[CUDA] Add amdgpu sub archs

Patch by Greg Rodgers.
Revised and lit tests added by Yaxun Liu.

Differential Revision: https://reviews.llvm.org/D45277

llvm-svn: 329232

6 files changed, 152 insertions, 42 deletions

diff --git a/clang/lib/Basic/Cuda.cpp b/clang/lib/Basic/Cuda.cpp
index b7e04bf3f96..970c060c0be 100644
--- a/clang/lib/Basic/Cuda.cpp
+++ b/clang/lib/Basic/Cuda.cpp
@@ -58,6 +58,32 @@ const char *CudaArchToString(CudaArch A) {
     return "sm_70";
   case CudaArch::SM_72:
     return "sm_72";
+  case CudaArch::GFX600: // tahiti
+    return "gfx600";
+  case CudaArch::GFX601: // pitcairn, verde, oland,hainan
+    return "gfx601";
+  case CudaArch::GFX700: // kaveri
+    return "gfx700";
+  case CudaArch::GFX701: // hawaii
+    return "gfx701";
+  case CudaArch::GFX702: // 290,290x,R390,R390x
+    return "gfx702";
+  case CudaArch::GFX703: // kabini mullins
+    return "gfx703";
+  case CudaArch::GFX704: // bonaire
+    return "gfx704";
+  case CudaArch::GFX801: // carrizo
+    return "gfx801";
+  case CudaArch::GFX802: // tonga,iceland
+    return "gfx802";
+  case CudaArch::GFX803: // fiji,polaris10
+    return "gfx803";
+  case CudaArch::GFX810: // stoney
+    return "gfx810";
+  case CudaArch::GFX900: // vega, instinct
+    return "gfx900";
+  case CudaArch::GFX902: // TBA
+    return "gfx902";
   }
   llvm_unreachable("invalid enum");
 }
@@ -78,6 +104,19 @@ CudaArch StringToCudaArch(llvm::StringRef S) {
       .Case("sm_62", CudaArch::SM_62)
       .Case("sm_70", CudaArch::SM_70)
       .Case("sm_72", CudaArch::SM_72)
+      .Case("gfx600", CudaArch::GFX600)
+      .Case("gfx601", CudaArch::GFX601)
+      .Case("gfx700", CudaArch::GFX700)
+      .Case("gfx701", CudaArch::GFX701)
+      .Case("gfx702", CudaArch::GFX702)
+      .Case("gfx703", CudaArch::GFX703)
+      .Case("gfx704", CudaArch::GFX704)
+      .Case("gfx801", CudaArch::GFX801)
+      .Case("gfx802", CudaArch::GFX802)
+      .Case("gfx803", CudaArch::GFX803)
+      .Case("gfx810", CudaArch::GFX810)
+      .Case("gfx900", CudaArch::GFX900)
+      .Case("gfx902", CudaArch::GFX902)
       .Default(CudaArch::UNKNOWN);
 }
 
@@ -111,6 +150,8 @@ const char *CudaVirtualArchToString(CudaVirtualArch A) {
     return "compute_70";
   case CudaVirtualArch::COMPUTE_72:
     return "compute_72";
+  case CudaVirtualArch::COMPUTE_AMDGCN:
+    return "compute_amdgcn";
   }
   llvm_unreachable("invalid enum");
 }
@@ -130,6 +171,7 @@ CudaVirtualArch StringToCudaVirtualArch(llvm::StringRef S) {
       .Case("compute_62", CudaVirtualArch::COMPUTE_62)
       .Case("compute_70", CudaVirtualArch::COMPUTE_70)
       .Case("compute_72", CudaVirtualArch::COMPUTE_72)
+      .Case("compute_amdgcn", CudaVirtualArch::COMPUTE_AMDGCN)
       .Default(CudaVirtualArch::UNKNOWN);
 }
 
