Always use cudaGetDriverEntryPoint with CUDA 12 (#2086)

`cudaGetDriverEntryPointByVersion` has been added to drivers in 12.5, but we don't know at compile time the driver version.
In particular, we can build with nvcc 12.8 for a 12.2 driver for instance, and this was causing the following error:

```
undefined symbol: cudaGetDriverEntryPointByVersion,
```
This commit is contained in:
dan_the_3rd
2025-02-11 19:04:25 +01:00
committed by GitHub
parent ad6e1ec19c
commit e9627ce55b

View File

@ -104,8 +104,7 @@ namespace cutlass {
#else // defined(CUTLASS_ENABLE_DIRECT_CUDA_DRIVER_CALL)
#if ((__CUDACC_VER_MAJOR__ >= 13) || \
((__CUDACC_VER_MAJOR__ == 12) && (__CUDACC_VER_MINOR__ >= 5))) \
#if (__CUDACC_VER_MAJOR__ >= 13)
#define CUTLASS_CUDA_DRIVER_WRAPPER_DECL(func, ver) \
template <typename... Args> \