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

[Artem Belevich via llvm-commits]

================
@@ -26,24 +27,38 @@ static cl::opt<bool>
     NoF16Math("nvptx-no-f16-math", cl::Hidden,
               cl::desc("NVPTX Specific: Disable generation of f16 math ops."),
               cl::init(false));
+static cl::opt<unsigned>
+    NextSM("nvptx-next-sm", cl::Hidden,
+           cl::desc("NVPTX Specific: Override SM ID for sm_next."),
+           cl::init(90));
----------------
Artem-B wrote:

Supported values live in [llvm/lib/Target/NVPTX/NVPTX.td, ](https://github.com/llvm/llvm-project/blob/cdc193459d90fad83d2eafaccfe03368a9a8a160/llvm/lib/Target/NVPTX/NVPTX.td#L4)

I think setting it to zero, and erroring out if the exp[licit value is not specified when we're targeting `sm_next` would probably be the right thing to do here. Will fix.

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