Re-add the __try_wait API as pipeline uses internal symbols of barrier for some reason

Author: wmaxey

include/cuda/std/barrier

@@ -87,7 +87,7 @@ public:
     using arrival_token = typename __barrier_base::arrival_token;
 
     _LIBCUDACXX_INLINE_VISIBILITY
-    bool try_wait(arrival_token __phase) const {
+    bool __try_wait(arrival_token __phase) const {
 #if __CUDA_ARCH__ >= 800
         if (__isShared(&__barrier)) {
             int __ready = 0;
@@ -220,6 +220,11 @@ public:
         _CUDA_VSTD::__libcpp_thread_poll_with_backoff(std::__barrier_poll_tester<barrier>(this, _CUDA_VSTD::move(__phase)));
     }
 
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool try_wait(arrival_token __phase) const {
+        return __try_wait(_CUDA_VSTD::move(__phase));
+    }
+
     inline _LIBCUDACXX_INLINE_VISIBILITY
     bool try_wait_parity(bool __parity) const
     {

libcxx/include/barrier

@@ -362,6 +362,12 @@ public:
     __barrier_base(__barrier_base const&) = delete;
     __barrier_base& operator=(__barrier_base const&) = delete;
 
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool __try_wait(arrival_token __old) const
+    {
+        return __try_wait_phase(__old & __phase_bit);
+    }
+
     _LIBCUDACXX_INLINE_VISIBILITY
     bool try_wait_parity(bool __parity) const
     {
@@ -370,7 +376,7 @@ public:
     _LIBCUDACXX_INLINE_VISIBILITY
     bool try_wait(arrival_token __old) const
     {
-        return __try_wait_phase(__old & __phase_bit);
+        return __try_wait(__old);
     }
 
     _LIBCUDACXX_NODISCARD_ATTRIBUTE inline _LIBCUDACXX_INLINE_VISIBILITY