@@ -166,6 +208,20 @@ CudaVirtualArch VirtualArchForCudaArch(CudaArch A) {
     return CudaVirtualArch::COMPUTE_70;
   case CudaArch::SM_72:
     return CudaVirtualArch::COMPUTE_72;
+  case CudaArch::GFX600:
+  case CudaArch::GFX601:
+  case CudaArch::GFX700:
+  case CudaArch::GFX701:
+  case CudaArch::GFX702:
+  case CudaArch::GFX703:
+  case CudaArch::GFX704:
+  case CudaArch::GFX801:
+  case CudaArch::GFX802:
+  case CudaArch::GFX803:
+  case CudaArch::GFX810:
+  case CudaArch::GFX900:
+  case CudaArch::GFX902:
+    return CudaVirtualArch::COMPUTE_AMDGCN;
   }
   llvm_unreachable("invalid enum");
 }
@@ -194,6 +250,20 @@ CudaVersion MinVersionForCudaArch(CudaArch A) {
     return CudaVersion::CUDA_90;
   case CudaArch::SM_72:
     return CudaVersion::CUDA_91;
+  case CudaArch::GFX600:
+  case CudaArch::GFX601:
+  case CudaArch::GFX700:
+  case CudaArch::GFX701:
+  case CudaArch::GFX702:
+  case CudaArch::GFX703:
+  case CudaArch::GFX704:
+  case CudaArch::GFX801:
+  case CudaArch::GFX802:
+  case CudaArch::GFX803:
+  case CudaArch::GFX810:
+  case CudaArch::GFX900:
+  case CudaArch::GFX902:
+    return CudaVersion::CUDA_70;
   }
   llvm_unreachable("invalid enum");
 }
@@ -204,6 +274,19 @@ CudaVersion MaxVersionForCudaArch(CudaArch A) {
     return CudaVersion::UNKNOWN;
   case CudaArch::SM_20:
   case CudaArch::SM_21:
+  case CudaArch::GFX600:
+  case CudaArch::GFX601:
+  case CudaArch::GFX700:
+  case CudaArch::GFX701:
+  case CudaArch::GFX702:
+  case CudaArch::GFX703:
+  case CudaArch::GFX704:
+  case CudaArch::GFX801:
+  case CudaArch::GFX802:
+  case CudaArch::GFX803:
+  case CudaArch::GFX810:
+  case CudaArch::GFX900:
+  case CudaArch::GFX902:
     return CudaVersion::CUDA_80;
   default:
     return CudaVersion::LATEST;
diff --git a/clang/lib/Basic/Targets.cpp b/clang/lib/Basic/Targets.cpp
index e325403a1fa..4311095063c 100644
--- a/clang/lib/Basic/Targets.cpp
+++ b/clang/lib/Basic/Targets.cpp
@@ -112,6 +112,61 @@ void addMinGWDefines(const llvm::Triple &Triple, const LangOptions &Opts,
   addCygMingDefines(Opts, Builder);
 }
 
+void defineCudaArchMacro(CudaArch GPU, clang::MacroBuilder &Builder) {
+  std::string CUDAArchCode = [GPU] {
+    switch (GPU) {
+    case CudaArch::LAST:
+      break;
+    case CudaArch::SM_20:
+      return "200";
+    case CudaArch::SM_21:
+      return "210";
+    case CudaArch::SM_30:
+      return "300";
+    case CudaArch::SM_32:
+      return "320";
+    case CudaArch::SM_35:
+      return "350";
+    case CudaArch::SM_37:
+      return "370";
+    case CudaArch::SM_50:
+      return "500";
+    case CudaArch::SM_52:
+      return "520";
+    case CudaArch::SM_53:
+      return "530";
+    case CudaArch::SM_60:
+      return "600";
+    case CudaArch::SM_61:
+      return "610";
+    case CudaArch::SM_62:
+      return "620";
+    case CudaArch::SM_70:
+      return "700";
+    case CudaArch::SM_72:
+      return "720";
+    case CudaArch::GFX600:
+    case CudaArch::GFX601:
+    case CudaArch::GFX700:
+    case CudaArch::GFX701:
+    case CudaArch::GFX702:
+    case CudaArch::GFX703:
+    case CudaArch::GFX704:
+    case CudaArch::GFX801:
+    case CudaArch::GFX802:
+    case CudaArch::GFX803:
+    case CudaArch::GFX810:
+    case CudaArch::GFX900:
+    case CudaArch::GFX902:
+      return "320";
+    case CudaArch::UNKNOWN:
+      llvm_unreachable("unhandled Cuda/HIP Arch");
+    }
+    llvm_unreachable("unhandled Cuda/HIP Arch");
+  }();
+  Builder.defineMacro("__CUDA_ARCH__", CUDAArchCode);
+}
+
 //===----------------------------------------------------------------------===//
 // Driver code
 //===----------------------------------------------------------------------===//
diff --git a/clang/lib/Basic/Targets.h b/clang/lib/Basic/Targets.h
index 6fc967ddabe..bb933c00a34 100644
--- a/clang/lib/Basic/Targets.h
+++ b/clang/lib/Basic/Targets.h
@@ -16,6 +16,7 @@
 #ifndef LLVM_CLANG_LIB_BASIC_TARGETS_H
 #define LLVM_CLANG_LIB_BASIC_TARGETS_H
 
