang {
+
+enum class CudaVersion {
+  UNKNOWN,
+  CUDA_70,
+  CUDA_75,
+  CUDA_80,
+};
+const char *CudaVersionToString(CudaVersion V);
+
+// No string -> CudaVersion conversion function because there's no canonical
+// spelling of the various CUDA versions.
+
+enum class CudaArch {
+  UNKNOWN,
+  SM_20,
+  SM_21,
+  SM_30,
+  SM_32,
+  SM_35,
+  SM_37,
+  SM_50,
+  SM_52,
+  SM_53,
+  SM_60,
+  SM_61,
+  SM_62,
+};
+const char *CudaArchToString(CudaArch A);
+
+// The input should have the form "sm_20".
+CudaArch StringToCudaArch(llvm::StringRef S);
+
+enum class CudaVirtualArch {
+  UNKNOWN,
+  COMPUTE_20,
+  COMPUTE_30,
+  COMPUTE_32,
+  COMPUTE_35,
+  COMPUTE_37,
+  COMPUTE_50,
+  COMPUTE_52,
+  COMPUTE_53,
+  COMPUTE_60,
+  COMPUTE_61,
+  COMPUTE_62,
+};
+const char *CudaVirtualArchToString(CudaVirtualArch A);
+
+// The input should have the form "compute_20".
+CudaVirtualArch StringToCudaVirtualArch(llvm::StringRef S);
+
+/// Get the compute_xx corresponding to an sm_yy.
+CudaVirtualArch VirtualArchForCudaArch(CudaArch A);
+
+/// Get the earliest CudaVersion that supports the given CudaArch.
+CudaVersion MinVersionForCudaArch(CudaArch A);
+
+} // namespace clang
+
+#endif

cfe/trunk/include/clang/Driver/Action.h

 //===--- Action.h - Abstract compilation steps ------------------*- C++ -*-===//
 //
 //                     The LLVM Compiler Infrastructure
 //
 // This file is distributed under the University of Illinois Open Source
 // License. See LICENSE.TXT for details.
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef LLVM_CLANG_DRIVER_ACTION_H
 #define LLVM_CLANG_DRIVER_ACTION_H
 
+#include "clang/Basic/Cuda.h"
 #include "clang/Driver/Types.h"
 #include "clang/Driver/Util.h"
 #include "llvm/ADT/SmallVector.h"
 
 namespace llvm {
 
 class StringRef;
 
public:
 
   static bool classof(const Action *A) {
     return A->getKind() == BindArchClass;
   }
 };
 
 class CudaDeviceAction : public Action {
   virtual void anchor();
-  /// GPU architecture to bind.  Always of the form /sm_\d+/ or null (when the
-  /// action applies to multiple architectures).
-  const char *GpuArchName;
+
+  const CudaArch GpuArch;
+
   /// 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);
+  CudaDeviceAction(Action *Input, CudaArch Arch, bool AtTopLevel);
 
-  const char *getGpuArchName() const { return GpuArchName; }
-
-  /// Gets the compute_XX that corresponds to getGpuArchName().  Returns null
-  /// when getGpuArchName() is null.
-  const char *getComputeArchName() const;
+  /// Get the CUDA GPU architecture to which this Action corresponds.  Returns
+  /// UNKNOWN if this Action corresponds to multiple architectures.
+  CudaArch getGpuArch() const { return GpuArch; }
 
   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;

cfe/trunk/lib/Basic/CMakeLists.txt

set_source_files_properties(Version.cpp
       PROPERTIES COMPILE_DEFINITIONS "SVN_REVISION=\"${SVN_REVISION}\"")
   endif()
 endif()
 
 add_clang_library(clangBasic
   Attributes.cpp
   Builtins.cpp
   CharInfo.cpp
+  Cuda.cpp
   Diagnostic.cpp
   DiagnosticIDs.cpp
   DiagnosticOptions.cpp
   FileManager.cpp
   FileSystemStatCache.cpp
   IdentifierTable.cpp
   LangOptions.cpp
   Module.cpp

