New branch for SIMPLE/tree-ssa work

[Jan Hubicka]

> On Thu, Jun 20, 2002 at 09:31:01PM +0200, Jan Hubicka wrote:
> > CFG branch run into this problem as well.  Should I follow same scheme?
> 
> Well, we hope to get most of the stuff from CFG branch out.
> I'd suggest simply not starting major new work on that branch.

OK, but still there is quite imporatnt amount of code concerning the
variable tracking and midlevel RTL, and improved bb reorder pass.
Perhaps the bb reorder pass may go into the tree later in stage 2, as it
is not as major change, but first it needs some polishing and
benchmarking.  Last benchmarking shows degradation in performance, I've
addessed some of the problems but didn't have time to re-try it.
There is also the GCSE work and register coalescing pass.  I plan to
re-do the GCSE bits as they've rotten considerably in the last major
changes to gcse this month.  I am also thinking what to do with register
coalescing pass.  It is definitly more effective than copy propagation
we have right now and helps some other optimizations, but it's effect
can be probably mostly nulled by sane register allocator.

I plan to branch out midlevel RTL branch separately, so this will go
away, but I would like to keep the variable tracking maintained and
continue in the development on loop optimizer.  I can maintain two
branches, even when maintaining cfg branch takes quite considerable
amount of time, but I don't want to maintain more of them, so still I
would like to have branch for loop stuff+variable tracking+register
coalescing+GCSE.

Honza
> 
> 
> r~