[clang] [Clang] Do not emit intrinsic math functions on GPU targets (PR #98209)

[Matt Arsenault via cfe-commits]

arsenm wrote:

> Interesting, the important bit is that this is done before the LTO link so we know which functions to pull in. Unfortunately I don't think there's a good location that actually states which math calls are supported, but I could probably make a script that goes through each one. Unless @arsenm knows some secret backend location.

The problem is still that there are still multiple half baked mechanisms for telling what calls. RuntimeLibcalls is theoretically the set of what calls what the backend can emit. It's now been moved, but I thought it still doesn't handle the renamed cases in an IR observable way? It's also still unfortunate that we have this parallel mechanism to TargetLibraryInfo, which is almost the same thing.

https://github.com/llvm/llvm-project/pull/98209