cfe/trunk/lib/Basic/Cuda.cpp

+#include "clang/Basic/Cuda.h"
+
+#include "llvm/ADT/StringRef.h"
+#include "llvm/ADT/StringSwitch.h"
+
+namespace clang {
+
+const char *CudaVersionToString(CudaVersion V) {
+  switch (V) {
+  case CudaVersion::UNKNOWN:
+    return "unknown";
+  case CudaVersion::CUDA_70:
+    return "7.0";
+  case CudaVersion::CUDA_75:
+    return "7.5";
+  case CudaVersion::CUDA_80:
+    return "8.0";
+  }
+}
+
+const char *CudaArchToString(CudaArch A) {
+  switch (A) {
+  case CudaArch::UNKNOWN:
+    return "unknown";
+  case CudaArch::SM_20:
+    return "sm_20";
+  case CudaArch::SM_21:
+    return "sm_21";
+  case CudaArch::SM_30:
+    return "sm_30";
+  case CudaArch::SM_32:
+    return "sm_32";
+  case CudaArch::SM_35:
+    return "sm_35";
+  case CudaArch::SM_37:
+    return "sm_37";
+  case CudaArch::SM_50:
+    return "sm_50";
+  case CudaArch::SM_52:
+    return "sm_52";
+  case CudaArch::SM_53:
+    return "sm_53";
+  case CudaArch::SM_60:
+    return "sm_60";
+  case CudaArch::SM_61:
+    return "sm_61";
+  case CudaArch::SM_62:
+    return "sm_62";
+  }
+}
+
+CudaArch StringToCudaArch(llvm::StringRef S) {
+  return llvm::StringSwitch<CudaArch>(S)
+      .Case("sm_20", CudaArch::SM_20)
+      .Case("sm_21", CudaArch::SM_21)
+      .Case("sm_30", CudaArch::SM_30)
+      .Case("sm_32", CudaArch::SM_32)
+      .Case("sm_35", CudaArch::SM_35)
+      .Case("sm_37", CudaArch::SM_37)
+      .Case("sm_50", CudaArch::SM_50)
+      .Case("sm_52", CudaArch::SM_52)
+      .Case("sm_53", CudaArch::SM_53)
+      .Case("sm_60", CudaArch::SM_60)
+      .Case("sm_61", CudaArch::SM_61)
+      .Case("sm_62", CudaArch::SM_62)
+      .Default(CudaArch::UNKNOWN);
+}
+
+const char *CudaVirtualArchToString(CudaVirtualArch A) {
+  switch (A) {
+  case CudaVirtualArch::UNKNOWN:
+    return "unknown";
+  case CudaVirtualArch::COMPUTE_20:
+    return "compute_20";
+  case CudaVirtualArch::COMPUTE_30:
+    return "compute_30";
+  case CudaVirtualArch::COMPUTE_32:
+    return "compute_32";
+  case CudaVirtualArch::COMPUTE_35:
+    return "compute_35";
+  case CudaVirtualArch::COMPUTE_37:
+    return "compute_37";
+  case CudaVirtualArch::COMPUTE_50:
+    return "compute_50";
+  case CudaVirtualArch::COMPUTE_52:
+    return "compute_52";
+  case CudaVirtualArch::COMPUTE_53:
+    return "compute_53";
+  case CudaVirtualArch::COMPUTE_60:
+    return "compute_60";
+  case CudaVirtualArch::COMPUTE_61:
+    return "compute_61";
+  case CudaVirtualArch::COMPUTE_62:
+    return "compute_62";
+  }
+}
+
+CudaVirtualArch StringToCudaVirtualArch(llvm::StringRef S) {
+  return llvm::StringSwitch<CudaVirtualArch>(S)
+      .Case("compute_20", CudaVirtualArch::COMPUTE_20)
+      .Case("compute_30", CudaVirtualArch::COMPUTE_30)
+      .Case("compute_32", CudaVirtualArch::COMPUTE_32)
+      .Case("compute_35", CudaVirtualArch::COMPUTE_35)
+      .Case("compute_37", CudaVirtualArch::COMPUTE_37)
+      .Case("compute_50", CudaVirtualArch::COMPUTE_50)
+      .Case("compute_52", CudaVirtualArch::COMPUTE_52)
+      .Case("compute_53", CudaVirtualArch::COMPUTE_53)
+      .Case("compute_60", CudaVirtualArch::COMPUTE_60)
+      .Case("compute_61", CudaVirtualArch::COMPUTE_61)
+      .Case("compute_62", CudaVirtualArch::COMPUTE_62)
+      .Default(CudaVirtualArch::UNKNOWN);
+}
+
+CudaVirtualArch VirtualArchForCudaArch(CudaArch A) {
+  switch (A) {
+  case CudaArch::UNKNOWN:
+    return CudaVirtualArch::UNKNOWN;
+  case CudaArch::SM_20:
+  case CudaArch::SM_21:
+    return CudaVirtualArch::COMPUTE_20;
+  case CudaArch::SM_30:
+    return CudaVirtualArch::COMPUTE_30;
+  case CudaArch::SM_32:
+    return CudaVirtualArch::COMPUTE_32;
+  case CudaArch::SM_35:
+    return CudaVirtualArch::COMPUTE_35;
+  case CudaArch::SM_37:
+    return CudaVirtualArch::COMPUTE_37;
+  case CudaArch::SM_50:
+    return CudaVirtualArch::COMPUTE_50;
+  case CudaArch::SM_52:
+    return CudaVirtualArch::COMPUTE_52;
+  case CudaArch::SM_53:
+    return CudaVirtualArch::COMPUTE_53;
+  case CudaArch::SM_60:
+    return CudaVirtualArch::COMPUTE_60;
+  case CudaArch::SM_61:
+    return CudaVirtualArch::COMPUTE_61;
+  case CudaArch::SM_62:
+    return CudaVirtualArch::COMPUTE_62;
+  }
+}
+
+CudaVersion MinVersionForCudaArch(CudaArch A) {
+  switch (A) {
+  case CudaArch::UNKNOWN:
+    return CudaVersion::UNKNOWN;
+  case CudaArch::SM_20:
+  case CudaArch::SM_21:
+  case CudaArch::SM_30:
+  case CudaArch::SM_32:
+  case CudaArch::SM_35:
+  case CudaArch::SM_37:
+  case CudaArch::SM_50:
+  case CudaArch::SM_52:
+  case CudaArch::SM_53:
+    return CudaVersion::CUDA_70;
+  case CudaArch::SM_60:
+  case CudaArch::SM_61:
+  case CudaArch::SM_62:
+    return CudaVersion::CUDA_80;
+  }
+}
+
+} // namespace clang