+#include "clang/Basic/Cuda.h"
 #include "clang/Basic/LangOptions.h"
 #include "clang/Basic/MacroBuilder.h"
 #include "clang/Basic/TargetInfo.h"
@@ -46,6 +47,9 @@ void addMinGWDefines(const llvm::Triple &Triple, const clang::LangOptions &Opts,
 LLVM_LIBRARY_VISIBILITY
 void addCygMingDefines(const clang::LangOptions &Opts,
                        clang::MacroBuilder &Builder);
+
+LLVM_LIBRARY_VISIBILITY
+void defineCudaArchMacro(CudaArch GPU, clang::MacroBuilder &Builder);
 } // namespace targets
 } // namespace clang
 #endif // LLVM_CLANG_LIB_BASIC_TARGETS_H
diff --git a/clang/lib/Basic/Targets/AMDGPU.cpp b/clang/lib/Basic/Targets/AMDGPU.cpp
index d956d132a5b..51943c737a1 100644
--- a/clang/lib/Basic/Targets/AMDGPU.cpp
+++ b/clang/lib/Basic/Targets/AMDGPU.cpp
@@ -12,6 +12,7 @@
 //===----------------------------------------------------------------------===//
 
 #include "AMDGPU.h"
+#include "Targets.h"
 #include "clang/Basic/Builtins.h"
 #include "clang/Basic/LangOptions.h"
 #include "clang/Basic/MacroBuilder.h"
@@ -263,6 +264,7 @@ AMDGPUTargetInfo::AMDGPUTargetInfo(const llvm::Triple &Triple,
   resetDataLayout(isAMDGCN(getTriple()) ? DataLayoutStringAMDGCN
                                         : DataLayoutStringR600);
   assert(DataLayout->getAllocaAddrSpace() == Private);
+  GCN_Subarch = CudaArch::GFX803; // Default to fiji
 
   setAddressSpaceMap(Triple.getOS() == llvm::Triple::Mesa3D ||
                      !isAMDGCN(Triple));
@@ -307,6 +309,9 @@ void AMDGPUTargetInfo::getTargetDefines(const LangOptions &Opts,
   if (GPU.Kind != GK_NONE)
     Builder.defineMacro(Twine("__") + Twine(GPU.CanonicalName) + Twine("__"));
 
+  if (Opts.CUDAIsDevice)
+    defineCudaArchMacro(GCN_Subarch, Builder);
+
   // TODO: __HAS_FMAF__, __HAS_LDEXPF__, __HAS_FP64__ are deprecated and will be
   // removed in the near future.
   if (GPU.HasFMAF)
diff --git a/clang/lib/Basic/Targets/AMDGPU.h b/clang/lib/Basic/Targets/AMDGPU.h
index f7c477659dc..1249031d066 100644
--- a/clang/lib/Basic/Targets/AMDGPU.h
+++ b/clang/lib/Basic/Targets/AMDGPU.h
@@ -14,6 +14,7 @@
 #ifndef LLVM_CLANG_LIB_BASIC_TARGETS_AMDGPU_H
 #define LLVM_CLANG_LIB_BASIC_TARGETS_AMDGPU_H
 
+#include "clang/Basic/Cuda.h"
 #include "clang/Basic/TargetInfo.h"
 #include "clang/Basic/TargetOptions.h"
 #include "llvm/ADT/StringSet.h"
@@ -174,6 +175,7 @@ class LLVM_LIBRARY_VISIBILITY AMDGPUTargetInfo final : public TargetInfo {
   static bool isAMDGCN(const llvm::Triple &TT) {
     return TT.getArch() == llvm::Triple::amdgcn;
   }
+  CudaArch GCN_Subarch;
 
 public:
   AMDGPUTargetInfo(const llvm::Triple &Triple, const TargetOptions &Opts);
@@ -330,6 +332,7 @@ public:
     else
       GPU = parseR600Name(Name);
 
+    GCN_Subarch = StringToCudaArch(Name);
     return GK_NONE != GPU.Kind;
   }
 
diff --git a/clang/lib/Basic/Targets/NVPTX.cpp b/clang/lib/Basic/Targets/NVPTX.cpp
index e8e6d2f2f74..cb24fd9c3e9 100644
--- a/clang/lib/Basic/Targets/NVPTX.cpp
+++ b/clang/lib/Basic/Targets/NVPTX.cpp
@@ -153,48 +153,8 @@ 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:
-        return "300";
-      case CudaArch::SM_32:
-        return "320";
-      case CudaArch::SM_35:
-        return "350";
-      case CudaArch::SM_37:
-        return "370";
-      case CudaArch::SM_50:
-        return "500";
-      case CudaArch::SM_52:
-        return "520";
-      case CudaArch::SM_53:
-        return "530";
-      case CudaArch::SM_60:
-        return "600";
-      case CudaArch::SM_61:
-        return "610";
-      case CudaArch::SM_62:
-        return "620";
-      case CudaArch::SM_70:
-        return "700";
-      case CudaArch::SM_72:
-        return "720";
-      }
-      llvm_unreachable("unhandled CudaArch");
-    }();
-    Builder.defineMacro("__CUDA_ARCH__", CUDAArchCode);
-  }
+  if (Opts.CUDAIsDevice)
+    defineCudaArchMacro(GPU, Builder);
 }
 
 ArrayRef<Builtin::Info> NVPTXTargetInfo::getTargetBuiltins() const {