[clang] Revert "[Clang][AMDGPU] Change __fp16 to _Float16 in builtin definitions" (PR #185861)

[Jan Patrick Lehr via cfe-commits]

jplehr wrote:

> We are moving toward _Float16 for the half type to make its use consistent in downstream code

I read that we will have more such patches coming, is that right?

https://github.com/llvm/llvm-project/pull/185861