cfe/trunk/lib/Basic/Targets.cpp

 //===--- Targets.cpp - Implement target feature support -------------------===//
 //
 //                     The LLVM Compiler Infrastructure
 //
 // This file is distributed under the University of Illinois Open Source
 // License. See LICENSE.TXT for details.
 //
 //===----------------------------------------------------------------------===//
 //
 // This file implements construction of a TargetInfo object from a
 // target triple.
 //
 //===----------------------------------------------------------------------===//
 
-#include "clang/Basic/TargetInfo.h"
 #include "clang/Basic/Builtins.h"
+#include "clang/Basic/Cuda.h"
 #include "clang/Basic/Diagnostic.h"
 #include "clang/Basic/LangOptions.h"
 #include "clang/Basic/MacroBuilder.h"
 #include "clang/Basic/TargetBuiltins.h"
+#include "clang/Basic/TargetInfo.h"
 #include "clang/Basic/TargetOptions.h"
 #include "clang/Basic/Version.h"
 #include "llvm/ADT/APFloat.h"
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/ADT/StringExtras.h"
 #include "llvm/ADT/StringRef.h"
 #include "llvm/ADT/StringSwitch.h"
 #include "llvm/ADT/Triple.h"
static const unsigned NVPTXAddrSpaceMap[] = {
     1, // cuda_device
     4, // cuda_constant
     3, // cuda_shared
 };
 
 class NVPTXTargetInfo : public TargetInfo {
   static const char *const GCCRegNames[];
   static const Builtin::Info BuiltinInfo[];
-
-  // The GPU profiles supported by the NVPTX backend
-  enum GPUKind {
-    GK_NONE,
-    GK_SM20,
-    GK_SM21,
-    GK_SM30,
-    GK_SM32,
-    GK_SM35,
-    GK_SM37,
-    GK_SM50,
-    GK_SM52,
-    GK_SM53,
-    GK_SM60,
-    GK_SM61,
-    GK_SM62,
-  } GPU;
+  CudaArch GPU;
 
 public:
   NVPTXTargetInfo(const llvm::Triple &Triple, const TargetOptions &Opts)
       : TargetInfo(Triple) {
     BigEndian = false;
     TLSSupported = false;
     LongWidth = LongAlign = 64;
     AddrSpaceMap = &NVPTXAddrSpaceMap;
     UseAddrSpaceMapMangling = true;
     // Define available target features
     // These must be defined in sorted order!
     NoAsmVariants = true;
-    // Set the default GPU to sm20
-    GPU = GK_SM20;
+    GPU = CudaArch::SM_20;
 
