Re: [tree-ssa] Organization of tree-ssa optimization passes

[Andrew MacLeod <amacleod at redhat dot com>]

On Wed, 2003-12-03 at 11:15, Diego Novillo wrote:
> On Wed, 2003-12-03 at 11:07, Zdenek Dvorak wrote:
> 
> > I think that the fact that even in the rtl optimization part of the gcc
> > we have created nothing that would not be easy to express by a slightly
> > extended formalism of this type shows that this objection is not so
> > serious.
> > 
> I'm not fundamentally opposed.  I just don't care enough.  I haven't
> found a big problem adding/modifying passes.  Maybe it's just me.  It's
> more of a stylistic thing.  What do others think?
> 
Similarly, I am fairly ambivilent, but I would like to be consistent
with the rtl optimizer.

If we can do it for rtl in mainline, then it ought to be a trivial
extension to do if for the tree optimizer too. I would expect similar
"complications" with the tree optimizer as would be encountered with the
current rtl optimizers.  

My suggestion would be to get this accepted for rtl on mainline, then
we'll follow in the tree optimizer with whatever decision the rest of
the community agrees to for rtl. That way we wont get stuck trying to
extend the definition the first time we try to do something not quite
covered.

Andrew