[flang-commits] [flang] [flang][OpenMP] Use maxnum/minnum for lowering of max/min reduction operators (PR #89258)

[Kiran Chandramohan via flang-commits]

kiranchandramohan wrote:

> > > > If this fixes #87955 then can you add a fixes tag?
> > > 
> > > 
> > > This works for me on AArch64. I think it lowers the 128 version early to compare and select. Thanks for the fix.
> > 
> > 
> > Sorry, I got that wrong. For #87955 it is a min reduction and the backend can convert minnum to a `fminl` call hence it works.
> 
> Okay, so does this PR fix anything?

Yes it fixes https://github.com/llvm/llvm-project/issues/87955 

https://github.com/llvm/llvm-project/pull/89258