Closed dfukalov closed 9 months ago
Just directly call the underlying intrinsics, there's no reason to go through these wrappers (and a lot of extra headaches by doing so)
Is this still needed?
If still needed, move this PR to https://github.com/ROCm/llvm-project/tree/amd-staging/amd/device-libs
The library is needed in https://github.com/ROCmSoftwarePlatform/triton as a wrapper for CUDA library functions.