[FORTRAN PATCH] Improved scalarization of constant array constructors

[Paul Thomas]

Jack,
> Paul,
>    Is this patch well defined enough that it might be safe for gcc 4.2
> branch as well? If it provides a significant performance improvement
> that might merit the risk.
>                  Jack
>
>   
If it makes a noticable difference to Polyhedron/SPEC2006 and does not
damage anything on trunk over a ~1 week timescale, yes I would say that
it would be worth it.  I do not think that there is any risk on this one.

Best regards

Paul