[patch, fortran] PR 40628, front-end optimization pass

[Thomas Koenig]

Paolo Bonzini wrote:

> You could test HONOR_NANS for that, though I don't know whether modes 
> are already available in the Fortran front-end (failing that, there
> is 
> flag_finite_math_only).

That's right, I'll probably do this in a future patch.

> What is the default definition of equality for derived types, if any? 

There is no intrinsic equality operator for derived types.  You can
define your own, but this will end up as the function actually being
invoked in the gfc_* structures.

	Thomas