Add support for sm101 target architecture (Tegra Blackwell)

Author: Sebastian Jodlowski

clang/include/clang/Basic/BuiltinsNVPTX.td

@@ -21,6 +21,10 @@ class SM<string version, list<SMFeatures> newer_list> : SMFeatures {
                         !strconcat(f, "|", newer.Features));
 }
 
+let Features = "sm_101a" in def SM_101a : SMFeatures;
+
+def SM_101 : SM<"101", [SM_101a]>;
+
 let Features = "sm_100a" in def SM_100a : SMFeatures;
 
 def SM_100 : SM<"100", [SM_100a]>;

clang/include/clang/Basic/Cuda.h

@@ -82,6 +82,8 @@ enum class OffloadArch {
   SM_90a,
   SM_100,
   SM_100a,
+  SM_101,
+  SM_101a,
   GFX600,
   GFX601,
   GFX602,

clang/lib/Basic/Cuda.cpp

@@ -100,6 +100,8 @@ static const OffloadArchToStringMap arch_names[] = {
     SM(90a),                         // Hopper
     SM(100),                         // Blackwell
     SM(100a),                        // Blackwell
+    SM(101),                         // Blackwell
+    SM(101a),                        // Blackwell
     GFX(600),  // gfx600
     GFX(601),  // gfx601
     GFX(602),  // gfx602
@@ -230,6 +232,8 @@ CudaVersion MinVersionForOffloadArch(OffloadArch A) {
     return CudaVersion::CUDA_120;
   case OffloadArch::SM_100:
   case OffloadArch::SM_100a:
+  case OffloadArch::SM_101:
+  case OffloadArch::SM_101a:
     return CudaVersion::CUDA_128;
   default:
     llvm_unreachable("invalid enum");

clang/lib/Basic/Targets/NVPTX.cpp

@@ -292,6 +292,9 @@ void NVPTXTargetInfo::getTargetDefines(const LangOptions &Opts,
       case OffloadArch::SM_100:
       case OffloadArch::SM_100a:
         return "1000";
+      case OffloadArch::SM_101:
+      case OffloadArch::SM_101a:
+         return "1010";
       }
       llvm_unreachable("unhandled OffloadArch");
     }();
@@ -300,6 +303,8 @@ void NVPTXTargetInfo::getTargetDefines(const LangOptions &Opts,
       Builder.defineMacro("__CUDA_ARCH_FEAT_SM90_ALL", "1");
     if (GPU == OffloadArch::SM_100a)
       Builder.defineMacro("__CUDA_ARCH_FEAT_SM100_ALL", "1");
+    if (GPU == OffloadArch::SM_101a)
+      Builder.defineMacro("__CUDA_ARCH_FEAT_SM101_ALL", "1");
   }
 }
 

clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp

@@ -2278,6 +2278,8 @@ void CGOpenMPRuntimeGPU::processRequiresDirective(const OMPRequiresDecl *D) {
       case OffloadArch::SM_90a:
       case OffloadArch::SM_100:
       case OffloadArch::SM_100a:
+      case OffloadArch::SM_101:
+      case OffloadArch::SM_101a:
       case OffloadArch::GFX600:
       case OffloadArch::GFX601:
       case OffloadArch::GFX602:

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

@@ -28,6 +28,8 @@
 // CHECK-SAME: {{^}}, sm_90a
 // CHECK-SAME: {{^}}, sm_100
 // CHECK-SAME: {{^}}, sm_100a
+// CHECK-SAME: {{^}}, sm_101
+// CHECK-SAME: {{^}}, sm_101a
 // CHECK-SAME: {{^}}, gfx600
 // CHECK-SAME: {{^}}, gfx601
 // CHECK-SAME: {{^}}, gfx602