Add NVPTX Support to ValidCPUList (enabling march notes)

[erichkeane]

Summary

A followup to: https://reviews.llvm.org/D42978
This patch adds NVPTX support for
enabling the march notes.

[erichkeane]

rebased, reflowed tests.

[echristo]

LGTM with an inline comment.

include/clang/Basic/Cuda.h

49

We have last, invalid, etc... maybe we should pick one among all the targets? :)

[tra]

test/Misc/target-invalid-cpu-note.c

38

Nit: Generally speaking this note is false. For any given version of CUDA, some of the listed GPU variants will not be accepted. E.g. CUDA versions before 9.1 do not know anything about sm_72, but CUDA-9.1 does not supports sm_20.

[erichkeane]

test/Misc/target-invalid-cpu-note.c

38

Is there somewhere else that this is checked? It seems that the 'setCPU' function here checks against this same list.

[tra]

test/Misc/target-invalid-cpu-note.c

38

We have CheckCudaVersionSupportsArch() in lib/Driver/ToolChains/Cuda.cpp.
I'm not sure if CUDA version is already known at the time we check -target-cpu, though.

I'm OK with the ful list for now if it's hard to get to the CUDA version.

include/clang/Basic/Cuda.h

enum class CudaArch {
   SM_50,
   SM_52,
   SM_53,
   SM_60,
   SM_61,
   SM_62,
   SM_70,
   SM_72,
+  LAST,
 };
 const char *CudaArchToString(CudaArch A);
 
 // The input should have the form "sm_20".
 CudaArch StringToCudaArch(llvm::StringRef S);
 
 enum class CudaVirtualArch {
   UNKNOWN,

lib/Basic/Cuda.cpp

const char *CudaVersionToString(CudaVersion V) {
   case CudaVersion::CUDA_91:
     return "9.1";
   }
   llvm_unreachable("invalid enum");
 }
 
 const char *CudaArchToString(CudaArch A) {
   switch (A) {
+  case CudaArch::LAST:
+    break;
   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";
return llvm::StringSwitch<CudaVirtualArch>(S)
       .Case("compute_62", CudaVirtualArch::COMPUTE_62)
       .Case("compute_70", CudaVirtualArch::COMPUTE_70)
       .Case("compute_72", CudaVirtualArch::COMPUTE_72)
       .Default(CudaVirtualArch::UNKNOWN);
 }
 
 CudaVirtualArch VirtualArchForCudaArch(CudaArch A) {
   switch (A) {
+  case CudaArch::LAST:
+    break;
   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:
CudaVirtualArch VirtualArchForCudaArch(CudaArch A) {
   case CudaArch::SM_72:
     return CudaVirtualArch::COMPUTE_72;
   }
   llvm_unreachable("invalid enum");
 }
 
 CudaVersion MinVersionForCudaArch(CudaArch A) {
   switch (A) {
+  case CudaArch::LAST:
+    break;
   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:

lib/Basic/Targets/NVPTX.h

BuiltinVaListKind getBuiltinVaListKind() const override {
     // FIXME: implement
     return TargetInfo::CharPtrBuiltinVaList;
   }
 
   bool isValidCPUName(StringRef Name) const override {
     return StringToCudaArch(Name) != CudaArch::UNKNOWN;
   }
 
+  void fillValidCPUList(SmallVectorImpl<StringRef> &Values) const override {
+    for (int i = static_cast<int>(CudaArch::SM_20);
+         i < static_cast<int>(CudaArch::LAST); ++i)
+      Values.emplace_back(CudaArchToString(static_cast<CudaArch>(i)));
+  }
+
   bool setCPU(const std::string &Name) override {
     GPU = StringToCudaArch(Name);
     return GPU != CudaArch::UNKNOWN;
   }
 
   void setSupportedOpenCLOpts() override {
     auto &Opts = getSupportedOpenCLOpts();
     Opts.support("cl_clang_storage_class_specifiers");

lib/Basic/Targets/NVPTX.cpp

 void NVPTXTargetInfo::getTargetDefines(const LangOptions &Opts,
                                        MacroBuilder &Builder) const {
   Builder.defineMacro("__PTX__");
   Builder.defineMacro("__NVPTX__");
   if (Opts.CUDAIsDevice) {
     // Set __CUDA_ARCH__ for the GPU specified.
     std::string CUDAArchCode = [this] {
       switch (GPU) {
+      case CudaArch::LAST:
+        break;
       case CudaArch::UNKNOWN:
         assert(false && "No GPU arch when compiling CUDA device code.");
         return "";
       case CudaArch::SM_20:
         return "200";
       case CudaArch::SM_21:
         return "210";
       case CudaArch::SM_30:

test/Misc/target-invalid-cpu-note.c

 // X86_64: error: unknown target CPU 'not-a-cpu'
 // X86_64: note: valid target CPU values are: nocona, core2, penryn, bonnell,
 // X86_64-SAME: atom, silvermont, slm, goldmont, nehalem, corei7, westmere,
 // X86_64-SAME: sandybridge, corei7-avx, ivybridge, core-avx-i, haswell,
 // X86_64-SAME: core-avx2, broadwell, skylake, skylake-avx512, skx, cannonlake,
 // X86_64-SAME: icelake, knl, knm, k8, athlon64, athlon-fx, opteron, k8-sse3,
 // X86_64-SAME: athlon64-sse3, opteron-sse3, amdfam10, barcelona, btver1,
 // X86_64-SAME: btver2, bdver1, bdver2, bdver3, bdver4, znver1, x86-64
+
+// RUN: not %clang_cc1 -triple nvptx--- -target-cpu not-a-cpu -fsyntax-only %s 2>&1 | FileCheck %s --check-prefix NVPTX
+// NVPTX: error: unknown target CPU 'not-a-cpu'
+// NVPTX: note: valid target CPU values are: sm_20, sm_21, sm_30, sm_32, sm_35,
+// NVPTX-SAME: sm_37, sm_50, sm_52, sm_53, sm_60, sm_61, sm_62, sm_70, sm_72