[tree-ssa] Fix problem with loop notes emitted by targets

[Pop Sébastian]

On Wed, Dec 17, 2003 at 11:39:11AM +0100, Zdenek Dvorak wrote:
> 
> this takes me to my favourite point: what about the loop optimizer?
> Tree-ssa cfg stuff is now in reasonable shape, so I would like to begin
> with getting rid of the old one. This obviously is a work for a
> subbranch, since it will get some time to make it work without
> regressions.  Pop Sébastian will probably be creating a LNO branch for
> his loop optimization related stuff, right?  We could either use the
> same branch for both, or use separated branches.
> 
> I would slightly prefer the later, since it would make evaluation of the
> changes easier and also the works seem to be reasonably independent -- I
> plan to focus on replacing the rtl stuff and on creating means to
> communicate results of tree-level analyses to the rtl level, while I

This will be useful for the dependences analysis to transmit the
results to the scheduling pass.

> assume Pop's work will be more tree-level oriented.  Of course the
> drawback is the chance of duplicating some stuff & creating conflicts.
> 
It is possible to pick and adapt patches from one subbranch to the other, 
so I'm fine with both solutions, creating or not the LNO branch.