7c3fdcc276
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.
32 lines
1.3 KiB
Plaintext
32 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: [[META0:![0-9]+]] = !{ptr @_Z6kernel1Sii, !"kernel", i32 1, !"grid_constant", [[META1:![0-9]+]]}
|
|
// CHECK: [[META1]] = !{i32 1, i32 3}
|
|
// CHECK: [[META2:![0-9]+]] = !{ptr @_Z13tkernel_constIK1SEvT_, !"kernel", i32 1, !"grid_constant", [[META3:![0-9]+]]}
|
|
// CHECK: [[META3]] = !{i32 1}
|
|
// CHECK: [[META4:![0-9]+]] = !{ptr @_Z13tkernel_constI1SEvT_, !"kernel", i32 1, !"grid_constant", [[META3]]}
|
|
// CHECK: [[META5:![0-9]+]] = !{ptr @_Z7tkernelIK1SEviT_, !"kernel", i32 1, !"grid_constant", [[META6:![0-9]+]]}
|
|
// CHECK: [[META6]] = !{i32 2}
|
|
//.
|