[clang] [llvm] Intrinsic: introduce minimumnum and maximumnum (PR #93841)

[YunQiang Su via cfe-commits]

================
@@ -16049,6 +16094,84 @@ of the two arguments. -0.0 is considered to be less than +0.0 for this
 intrinsic. Note that these are the semantics specified in the draft of
 IEEE 754-2019.
 
+.. _i_minimumnum:
+
+'``llvm.minimumnum.*``' Intrinsic
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+Syntax:
+"""""""
+
+This is an overloaded intrinsic. You can use ``llvm.minimumnum`` on any
+floating-point or vector of floating-point type. Not all targets support
+all types however.
+
+::
+
+      declare float     @llvm.minimumnum.f32(float %Val0, float %Val1)
+      declare double    @llvm.minimumnum.f64(double %Val0, double %Val1)
+      declare x86_fp80  @llvm.minimumnum.f80(x86_fp80 %Val0, x86_fp80 %Val1)
+      declare fp128     @llvm.minimumnum.f128(fp128 %Val0, fp128 %Val1)
+      declare ppc_fp128 @llvm.minimumnum.ppcf128(ppc_fp128 %Val0, ppc_fp128 %Val1)
+
+Overview:
+"""""""""
+
+The '``llvm.minimumnum.*``' intrinsics return the minimum of the two
+arguments, not propagating NaNs and treating -0.0 as less than +0.0.
+
+
+Arguments:
+""""""""""
+
+The arguments and return value are floating-point numbers of the same
+type.
+
+Semantics:
+""""""""""
+If both operands are NaNs, returns qNaN. Otherwise returns the lesser
+of the two arguments. -0.0 is considered to be less than +0.0 for this
+intrinsic. Note that these are the semantics specified in IEEE 754-2019.
----------------
wzssyqa wrote:

> Needs to spell out the signaling nan behavior. If we're fixing minnum's snan behavior to match IEEE, this is identical except with the stronger guarantee for signed zero ordering. The documentation should also explicitly state this is the only difference, to help reduce confusion. Alternatively, we could add an immediate bool parameter to minnum/maxnum for whether the ordering of 0 is guaranteed
> 

OK. I will add more details here.

> I hate the naming mess we've ended up with here, but I guess C23 has damned us. If you're going to match the C23 names, this should be `llvm.minimum.num` with an extra _

In fact, in my initial version of patch, I use _, then I get `llvm.minimum.num`. I considered it as a problem, since it cost some time to know in fact I should use `llvm.minimum.num` instead of `llvm.minimum_num` :(

https://github.com/llvm/llvm-project/pull/93841