Fix GCC 9.4 build: gate std::atomic<shared_ptr> on C++20 feature macro

NAM/wavenet/slimmable.cpp

@@ -291,45 +291,45 @@ bool is_full_size(const std::vector<wavenet::LayerArrayParams>& params, const st
 
 } // anonymous namespace
 
-#ifdef _LIBCPP_VERSION
+#if NAM_HAS_ATOMIC_SHARED_PTR
 void SlimmableWavenet::_pending_clear_release()
 {
-  std::atomic_store_explicit(&_pending_staged, std::shared_ptr<StagedSlimModel>{}, std::memory_order_release);
+  _pending_staged.store({}, std::memory_order_release);
 }
 
 std::shared_ptr<SlimmableWavenet::StagedSlimModel> SlimmableWavenet::_pending_load_acquire() const
 {
-  return std::atomic_load_explicit(&_pending_staged, std::memory_order_acquire);
+  return _pending_staged.load(std::memory_order_acquire);
 }
 
 void SlimmableWavenet::_pending_store_release(std::shared_ptr<StagedSlimModel> p)
 {
-  std::atomic_store_explicit(&_pending_staged, std::move(p), std::memory_order_release);
+  _pending_staged.store(std::move(p), std::memory_order_release);
 }
 
 std::shared_ptr<SlimmableWavenet::StagedSlimModel> SlimmableWavenet::_pending_exchange_take_acq_rel()
 {
-  return std::atomic_exchange_explicit(&_pending_staged, std::shared_ptr<StagedSlimModel>{}, std::memory_order_acq_rel);
+  return _pending_staged.exchange({}, std::memory_order_acq_rel);
 }
 #else
 void SlimmableWavenet::_pending_clear_release()
 {
-  _pending_staged.store({}, std::memory_order_release);
+  std::atomic_store_explicit(&_pending_staged, std::shared_ptr<StagedSlimModel>{}, std::memory_order_release);
 }
 
 std::shared_ptr<SlimmableWavenet::StagedSlimModel> SlimmableWavenet::_pending_load_acquire() const
 {
-  return _pending_staged.load(std::memory_order_acquire);
+  return std::atomic_load_explicit(&_pending_staged, std::memory_order_acquire);
 }
 
 void SlimmableWavenet::_pending_store_release(std::shared_ptr<StagedSlimModel> p)
 {
-  _pending_staged.store(std::move(p), std::memory_order_release);
+  std::atomic_store_explicit(&_pending_staged, std::move(p), std::memory_order_release);
 }
 
 std::shared_ptr<SlimmableWavenet::StagedSlimModel> SlimmableWavenet::_pending_exchange_take_acq_rel()
 {
-  return _pending_staged.exchange({}, std::memory_order_acq_rel);
+  return std::atomic_exchange_explicit(&_pending_staged, std::shared_ptr<StagedSlimModel>{}, std::memory_order_acq_rel);
 }
 #endif
 

NAM/wavenet/slimmable.h

@@ -3,9 +3,19 @@
 #include <memory>
 #include <vector>
 
-#ifdef _LIBCPP_VERSION
-// libc++: std::atomic<std::shared_ptr<T>> is not viable; staging uses deprecated atomic_* free functions.
+// std::atomic<std::shared_ptr<T>> requires C++20 library support (libstdc++ >= GCC 12).
+// Where it is unavailable -- libc++ (any version so far) and older libstdc++ such as the
+// one shipped with GCC 9 -- fall back to the deprecated std::atomic_* free-function
+// overloads for shared_ptr, which provide the same acquire/release semantics. Keyed on the
+// C++20 feature-test macro rather than on a specific standard library so the right path is
+// chosen for every compiler.
+#if defined(__cpp_lib_atomic_shared_ptr) && __cpp_lib_atomic_shared_ptr >= 201711L
+  #define NAM_HAS_ATOMIC_SHARED_PTR 1
 #else
+  #define NAM_HAS_ATOMIC_SHARED_PTR 0
+#endif
+
+#if NAM_HAS_ATOMIC_SHARED_PTR
   #include <atomic>
 #endif
 
@@ -71,11 +81,11 @@ class SlimmableWavenet : public DSP, public SlimmableModel
     std::shared_ptr<DSP> model;
     std::vector<int> channels;
   };
-#ifdef _LIBCPP_VERSION
+#if NAM_HAS_ATOMIC_SHARED_PTR
+  std::atomic<std::shared_ptr<StagedSlimModel>> _pending_staged;
+#else
   /// Staged model; synchronized via deprecated std::atomic_* overloads for shared_ptr only.
   std::shared_ptr<StagedSlimModel> _pending_staged;
-#else
-  std::atomic<std::shared_ptr<StagedSlimModel>> _pending_staged;
 #endif
 
   std::vector<int> _current_channels;