llvm-project/clang/test/SemaCUDA/grid-constant.cu
Artem Belevich 7c3fdcc276
[CUDA] Add support for __grid_constant__ attribute (#114589)
LLVM support for the attribute has been implemented already, so it just
plumbs it through to the CUDA front-end.

One notable difference from NVCC is that the attribute can be used
regardless of the targeted GPU. On the older GPUs it will just be
ignored. The attribute is a performance hint, and does not warrant a
hard error if compiler can't benefit from it on a particular GPU
variant.
2024-11-05 10:48:54 -08:00

34 lines
1.5 KiB
Plaintext

// RUN: %clang_cc1 -fsyntax-only -verify %s
// RUN: %clang_cc1 -fsyntax-only -fcuda-is-device -verify %s
#include "Inputs/cuda.h"
struct S {};
__global__ void kernel_struct(__grid_constant__ const S arg) {}
__global__ void kernel_scalar(__grid_constant__ const int arg) {}
__global__ void gc_kernel_non_const(__grid_constant__ S arg) {} // expected-error {{__grid_constant__ is only allowed on const-qualified kernel parameters}}
void non_kernel(__grid_constant__ S arg) {} // expected-error {{__grid_constant__ is only allowed on const-qualified kernel parameters}}
// templates w/ non-dependent argument types get diagnosed right
// away, without instantiation.
template <typename T>
__global__ void tkernel_nd_const(__grid_constant__ const S arg, T dummy) {}
template <typename T>
__global__ void tkernel_nd_non_const(__grid_constant__ S arg, T dummy) {} // expected-error {{__grid_constant__ is only allowed on const-qualified kernel parameters}}
// dependent arguments get diagnosed after instantiation.
template <typename T>
__global__ void tkernel_const(__grid_constant__ const T arg) {}
template <typename T>
__global__ void tkernel(__grid_constant__ T arg) {} // expected-error {{__grid_constant__ is only allowed on const-qualified kernel parameters}}
void foo() {
tkernel_const<const S><<<1,1>>>({});
tkernel_const<S><<<1,1>>>({});
tkernel<const S><<<1,1>>>({});
tkernel<S><<<1,1>>>({}); // expected-note {{in instantiation of function template specialization 'tkernel<S>' requested here}}
}