[llvm] Intrinsic: Fix minnum and introduce minimumnum (PR #93373)

[YunQiang Su via llvm-commits]

wzssyqa wrote:

> Needs LangRef update and an RFC.
> 
> I only briefly looked at this, but I think we can probably get away with just changing minnum to follow IEEE behavior and then make use of the following general rule about our FP operations:
> 

RISC-V/Hexagon has  IEEE754-2019 `minimumNUMBER` operation only.
Such architectures may get benefits with `ISD::FMINIMUMNUM`.

> > Floating-point math operations are allowed to treat all NaNs as if they were quiet NaNs. For example, “pow(1.0, SNaN)” may be simplified to 1.0.
> 
> The effective outcome of this would be that both behaviors become valid. The sNaN -> qNaN case would be the plain intrinsic definition and the sNaN -> other operand case would be via sNaN -> qNaN relaxation.
> 
> Only the constrained minnum intrinsics would require precise sNaN/qNaN distinction, as usual.
> 

My major concern is C code like
```C
double fmin(double, double);
double f(double a, double b) {
        return fmin(a, b);
}
```
This code may expect the same behavior as libc's fmin(3), while we convert it to `llvm.minnum.f64`.
It may break something.

> I think this would be in the spirit of the general FP handling in LLVM.



https://github.com/llvm/llvm-project/pull/93373