[llvm] [RFC][SPIR-V] Add intrinsics to convert to/from ap.float (PR #164252)

[Matt Arsenault via llvm-commits]

================

----------------
arsenm wrote:

Why would this be challenging? Isn't it a matter of just shuffling the bits into the right positions, and then doing a rounding bit adjustment? 

https://github.com/llvm/llvm-project/pull/164252