Compilation time has more than doubled on some Polyhedron tests

[Tobias Schlüter]

Jack Howarth wrote:
> Tobi,
>    On a slightly different but related topic, what happened to the patch
> for having the scalarizer generate vectorizable loops?

Well, I don't know why you singled me out to ask me, but ...

> http://gcc.gnu.org/ml/fortran/2006-01/msg00080.html
> 
> I haven't seen it appear in gcc 4.1 branch yet. Although I must admit I
> was a little confused about why it was being discussed and proposed in
> the gfortran mailing list when the changes and tests seemed to be in the
> gcc compiler code.

... the question bears the answer ;-)  The patch in question is outside the
Fortran FE, and outside the Fortran FE the policy is to only allow regression
fixes on release branches.  Hence, it won't appear on the 4.1 branch.

This policy is in place in order to not destabilize release branches, only in
the Fortran frontend are we more lenient, for its quality is not part of the
release criteria, anyway.

(The thread was CC'ed to the fortran list, as it originated there.)

- Tobi