// RUN: echo -n "GPU binary would be here." > %t // RUN: %clang_cc1 -triple x86_64-linux-gnu -emit-llvm %s \ // RUN: -target-sdk-version=11.0 -fcuda-include-gpubinary %t -o - \ // RUN: | FileCheck %s --check-prefixes CUDA // RUN: %clang_cc1 -triple x86_64-linux-gnu -emit-llvm %s -x hip \ // RUN: -fcuda-include-gpubinary %t -o - \ // RUN: | FileCheck %s --check-prefixes HIP #include "Inputs/cuda.h" template struct S { T t; }; template static __global__ void Kernel(S) {} // For some reason it takes three or more instantiations of Kernel to trigger a // crash during CUDA compilation. auto x = &Kernel; auto y = &Kernel; auto z = &Kernel; // This triggers HIP-specific code path. void func (){ Kernel<<<1,1>>>({1}); } // CUDA-LABEL: @__cuda_register_globals( // CUDA: call i32 @__cudaRegisterFunction(ptr %0, ptr @_ZL21__device_stub__KernelIdEv1SIT_E // CUDA: call i32 @__cudaRegisterFunction(ptr %0, ptr @_ZL21__device_stub__KernelIfEv1SIT_E // CUDA: call i32 @__cudaRegisterFunction(ptr %0, ptr @_ZL21__device_stub__KernelIiEv1SIT_E // CUDA: ret void // HIP-LABEL: @__hip_register_globals( // HIP: call i32 @__hipRegisterFunction(ptr %0, ptr @_ZL6KernelIdEv1SIT_E // HIP: call i32 @__hipRegisterFunction(ptr %0, ptr @_ZL6KernelIfEv1SIT_E // HIP: call i32 @__hipRegisterFunction(ptr %0, ptr @_ZL6KernelIiEv1SIT_E // HIP: ret void