     // If possible, get a TargetInfo for our host triple, so we can match its
     // types.
     llvm::Triple HostTriple(Opts.HostTriple);
     if (HostTriple.isNVPTX())
       return;
     std::unique_ptr<TargetInfo> HostTarget(
         AllocateTarget(llvm::Triple(Opts.HostTriple), Opts));
public:
   void getTargetDefines(const LangOptions &Opts,
                         MacroBuilder &Builder) const override {
     Builder.defineMacro("__PTX__");
     Builder.defineMacro("__NVPTX__");
     if (Opts.CUDAIsDevice) {
       // Set __CUDA_ARCH__ for the GPU specified.
       std::string CUDAArchCode = [this] {
         switch (GPU) {
-        case GK_NONE:
+        case CudaArch::UNKNOWN:
           assert(false && "No GPU arch when compiling CUDA device code.");
           return "";
-        case GK_SM20:
+        case CudaArch::SM_20:
           return "200";
-        case GK_SM21:
+        case CudaArch::SM_21:
           return "210";
-        case GK_SM30:
+        case CudaArch::SM_30:
           return "300";
-        case GK_SM32:
+        case CudaArch::SM_32:
           return "320";
-        case GK_SM35:
+        case CudaArch::SM_35:
           return "350";
-        case GK_SM37:
+        case CudaArch::SM_37:
           return "370";
-        case GK_SM50:
+        case CudaArch::SM_50:
           return "500";
-        case GK_SM52:
+        case CudaArch::SM_52:
           return "520";
-        case GK_SM53:
+        case CudaArch::SM_53:
           return "530";
-        case GK_SM60:
+        case CudaArch::SM_60:
           return "600";
-        case GK_SM61:
+        case CudaArch::SM_61:
           return "610";
-        case GK_SM62:
+        case CudaArch::SM_62:
           return "620";
         }
       }();
       Builder.defineMacro("__CUDA_ARCH__", CUDAArchCode);
     }
   }
   ArrayRef<Builtin::Info> getTargetBuiltins() const override {
     return llvm::makeArrayRef(BuiltinInfo,
const char *getClobbers() const override {
     // FIXME: Is this really right?
     return "";
   }
   BuiltinVaListKind getBuiltinVaListKind() const override {
     // FIXME: implement
     return TargetInfo::CharPtrBuiltinVaList;
   }
   bool setCPU(const std::string &Name) override {
-    GPU = llvm::StringSwitch<GPUKind>(Name)
-              .Case("sm_20", GK_SM20)
-              .Case("sm_21", GK_SM21)
-              .Case("sm_30", GK_SM30)
-              .Case("sm_32", GK_SM32)
-              .Case("sm_35", GK_SM35)
-              .Case("sm_37", GK_SM37)
-              .Case("sm_50", GK_SM50)
-              .Case("sm_52", GK_SM52)
-              .Case("sm_53", GK_SM53)
-              .Case("sm_60", GK_SM60)
-              .Case("sm_61", GK_SM61)
-              .Case("sm_62", GK_SM62)
-              .Default(GK_NONE);
-
-    return GPU != GK_NONE;
+    GPU = StringToCudaArch(Name);
+    return GPU != CudaArch::UNKNOWN;
   }
   void setSupportedOpenCLOpts() override {
     auto &Opts = getSupportedOpenCLOpts();
     Opts.cl_clang_storage_class_specifiers = 1;
     Opts.cl_khr_gl_sharing = 1;
     Opts.cl_khr_icd = 1;
 
     Opts.cl_khr_fp64 = 1;

cfe/trunk/lib/Driver/Action.cpp

InputAction::InputAction(const Arg &_Input, types::ID _Type)
   : Action(InputClass, _Type), Input(_Input) {
 }
 
 void BindArchAction::anchor() {}
 
 BindArchAction::BindArchAction(Action *Input, const char *_ArchName)
     : Action(BindArchClass, Input), ArchName(_ArchName) {}
 
-// Converts CUDA GPU architecture, e.g. "sm_21", to its corresponding virtual
-// compute arch, e.g. "compute_20".  Returns null if the input arch is null or
-// doesn't match an existing arch.
-static const char* GpuArchToComputeName(const char *ArchName) {
-  if (!ArchName)
-    return nullptr;
-  return llvm::StringSwitch<const char *>(ArchName)
-      .Cases("sm_20", "sm_21", "compute_20")
-      .Case("sm_30", "compute_30")
-      .Case("sm_32", "compute_32")
-      .Case("sm_35", "compute_35")
-      .Case("sm_37", "compute_37")
-      .Case("sm_50", "compute_50")
-      .Case("sm_52", "compute_52")
-      .Case("sm_53", "compute_53")
-      .Case("sm_60", "compute_60")
-      .Case("sm_61", "compute_61")
-      .Case("sm_62", "compute_62")
-      .Default(nullptr);
-}
-
 void CudaDeviceAction::anchor() {}
 
