[clang] [llvm] [OpenMP]Initial parsing/sema support for target_device selector set (PR #118471)

[via llvm-commits]

================
@@ -2903,7 +2903,10 @@ TargetOMPContext::TargetOMPContext(
     const FunctionDecl *CurrentFunctionDecl,
     ArrayRef<llvm::omp::TraitProperty> ConstructTraits)
     : OMPContext(ASTCtx.getLangOpts().OpenMPIsTargetDevice,
-                 ASTCtx.getTargetInfo().getTriple()),
+                 ASTCtx.getTargetInfo().getTriple(),
+                 ASTCtx.getLangOpts().OMPTargetTriples.empty()
+                     ? llvm::Triple()
+                     : ASTCtx.getLangOpts().OMPTargetTriples[0]),
----------------
Ritanya-B-Bharadwaj wrote:

>From what I understand, the --offload-arch option is mostly for non-host devices like GPUs and doesn’t apply to the host. For the host, the architecture is determined by the default target triple. Since the host isn’t treated as an offloading target, --offload-arch doesn’t affect its code generation.

https://github.com/llvm/llvm-project/pull/118471