[llvm] PowerPC/VSX: Select FMINNUM and FMAXNUM (PR #135739)

[YunQiang Su via llvm-commits]

================
@@ -2731,6 +2731,8 @@ def : Pat<(f32 (fneg (fabs f32:$S))),
                (COPY_TO_REGCLASS $S, VSFRC)), VSSRC))>;
 
 // f32 Min.
+def : Pat<(f32 (fminnum f32:$A, f32:$B)),
+          (f32 FpMinMax.F32Min)>;
----------------
wzssyqa wrote:

I have a try. If we drop FMAXNUM_IEEE/FMINNUM_IEEE now,
`llvm/test/CodeGen/PowerPC/fminimum-fmaximum.ll` will not emit correct instructions, as
`expandFMINIMUM_FMAXIMUM` uses FMINNUM_IEEE/FMAXNUM_IEEE.

https://github.com/llvm/llvm-project/pull/135739