[flang-commits] [clang] [flang] [llvm] clang: Start using new amdgpu subarch triples (PR #206482)

[Matt Arsenault via flang-commits]

================
@@ -2138,6 +2154,12 @@ void ToolChain::TranslateXarchArgs(
 static bool isXArchCompatibleTripleArch(const llvm::Triple &TT,
                                         StringRef XArchVal) {
   llvm::Triple ParsedTriple(XArchVal);
+
+  // Accept -Xarch_amdgcn for all amdgpu subarches, and -Xarch_amdgpu9 for
+  // amdgpu9.xx
+  if (TT.isAMDGCN() && ParsedTriple.isAMDGCN())
----------------
arsenm wrote:

I don't think that ever worked

https://github.com/llvm/llvm-project/pull/206482