Use thread_local in GetGlobalRuntimeDeviceTracker if possible

It will reduce the cost of getting the thread runtime device tracker,
and will have a better runtime overhead if user constructs a lot of
short lived threads that use VTK-m.

docs/changelog/use-thread_local-in-GetGlobalRuntimeDeviceTracker.md

@@ -0,0 +1,5 @@
+# Use thread_local in GetGlobalRuntimeDeviceTracker function if possible
+
+It will reduce the cost of getting the thread runtime device tracker,
+and will have a better runtime overhead if user constructs a lot of
+short lived threads that use VTK-m.

vtkm/cont/RuntimeDeviceTracker.cxx

@@ -158,6 +158,7 @@ void RuntimeDeviceTracker::ForceDeviceImpl(vtkm::cont::DeviceAdapterId deviceId,
 VTKM_CONT
 vtkm::cont::RuntimeDeviceTracker GetGlobalRuntimeDeviceTracker()
 {
+#if defined(VTKM_CLANG) && (__apple_build_version__ < 8000000)
   static std::mutex mtx;
   static std::map<std::thread::id, vtkm::cont::RuntimeDeviceTracker> globalTrackers;
   std::thread::id this_id = std::this_thread::get_id();
@@ -174,6 +175,9 @@ vtkm::cont::RuntimeDeviceTracker GetGlobalRuntimeDeviceTracker()
     globalTrackers[this_id] = tracker;
     return tracker;
   }
+#else
+  return runtimeDeviceTracker;
+#endif
 }
 }
 } // namespace vtkm::cont

vtkm/cont/RuntimeDeviceTracker.h

@@ -237,6 +237,10 @@ private:
 /// to check over and over again, VTK-m uses per thread runtime device tracker
 /// so that these choices are marked and shared.
 ///
+/// Xcode's clang only supports thread_local from version 8
+#if !(defined(VTKM_CLANG) && (__apple_build_version__ < 8000000))
+thread_local static vtkm::cont::RuntimeDeviceTracker runtimeDeviceTracker;
+#endif
 VTKM_CONT_EXPORT
 VTKM_CONT
 vtkm::cont::RuntimeDeviceTracker GetGlobalRuntimeDeviceTracker();