diff --git a/llvm/lib/Target/AMDGPU/AMDGPUCtorDtorLowering.cpp b/llvm/lib/Target/AMDGPU/AMDGPUCtorDtorLowering.cpp
--- a/llvm/lib/Target/AMDGPU/AMDGPUCtorDtorLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUCtorDtorLowering.cpp
@@ -52,8 +52,8 @@
   bool createInitOrFiniKernel(Module &M, GlobalVariable *GV, bool IsCtor) {
     if (!GV)
       return false;
-    ConstantArray *GA = cast<ConstantArray>(GV->getInitializer());
-    if (GA->getNumOperands() == 0)
+    ConstantArray *GA = dyn_cast<ConstantArray>(GV->getInitializer());
+    if (!GA || GA->getNumOperands() == 0)
       return false;
     Function *InitOrFiniKernel = createInitOrFiniKernelFunction(M, IsCtor);
     IRBuilder<> IRB(InitOrFiniKernel->getEntryBlock().getTerminator());
diff --git a/llvm/test/CodeGen/AMDGPU/lower-empty-ctor-dtor.ll b/llvm/test/CodeGen/AMDGPU/lower-empty-ctor-dtor.ll
new file mode 100644
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/lower-empty-ctor-dtor.ll
@@ -0,0 +1,9 @@
+; RUN: opt -S -mtriple=amdgcn--  -amdgpu-lower-ctor-dtor < %s | FileCheck %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx700 -filetype=obj -o - < %s | llvm-readelf -s - 2>&1 | FileCheck %s
+
+@llvm.global_ctors = appending global [0 x { i32, void ()*, i8* }] zeroinitializer
+@llvm.global_dtors = appending global [0 x { i32, void ()*, i8* }] zeroinitializer
+
+; No amdgpu_kernels emitted for empty global_ctors
+; CHECK-NOT: amdgcn.device.init
+; CHECK-NOT: amdgcn.device.fini