[PATCH] D48613: [CUDA] Use atexit() to call module destructor.

[Artem Belevich via Phabricator via cfe-commits]

tra updated this revision to Diff 153138.
tra added a comment.

Make destructor registration conditional -- we generate no destructor with -fcuda-rdc.
Updated the test to reflect changes in destructor registration.


https://reviews.llvm.org/D48613

Files:
  clang/lib/CodeGen/CGCUDANV.cpp
  clang/lib/CodeGen/CodeGenModule.cpp
  clang/test/CodeGenCUDA/device-stub.cu


Index: clang/test/CodeGenCUDA/device-stub.cu
===================================================================
--- clang/test/CodeGenCUDA/device-stub.cu
+++ clang/test/CodeGenCUDA/device-stub.cu
@@ -86,8 +86,6 @@
 // HIPRDC-SAME: c"[[MODULE_ID:.+]]\00", section "__hip_module_id", align 32
 // * Make sure our constructor was added to global ctor list.
 // ALL: @llvm.global_ctors = appending global {{.*}}@__[[PREFIX]]_module_ctor
-// * In separate mode we also register a destructor.
-// NORDC: @llvm.global_dtors = appending global {{.*}}@__[[PREFIX]]_module_dtor
 // * Alias to global symbol containing the NVModuleID.
 // RDC: @__fatbinwrap[[MODULE_ID]] = alias { i32, i32, i8*, i8* }
 // RDC-SAME: { i32, i32, i8*, i8* }* @__[[PREFIX]]_fatbin_wrapper
@@ -127,6 +125,8 @@
 // NORDC-NEXT: store{{.*}}__[[PREFIX]]_gpubin_handle
 //   .. and then calls __[[PREFIX]]_register_globals
 // NORDC-NEXT: call void @__[[PREFIX]]_register_globals
+// * In separate mode we also register a destructor.
+// NORDC-NEXT: call i32 @atexit(void (i8*)* @__[[PREFIX]]_module_dtor)
 
 // With relocatable device code we call __[[PREFIX]]RegisterLinkedBinary%NVModuleID%
 // RDC: call{{.*}}__[[PREFIX]]RegisterLinkedBinary[[MODULE_ID]](
Index: clang/lib/CodeGen/CodeGenModule.cpp
===================================================================
--- clang/lib/CodeGen/CodeGenModule.cpp
+++ clang/lib/CodeGen/CodeGenModule.cpp
@@ -404,10 +404,9 @@
       AddGlobalCtor(ObjCInitFunction);
   if (Context.getLangOpts().CUDA && !Context.getLangOpts().CUDAIsDevice &&
       CUDARuntime) {
-    if (llvm::Function *CudaCtorFunction = CUDARuntime->makeModuleCtorFunction())
+    if (llvm::Function *CudaCtorFunction =
+            CUDARuntime->makeModuleCtorFunction())
       AddGlobalCtor(CudaCtorFunction);
-    if (llvm::Function *CudaDtorFunction = CUDARuntime->makeModuleDtorFunction())
-      AddGlobalDtor(CudaDtorFunction);
   }
   if (OpenMPRuntime) {
     if (llvm::Function *OpenMPRegistrationFunction =
Index: clang/lib/CodeGen/CGCUDANV.cpp
===================================================================
--- clang/lib/CodeGen/CGCUDANV.cpp
+++ clang/lib/CodeGen/CGCUDANV.cpp
@@ -472,6 +472,19 @@
     CtorBuilder.CreateCall(RegisterLinkedBinaryFunc, Args);
   }
 
+  // Create destructor and register it with atexit() the way NVCC does it. Doing
+  // it during regular destructor phase worked in CUDA before 9.2 but results in
+  // double-free in 9.2.
+  if (llvm::Function *CleanupFn = makeModuleDtorFunction()) {
+    // extern "C" int atexit(void (*f)(void));
+    llvm::FunctionType *AtExitTy =
+        llvm::FunctionType::get(IntTy, CleanupFn->getType(), false);
+    llvm::Constant *AtExitFunc =
+        CGM.CreateRuntimeFunction(AtExitTy, "atexit", llvm::AttributeList(),
+                                  /*Local=*/true);
+    CtorBuilder.CreateCall(AtExitFunc, CleanupFn);
+  }
+
   CtorBuilder.CreateRetVoid();
   return ModuleCtorFunc;
 }


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D48613.153138.patch
Type: text/x-patch
Size: 2955 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20180627/cecb4f6a/attachment.bin>