[clang] [llvm] [CUDA] Add a pseudo GPU sm_next which allows overriding for SM/PTX version. (PR #100247)

[Artem Belevich via cfe-commits]

================
@@ -52,6 +53,42 @@ const char *CudaVersionToString(CudaVersion V);
 // Input is "Major.Minor"
 CudaVersion CudaStringToVersion(const llvm::Twine &S);
 
+enum class PTXVersion {
+  PTX_UNKNOWN = 0,
+  PTX_32 = 32,
+  PTX_40 = 40,
+  PTX_41,
+  PTX_42,
+  PTX_43,
+  PTX_50 = 50,
+  PTX_60 = 60,
+  PTX_61,
+  PTX_62,
+  PTX_63,
+  PTX_64,
+  PTX_65,
+  PTX_70 = 70,
+  PTX_71,
+  PTX_72,
+  PTX_73,
+  PTX_74,
+  PTX_75,
+  PTX_76,
+  PTX_77,
+  PTX_78,
+  PTX_80 = 80,
+  PTX_81,
+  PTX_82,
+  PTX_83,
+  PTX_84,
+  PTX_85,
+  PTX_LAST = PTX_85,
+  PTX_custom = 9999, // placeholder for an unknown future version.
----------------
Artem-B wrote:

I've just used the same value I use in NVPTX.td on LLVM side, where I can't use UINT_MAX, though it's just for consistency. In case someone searches for 9999, they will find all related occurrences.

Specific values do not matter in either place, as long as they are distinct from other values. 

https://github.com/llvm/llvm-project/pull/100247