Add a comment to explain __CUDA_ARCH__ handling

Signed-off-by: John Pennycook <john.pennycook@intel.com>

Author: Pennycook

codebasin/config.py

@@ -322,6 +322,9 @@ def parse_args(self, argv: list[str]) -> list[PreprocessorConfiguration]:
             defines.append("__CUDACC__")
 
             if pass_ != "default":
+                # The __CUDA_ARCH__ macro always has three digits.
+                # Multiplying the SM version by 10 gives the macro value.
+                # e.g., sm_71 corresponds to __CUDA_ARCH__=710.
                 arch = int(pass_) * 10
                 defines.append(f"__CUDA_ARCH__={arch}")