| Issue |
114048
|
| Summary |
Error when compiling CUDA code with libc++ and C++ standard 26
|
| Labels |
libc++
|
| Assignees |
|
| Reporter |
OgnianM
|
The error:
```
/usr/local/lib/clang/20/include/cuda_wrappers/new:94:25: error: __device__ function 'operator new' cannot overload __host__ __device__ function 'operator new'
94 | __device__ inline void *operator new(__SIZE_TYPE__, void *__ptr) CUDA_NOEXCEPT {
| ^
/usr/local/bin/../include/c++/v1/new:252:1: note: previous declaration is here
252 | operator new(std::size_t, void* __p) _NOEXCEPT {
| ^
```
Looks like it's caused by the `_LIBCPP_CONSTEXPR_SINCE_CXX26` macro making `operator new(std::size_t, void* __p)` implicitly a __host__ __device__ function when CXX >= 26.
The simplest fix would be wrapping the device overrides for placement new in `cuda_wrappers/new` in an `#if` like so
```
// Device overrides for placement new and delete.
#if _LIBCPP_STD_VER < 26
__device__ inline void *operator new(__SIZE_TYPE__, void *__ptr) CUDA_NOEXCEPT {
return __ptr;
}
__device__ inline void *operator new[](__SIZE_TYPE__, void *__ptr) CUDA_NOEXCEPT {
return __ptr;
}
#endif
```
![]()
_______________________________________________
llvm-bugs mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs
