Add support for sm120 as well.

Author: Sebastian Jodlowski

clang/include/clang/Basic/BuiltinsNVPTX.td

@@ -21,13 +21,17 @@ class SM<string version, list<SMFeatures> newer_list> : SMFeatures {
                         !strconcat(f, "|", newer.Features));
 }
 
+let Features = "sm_120a" in def SM_120a : SMFeatures;
+
+def SM_120 : SM<"120", [SM_120a]>;
+
 let Features = "sm_101a" in def SM_101a : SMFeatures;
 
-def SM_101 : SM<"101", [SM_101a]>;
+def SM_101 : SM<"101", [SM_101a, SM_120]>;
 
 let Features = "sm_100a" in def SM_100a : SMFeatures;
 
-def SM_100 : SM<"100", [SM_100a]>;
+def SM_100 : SM<"100", [SM_100a, SM_101]>;
 
 let Features = "sm_90a" in def SM_90a : SMFeatures;
 

clang/include/clang/Basic/Cuda.h

@@ -84,6 +84,8 @@ enum class OffloadArch {
   SM_100a,
   SM_101,
   SM_101a,
+  SM_120,
+  SM_120a,
   GFX600,
   GFX601,
   GFX602,

clang/lib/Basic/Cuda.cpp

@@ -102,6 +102,8 @@ static const OffloadArchToStringMap arch_names[] = {
     SM(100a),                        // Blackwell
     SM(101),                         // Blackwell
     SM(101a),                        // Blackwell
+    SM(120),                         // Blackwell
+    SM(120a),                        // Blackwell
     GFX(600),  // gfx600
     GFX(601),  // gfx601
     GFX(602),  // gfx602
@@ -234,6 +236,8 @@ CudaVersion MinVersionForOffloadArch(OffloadArch A) {
   case OffloadArch::SM_100a:
   case OffloadArch::SM_101:
   case OffloadArch::SM_101a:
+  case OffloadArch::SM_120:
+  case OffloadArch::SM_120a:
     return CudaVersion::CUDA_128;
   default:
     llvm_unreachable("invalid enum");

clang/lib/Basic/Targets/NVPTX.cpp

@@ -295,6 +295,9 @@ void NVPTXTargetInfo::getTargetDefines(const LangOptions &Opts,
       case OffloadArch::SM_101:
       case OffloadArch::SM_101a:
          return "1010";
+      case OffloadArch::SM_120:
+      case OffloadArch::SM_120a:
+         return "1200";
       }
       llvm_unreachable("unhandled OffloadArch");
     }();
@@ -305,6 +308,8 @@ void NVPTXTargetInfo::getTargetDefines(const LangOptions &Opts,
       Builder.defineMacro("__CUDA_ARCH_FEAT_SM100_ALL", "1");
     if (GPU == OffloadArch::SM_101a)
       Builder.defineMacro("__CUDA_ARCH_FEAT_SM101_ALL", "1");
+    if (GPU == OffloadArch::SM_120a)
+      Builder.defineMacro("__CUDA_ARCH_FEAT_SM120_ALL", "1");
   }
 }
 

clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp

@@ -2280,6 +2280,8 @@ void CGOpenMPRuntimeGPU::processRequiresDirective(const OMPRequiresDecl *D) {
       case OffloadArch::SM_100a:
       case OffloadArch::SM_101:
       case OffloadArch::SM_101a:
+      case OffloadArch::SM_120:
+      case OffloadArch::SM_120a:
       case OffloadArch::GFX600:
       case OffloadArch::GFX601:
       case OffloadArch::GFX602:

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

@@ -30,6 +30,8 @@
 // CHECK-SAME: {{^}}, sm_100a
 // CHECK-SAME: {{^}}, sm_101
 // CHECK-SAME: {{^}}, sm_101a
+// CHECK-SAME: {{^}}, sm_120
+// CHECK-SAME: {{^}}, sm_120a
 // CHECK-SAME: {{^}}, gfx600
 // CHECK-SAME: {{^}}, gfx601
 // CHECK-SAME: {{^}}, gfx602