[llvm] Emit float128 libcalls for some F128 variants (PR #79611)

[Pranav Kant via llvm-commits]

pranavk wrote:

@MaskRay clang codegens correct LLVM IR. So I don't think we need any more tests in clang/test/CodeGen/. All the fp128 __builtin tests we have there already check for generation of appropriate fp128 LLVM IR calls (https://github.com/llvm/llvm-project/blob/3d71e4166de81bc3b86d127d9ac6607bda2b2755/clang/test/CodeGen/X86/math-builtins.c)

The problem is in lowering of LLVM IR into machine code which is what this patch is trying to fix for X86_64. I have changed the tests to check for that lowering.

https://github.com/llvm/llvm-project/pull/79611