-CudaDeviceAction::CudaDeviceAction(Action *Input, const char *ArchName,
+CudaDeviceAction::CudaDeviceAction(Action *Input, CudaArch Arch,
                                    bool AtTopLevel)
-    : Action(CudaDeviceClass, Input), GpuArchName(ArchName),
-      AtTopLevel(AtTopLevel) {
-  assert(!GpuArchName || IsValidGpuArchName(GpuArchName));
-}
-
-const char *CudaDeviceAction::getComputeArchName() const {
-  return GpuArchToComputeName(GpuArchName);
-}
-
-bool CudaDeviceAction::IsValidGpuArchName(llvm::StringRef ArchName) {
-  return GpuArchToComputeName(ArchName.data()) != nullptr;
-}
+    : Action(CudaDeviceClass, Input), GpuArch(Arch), AtTopLevel(AtTopLevel) {}
 
 void CudaHostAction::anchor() {}
 
 CudaHostAction::CudaHostAction(Action *Input, const ActionList &DeviceActions)
     : Action(CudaHostClass, Input), DeviceActions(DeviceActions) {}
 
 void JobAction::anchor() {}
 

cfe/trunk/lib/Driver/Driver.cpp

 #include "clang/Driver/DriverDiagnostic.h"
 #include "clang/Driver/Job.h"
 #include "clang/Driver/Options.h"
 #include "clang/Driver/SanitizerArgs.h"
 #include "clang/Driver/Tool.h"
 #include "clang/Driver/ToolChain.h"
 #include "llvm/ADT/ArrayRef.h"
 #include "llvm/ADT/STLExtras.h"
+#include "llvm/ADT/SmallSet.h"
 #include "llvm/ADT/StringExtras.h"
 #include "llvm/ADT/StringSet.h"
 #include "llvm/ADT/StringSwitch.h"
 #include "llvm/Option/Arg.h"
 #include "llvm/Option/ArgList.h"
 #include "llvm/Option/OptSpecifier.h"
 #include "llvm/Option/OptTable.h"
 #include "llvm/Option/Option.h"
static unsigned PrintActions1(const Compilation &C, Action *A,
 
   os << Action::getClassName(A->getKind()) << ", ";
   if (InputAction *IA = dyn_cast<InputAction>(A)) {
     os << "\"" << IA->getInputArg().getValue() << "\"";
   } else if (BindArchAction *BIA = dyn_cast<BindArchAction>(A)) {
     os << '"' << BIA->getArchName() << '"' << ", {"
        << PrintActions1(C, *BIA->input_begin(), Ids) << "}";
   } else if (CudaDeviceAction *CDA = dyn_cast<CudaDeviceAction>(A)) {
-    os << '"'
-       << (CDA->getGpuArchName() ? CDA->getGpuArchName() : "(multiple archs)")
-       << '"' << ", {" << PrintActions1(C, *CDA->input_begin(), Ids) << "}";
+    CudaArch Arch = CDA->getGpuArch();
+    if (Arch != CudaArch::UNKNOWN)
+      os << "'" << CudaArchToString(Arch) << "', ";
+    os << "{" << PrintActions1(C, *CDA->input_begin(), Ids) << "}";
   } else {
     const ActionList *AL;
     if (CudaHostAction *CHA = dyn_cast<CudaHostAction>(A)) {
       os << "{" << PrintActions1(C, *CHA->input_begin(), Ids) << "}"
          << ", gpu binaries ";
       AL = &CHA->getDeviceActions();
     } else
       AL = &A->getInputs();
static Action *buildCudaActions(Compilation &C, DerivedArgList &Args,
   bool CompileDeviceOnly =
       PartialCompilationArg &&
       PartialCompilationArg->getOption().matches(options::OPT_cuda_device_only);
 
   if (CompileHostOnly)
     return C.MakeAction<CudaHostAction>(HostAction, ActionList());
 
   // Collect all cuda_gpu_arch parameters, removing duplicates.
-  SmallVector<const char *, 4> GpuArchList;
-  llvm::StringSet<> GpuArchNames;
+  SmallVector<CudaArch, 4> GpuArchList;
+  llvm::SmallSet<CudaArch, 4> GpuArchs;
   for (Arg *A : Args) {
     if (!A->getOption().matches(options::OPT_cuda_gpu_arch_EQ))
       continue;
     A->claim();
 
