[HIP-Clang] Fat binary should not be produced for non GPU code

[ashi1]

Summary

Do not produce Fat binary functions for HIP when no device code is present.

[yaxunl]

LGTM. Thanks!

[tra]

General nit: please use diffs with very large context when you submit patches with Phabricator.
https://llvm.org/docs/Phabricator.html#requesting-a-review-via-the-web-interface

lib/CodeGen/CGCUDANV.cpp

475–476

I think this would make sense for CUDA, too.

[yaxunl]

lib/CodeGen/CGCUDANV.cpp

475–476

https://github.com/llvm-mirror/clang/blob/master/lib/CodeGen/CGCUDANV.cpp#L482

CUDA generates a dummy register function call for -fgpu-rdc. So probably only do this when RelocatableDeviceCode is false for CUDA?

[tra]

lib/CodeGen/CGCUDANV.cpp

475–476

Good point. Off the top of my head I can't tell why -fcudardc needs this. Keeping the change HIP-only is fine.

[ashi1]

Hi Artem, I had just committed the change. IS this change OK or should I revert it?

[tra]

In D60141#1452019, @ashi1 wrote:

Hi Artem, I had just committed the change. IS this change OK or should I revert it?

The if condition could use some clang-formatting, but other than that the patch still looks OK to me.

[ashi1]

Thank you, I submitted a clang-formatted fix for the if statement.

lib/CodeGen/CGCUDANV.cpp

 ///         __hip_register_globals(__hip_gpubin_handle);
 ///     }
 /// }
 /// \endcode
 llvm::Function *CGNVCUDARuntime::makeModuleCtorFunction() {
   bool IsHIP = CGM.getLangOpts().HIP;
   // No need to generate ctors/dtors if there is no GPU binary.
   StringRef CudaGpuBinaryFileName = CGM.getCodeGenOpts().CudaGpuBinaryFileName;
   if (CudaGpuBinaryFileName.empty() && !IsHIP)
     return nullptr;
+  if (IsHIP && EmittedKernels.empty() && DeviceVars.empty())
+    return nullptr;
 
   // void __{cuda|hip}_register_globals(void* handle);
   llvm::Function *RegisterGlobalsFunc = makeRegisterGlobalsFn();
   // We always need a function to pass in as callback. Create a dummy
   // implementation if we don't need to register anything.
   if (RelocatableDeviceCode && !RegisterGlobalsFunc)
     RegisterGlobalsFunc = makeDummyFunction(getRegisterGlobalsFnTy());
 
   // void ** __{cuda|hip}RegisterFatBinary(void *);
   llvm::FunctionCallee RegisterFatbinFunc = CGM.CreateRuntimeFunction(

test/CodeGenCUDA/device-stub.cu

 // HIP: if:
 // HIP-NEXT: call void @__[[PREFIX]]UnregisterFatBinary
 // HIP-NEXT: store i8** null, i8*** @__hip_gpubin_handle
 // HIP-NEXT: br label %exit
 // HIP: exit:
 
 // There should be no __[[PREFIX]]_register_globals if we have no
 // device-side globals, but we still need to register GPU binary.
 // Skip GPU binary string first.
 // CUDANOGLOBALS: @{{.*}} = private constant{{.*}}
-// HIPNOGLOBALS: @{{.*}} = internal constant{{.*}}
+// HIPNOGLOBALS-NOT: @{{.*}} = internal constant{{.*}}
 // NOGLOBALS-NOT: define internal void @__{{.*}}_register_globals
-// NOGLOBALS: define internal void @__[[PREFIX:cuda|hip]]_module_ctor
-// NOGLOBALS: call{{.*}}[[PREFIX]]RegisterFatBinary{{.*}}__[[PREFIX]]_fatbin_wrapper
+// CUDANOGLOBALS: define internal void @__[[PREFIX:cuda|hip]]_module_ctor
+// CUDANOGLOBALS: call{{.*}}[[PREFIX]]RegisterFatBinary{{.*}}__[[PREFIX]]_fatbin_wrapper
 // NOGLOBALS-NOT: call void @__[[PREFIX]]_register_globals
-// NOGLOBALS: define internal void @__[[PREFIX]]_module_dtor
-// NOGLOBALS: call void @__[[PREFIX]]UnregisterFatBinary
+// CUDANOGLOBALS: define internal void @__[[PREFIX]]_module_dtor
+// CUDANOGLOBALS: call void @__[[PREFIX]]UnregisterFatBinary
+
+// There should be no fat binary functions when no device-code is found for HIP.
+// HIPNOGLOBALS-NOT: define internal void @__[[PREFIX:cuda|hip]]_module_ctor
+// HIPNOGLOBALS-NOT: call{{.*}}[[PREFIX]]RegisterFatBinary{{.*}}__[[PREFIX]]_fatbin_wrapper
+// HIPNOGLOBALS-NOT: define internal void @__[[PREFIX]]_module_dtor
+// HIPNOGLOBALS-NOT: call void @__[[PREFIX]]UnregisterFatBinary
+
 
 // There should be no constructors/destructors if we have no GPU binary.
 // NOGPUBIN-NOT: define internal void @__[[PREFIX]]_register_globals
 // NOGPUBIN-NOT: define internal void @__[[PREFIX]]_module_ctor
 // NOGPUBIN-NOT: define internal void @__[[PREFIX]]_module_dtor