[clang] [clang][AMDGPU] Clean-up handling of named barrier type (PR #207687)

[Yaxun Liu via cfe-commits]

================
@@ -1,5 +1,6 @@
 // REQUIRES: amdgpu-registered-target
 // RUN: %clang_cc1 -fsyntax-only -verify -std=gnu++11 -triple amdgcn -Wno-unused-value %s
+// RUN: %clang_cc1 -fcuda-is-device -fsyntax-only -verify -std=gnu++11 -triple x86_64 -aux-triple amdgcn -Wno-unused-value %s
----------------
yxsamliu wrote:

Same issue here: this RUN line mixes device mode with a host primary triple.

For host-side aux-target coverage, this should not use `-fcuda-is-device`. For device-side coverage, AMDGPU should be the primary triple and the host should be the aux triple.

https://github.com/llvm/llvm-project/pull/207687