[clang] [llvm] [ARM] enable FENV_ACCESS pragma support for hard-float targets (PR #137101)

[Erik Enikeev via llvm-commits]

Varnike wrote:

- Added remaining tests for f16/32/64 intrinsic.
- Fixed lowering for several f16 ops, which leaded to errors at selection stage (https://godbolt.org/z/G8E9Ksv4P).
- The only untested intrinsics are maximum and minimum. At the moment they also lead to compilation fail (https://godbolt.org/z/cbdx5bj9v, same for llvm.maximum and minimum). Is there support for these intrinsics in the backend and should it be addressed within this patch?

I haven’t addressed FPSCR yet; before proceeding I'd like to hear your opinion on how should it be done.

https://github.com/llvm/llvm-project/pull/137101