-    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)
+    const auto &ArchStr = A->getValue();
+    CudaArch Arch = StringToCudaArch(ArchStr);
+    if (Arch == CudaArch::UNKNOWN)
+      C.getDriver().Diag(clang::diag::err_drv_cuda_bad_gpu_arch) << ArchStr;
+    else if (GpuArchs.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");
+    GpuArchList.push_back(CudaArch::SM_20);
 
   // Replicate inputs for each GPU architecture.
   Driver::InputList CudaDeviceInputs;
   for (unsigned I = 0, E = GpuArchList.size(); I != E; ++I)
     CudaDeviceInputs.push_back(std::make_pair(types::TY_CUDA_DEVICE, InputArg));
 
   // Build actions for all device inputs.
   assert(C.getSingleOffloadToolChain<Action::OFK_Cuda>() &&
for (unsigned I = 0, E = GpuArchList.size(); I != E; ++I) {
 
     for (const auto& A : {AssembleAction, BackendAction}) {
       DeviceActions.push_back(C.MakeAction<CudaDeviceAction>(
           A, GpuArchList[I], /* AtTopLevel */ false));
     }
   }
   auto FatbinAction = C.MakeAction<CudaDeviceAction>(
       C.MakeAction<LinkJobAction>(DeviceActions, types::TY_CUDA_FATBIN),
-      /* GpuArchName = */ nullptr,
+      CudaArch::UNKNOWN,
       /* AtTopLevel = */ false);
   // Return a new host action that incorporates original host action and all
   // device actions.
   return C.MakeAction<CudaHostAction>(std::move(HostAction),
                                       ActionList({FatbinAction}));
 }
 
 void Driver::BuildActions(Compilation &C, DerivedArgList &Args,
return BuildJobsForAction(C, *BAA->input_begin(), TC, ArchName, AtTopLevel,
                               MultipleArchs, LinkingOutput, CachedResults);
   }
 
   if (const CudaDeviceAction *CDA = dyn_cast<CudaDeviceAction>(A)) {
     // Initial processing of CudaDeviceAction carries host params.
     // Call BuildJobsForAction() again, now with correct device parameters.
     InputInfo II = BuildJobsForAction(
         C, *CDA->input_begin(), C.getSingleOffloadToolChain<Action::OFK_Cuda>(),
-        CDA->getGpuArchName(), CDA->isAtTopLevel(), /*MultipleArchs=*/true,
-        LinkingOutput, CachedResults);
+        CudaArchToString(CDA->getGpuArch()), CDA->isAtTopLevel(),
+        /*MultipleArchs=*/true, LinkingOutput, CachedResults);
     // Currently II's Action is *CDA->input_begin().  Set it to CDA instead, so
     // that one can retrieve II's GPU arch.
     II.setAction(A);
     return II;
   }
 
   const ActionList *Inputs = &A->getInputs();
 

cfe/trunk/lib/Driver/Tools.cpp

void NVPTX::Linker::ConstructJob(Compilation &C, const JobAction &JA,
   CmdArgs.push_back(TC.getTriple().isArch64Bit() ? "-64" : "-32");
   CmdArgs.push_back(Args.MakeArgString("--create"));
   CmdArgs.push_back(Args.MakeArgString(Output.getFilename()));
 
   for (const auto& II : Inputs) {
     auto* A = cast<const CudaDeviceAction>(II.getAction());
     // We need to pass an Arch of the form "sm_XX" for cubin files and
     // "compute_XX" for ptx.
-    const char *Arch = (II.getType() == types::TY_PP_Asm)
-                           ? A->getComputeArchName()
-                           : A->getGpuArchName();
+    const char *Arch =
+        (II.getType() == types::TY_PP_Asm)
+            ? CudaVirtualArchToString(VirtualArchForCudaArch(A->getGpuArch()))
+            : CudaArchToString(A->getGpuArch());
     CmdArgs.push_back(Args.MakeArgString(llvm::Twine("--image=profile=") +
                                          Arch + ",file=" + II.getFilename()));
   }
 
   for (const auto& A : Args.getAllArgValues(options::OPT_Xcuda_fatbinary))
     CmdArgs.push_back(Args.MakeArgString(A));
 
   const char *Exec = Args.MakeArgString(TC.GetProgramPath("fatbinary"));
   C.addCommand(llvm::make_unique<Command>(JA, *this, Exec, CmdArgs, Inputs));
 }