[CUDA] Mark CUDA-12.4 as supported and introduce ptx 8.4.

[Artem-B]

No description provided.

[llvmbot]

@llvm/pr-subscribers-clang

@llvm/pr-subscribers-clang-driver

Author: Artem Belevich (Artem-B)

Changes

---

Full diff: https://github.com/llvm/llvm-project/pull/91516.diff

6 Files Affected:

• (modified) clang/docs/ReleaseNotes.rst (+1)
• (modified) clang/include/clang/Basic/BuiltinsNVPTX.def (+4-1)
• (modified) clang/include/clang/Basic/Cuda.h (+2-1)
• (modified) clang/lib/Basic/Cuda.cpp (+3-2)
• (modified) clang/lib/Driver/ToolChains/Cuda.cpp (+3)
• (modified) llvm/lib/Target/NVPTX/NVPTX.td (+1-1)

diff --git a/clang/docs/ReleaseNotes.rst b/clang/docs/ReleaseNotes.rst
index 0f9728c00e648..a3c8e4141ca54 100644
--- a/clang/docs/ReleaseNotes.rst
+++ b/clang/docs/ReleaseNotes.rst
@@ -798,6 +798,7 @@ CUDA/HIP Language Changes
 
 CUDA Support
 ^^^^^^^^^^^^
+- Clang now supports CUDA SDK up to 12.4
 
 AIX Support
 ^^^^^^^^^^^
diff --git a/clang/include/clang/Basic/BuiltinsNVPTX.def b/clang/include/clang/Basic/BuiltinsNVPTX.def
index 8d3c5e69d55cf..9e243d740ed7a 100644
--- a/clang/include/clang/Basic/BuiltinsNVPTX.def
+++ b/clang/include/clang/Basic/BuiltinsNVPTX.def
@@ -61,7 +61,9 @@
 #pragma push_macro("PTX81")
 #pragma push_macro("PTX82")
 #pragma push_macro("PTX83")
-#define PTX83 "ptx83"
+#pragma push_macro("PTX84")
+#define PTX84 "ptx84"
+#define PTX83 "ptx83|" PTX84
 #define PTX82 "ptx82|" PTX83
 #define PTX81 "ptx81|" PTX82
 #define PTX80 "ptx80|" PTX81
@@ -1091,3 +1093,4 @@ TARGET_BUILTIN(__nvvm_getctarank_shared_cluster, "iv*3", "", AND(SM_90,PTX78))
 #pragma pop_macro("PTX81")
 #pragma pop_macro("PTX82")
 #pragma pop_macro("PTX83")
+#pragma pop_macro("PTX84")
diff --git a/clang/include/clang/Basic/Cuda.h b/clang/include/clang/Basic/Cuda.h
index ba0e4465a0f5a..2d67c4181d129 100644
--- a/clang/include/clang/Basic/Cuda.h
+++ b/clang/include/clang/Basic/Cuda.h
@@ -41,9 +41,10 @@ enum class CudaVersion {
   CUDA_121,
   CUDA_122,
   CUDA_123,
+  CUDA_124,
   FULLY_SUPPORTED = CUDA_123,
   PARTIALLY_SUPPORTED =
-      CUDA_123, // Partially supported. Proceed with a warning.
+      CUDA_124, // Partially supported. Proceed with a warning.
   NEW = 10000,  // Too new. Issue a warning, but allow using it.
 };
 const char *CudaVersionToString(CudaVersion V);
diff --git a/clang/lib/Basic/Cuda.cpp b/clang/lib/Basic/Cuda.cpp
index 113483db5729b..e8ce15eb0decb 100644
--- a/clang/lib/Basic/Cuda.cpp
+++ b/clang/lib/Basic/Cuda.cpp
@@ -14,7 +14,7 @@ struct CudaVersionMapEntry {
 };
 #define CUDA_ENTRY(major, minor)                                               \
   {                                                                            \
-#major "." #minor, CudaVersion::CUDA_##major##minor,                       \
+    #major "." #minor, CudaVersion::CUDA_##major##minor,                       \
         llvm::VersionTuple(major, minor)                                       \
   }
 
@@ -41,6 +41,7 @@ static const CudaVersionMapEntry CudaNameVersionMap[] = {
     CUDA_ENTRY(12, 1),
     CUDA_ENTRY(12, 2),
     CUDA_ENTRY(12, 3),
+    CUDA_ENTRY(12, 4),
     {"", CudaVersion::NEW, llvm::VersionTuple(std::numeric_limits<int>::max())},
     {"unknown", CudaVersion::UNKNOWN, {}} // End of list tombstone.
 };
@@ -241,7 +242,7 @@ CudaVersion MaxVersionForCudaArch(CudaArch A) {
   }
 }
 
-bool CudaFeatureEnabled(llvm::VersionTuple  Version, CudaFeature Feature) {
+bool CudaFeatureEnabled(llvm::VersionTuple Version, CudaFeature Feature) {
   return CudaFeatureEnabled(ToCudaVersion(Version), Feature);
 }
 
diff --git a/clang/lib/Driver/ToolChains/Cuda.cpp b/clang/lib/Driver/ToolChains/Cuda.cpp
index 6634e6d818b33..d5f93c9c830fa 100644
--- a/clang/lib/Driver/ToolChains/Cuda.cpp
+++ b/clang/lib/Driver/ToolChains/Cuda.cpp
@@ -82,6 +82,8 @@ CudaVersion getCudaVersion(uint32_t raw_version) {
     return CudaVersion::CUDA_122;
   if (raw_version < 12040)
     return CudaVersion::CUDA_123;
+  if (raw_version < 12050)
+    return CudaVersion::CUDA_124;
   return CudaVersion::NEW;
 }
 
@@ -688,6 +690,7 @@ void NVPTX::getNVPTXTargetFeatures(const Driver &D, const llvm::Triple &Triple,
   case CudaVersion::CUDA_##CUDA_VER:                                           \
     PtxFeature = "+ptx" #PTX_VER;                                              \
     break;
+    CASE_CUDA_VERSION(124, 84);
     CASE_CUDA_VERSION(123, 83);
     CASE_CUDA_VERSION(122, 82);
     CASE_CUDA_VERSION(121, 81);
diff --git a/llvm/lib/Target/NVPTX/NVPTX.td b/llvm/lib/Target/NVPTX/NVPTX.td
index 6aa98543e5e22..05457c71cd392 100644
--- a/llvm/lib/Target/NVPTX/NVPTX.td
+++ b/llvm/lib/Target/NVPTX/NVPTX.td
@@ -41,7 +41,7 @@ foreach sm = [20, 21, 30, 32, 35, 37, 50, 52, 53,
 def SM90a: FeatureSM<"90a", 901>;
 
 foreach version = [32, 40, 41, 42, 43, 50, 60, 61, 62, 63, 64, 65,
-                   70, 71, 72, 73, 74, 75, 76, 77, 78, 80, 81, 82, 83] in
+                   70, 71, 72, 73, 74, 75, 76, 77, 78, 80, 81, 82, 83, 84] in
   def PTX#version: FeaturePTX<version>;
 
 //===----------------------------------------------------------------------===//