[llvm] [llubi] Upstream existing floating-point intrinsics (PR #196034)

[Yingwei Zheng via llvm-commits]

================
@@ -84,6 +84,10 @@ static cl::opt<bool>
                   cl::desc("Disable interpreter-introduced non-determinism."),
                   cl::init(false), cl::cat(InterpreterCategory));
 
+static cl::opt<bool> FuseFMulAdd("fuse-fmuladd",
----------------
dtcxzyw wrote:

`-fuse-fmuladd=false` is not tested.

https://github.com/llvm/llvm-project/pull/196034