[clang] [openmp] [llvm] [OpenMP] Rework handling of global ctor/dtors in OpenMP (PR #71739)

[Johannes Doerfert via cfe-commits]

================
@@ -2627,6 +2637,38 @@ struct AMDGPUDeviceTy : public GenericDeviceTy, AMDGenericDeviceTy {
   using AMDGPUEventRef = AMDGPUResourceRef<AMDGPUEventTy>;
   using AMDGPUEventManagerTy = GenericDeviceResourceManagerTy<AMDGPUEventRef>;
 
+  /// Common method to invoke a single threaded constructor or destructor
+  /// kernel by name.
+  Error callGlobalCtorDtorCommon(GenericPluginTy &Plugin, DeviceImageTy &Image,
+                                 const char *Name) {
+    // Perform a quick check for the named kernel in the image. The kernel
+    // should be created by the 'amdgpu-lower-ctor-dtor' pass.
+    GenericGlobalHandlerTy &Handler = Plugin.getGlobalHandler();
+    GlobalTy Global(Name, sizeof(void *));
+    if (auto Err = Handler.getGlobalMetadataFromImage(*this, Image, Global)) {
+      consumeError(std::move(Err));
+      return Plugin::success();
+    }
+
+    // Allocate and construct the AMDGPU kernel.
+    AMDGPUKernelTy AMDGPUKernel(Name);
+    if (auto Err = AMDGPUKernel.initImpl(*this, Image))
----------------
jdoerfert wrote:

Generally, we should always call the generic entry points, so, init, not initImpl.
Assuming you have no specific reason not to. Also below for launch.

https://github.com/llvm/llvm-project/pull/71739