Engininja2
Engininja2
`__shfl_xor()` for half2 was added in ROCm 5.6. This PR implements it for HIP versions less than that. Fixes #7242
The current version of CUDA allows you to access the component halfs of half2 through half2.x and half2.y, but in HIP x and y are unsigned shorts and not half...
### Problem Description When using Visual Studio 2022 CMake to compile a project (llama.cpp PR#5966) that uses `enable_language(HIP)` with the HIP 5.7 SDK I get the following error: ``` CMake...