llvm-project/clang/test/CodeGenCUDA/grid-constant.cu
Alex MacLean 06bcc34e3d
[NVPTX] Auto-upgrade nvvm.grid_constant to param attribute (#155489)
Upgrade the !"grid_constant" !nvvm.annotation to a "nvvm.grid_constant"
attribute. This attribute is much simpler for front-ends to apply and
faster and simpler to query.
2025-08-27 16:32:28 -07:00

28 lines
1.3 KiB
Plaintext

// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --check-globals all --version 5
// RUN: %clang_cc1 "-triple" "nvptx64-nvidia-cuda" -emit-llvm -fcuda-is-device -o - %s | FileCheck %s
#include "Inputs/cuda.h"
struct S {};
__global__ void kernel(__grid_constant__ const S gc_arg1, int arg2, __grid_constant__ const int gc_arg3) {}
// dependent arguments get diagnosed after instantiation.
template <typename T>
__global__ void tkernel_const(__grid_constant__ const T arg) {}
template <typename T>
__global__ void tkernel(int dummy, __grid_constant__ T arg) {}
void foo() {
tkernel_const<const S><<<1,1>>>({});
tkernel_const<S><<<1,1>>>({});
tkernel<const S><<<1,1>>>(1, {});
}
// CHECK: define dso_local ptx_kernel void @_Z6kernel1Sii(ptr noundef byval(%struct.S) align 1 "nvvm.grid_constant" %gc_arg1, i32 noundef %arg2, i32 noundef "nvvm.grid_constant" %gc_arg3)
// CHECK: define ptx_kernel void @_Z13tkernel_constIK1SEvT_(ptr noundef byval(%struct.S) align 1 "nvvm.grid_constant" %arg)
// CHECK: define ptx_kernel void @_Z13tkernel_constI1SEvT_(ptr noundef byval(%struct.S) align 1 "nvvm.grid_constant" %arg)
// CHECK: define ptx_kernel void @_Z7tkernelIK1SEviT_(i32 noundef %dummy, ptr noundef byval(%struct.S) align 1 "nvvm.grid_constant" %arg)