A directx specific llvm intrinsic is introduced:
float @llvm.dx.sin.f32(float)
It is for https://github.com/Microsoft/DirectXShaderCompiler/blob/master/docs/DXIL.rst#sin.
This change the llvm intrinsic and test it can be generated from clang codeGen.
The clang builtin change is in https://reviews.llvm.org/D124593