[cutlass backend] check against arch >= 100 (#145812)

Summary:
Want to add a guard against silent fallback to SM90.

GenerateSM100 was just added 3 days ago. https://github.com/NVIDIA/cutlass/blame/main/python/cutlass_library/generator.py#L8896

It should show up in CUTLASS 3.8 (not pinned yet).

Test Plan: ci

Differential Revision: D68748705

Pull Request resolved: https://github.com/pytorch/pytorch/pull/145812
Approved by: https://github.com/chenyang78, https://github.com/ColinPeppler, https://github.com/Aidyn-A

torch/_inductor/codegen/cuda/cutlass_utils.py

@@ -114,8 +114,19 @@ def try_import_cutlass() -> bool:
     return False
 
 
+@functools.lru_cache(8)
 def _normalize_cuda_arch(arch: str) -> str:
-    if int(arch) >= 90:
+    if int(arch) >= 100:
+        log.warning(
+            "Detected CUDA architecture >= 100: %s. We will generate operations with "
+            "GenerateSM100 (if available) and GenerateSM90. Please file an "
+            "issue for any problems and feedback. ",
+            arch,
+        )
+
+    if int(arch) >= 100:
+        return "100"
+    elif int(arch) >= 90:
         return "90"
     elif int(arch) >= 80:
         return "80"
@@ -186,7 +197,15 @@ def _gen_ops_cached(arch, version) -> list[Any]:
     )
     manifest = cutlass_manifest.Manifest(args)
 
-    if arch == "90":
+    if arch == "100":
+        try:
+            from cutlass_generator import GenerateSM100  # type: ignore[import]
+
+            GenerateSM100(manifest, args.cuda_version)
+        except ImportError:
+            log.warning("Cannot find GenerateSM100. Only GenerateSM90 will be used. ")
+        cutlass_generator.GenerateSM90(manifest, args.cuda_version)
+    elif arch == "90":
         cutlass_generator.GenerateSM90(manifest, args.cuda_version)
         cutlass_generator.GenerateSM80(manifest, args.cuda_version)
     else: