Re: [Jan Hubicka] Memory reduction patches for 3.4, round 1

[Jan Hubicka <jh at suse dot cz>]

> Jan Hubicka wrote:
> 
> >>Jan --
> >>
> >>Zack asked me to review some of your changes for GCC 3.4.  In future, if 
> >>you're looking to make changes to the release branch, please copy me 
> >>directly; I tend to get a ways behind on GCC mail and read it in batch.
> >
> >
> >Thank you!
> >Sorry for forgetting add CC to you,  I plan a followup round to my
> >patches once the situation around cselib isses is stablized.  Did you
> >skipped the two cselib patches (cutting down it's usage in jump
> >threading) for reason or just by accident?
> 
> From your email, it didn't sound like those two were ready for prime 
> time yet; I thought you said they were still stabilizing.  I'd prefer 
> not to consider anything like this for the branch that hasn't been 
> proven on the mainline yet.
I see, I was bit confusing.
I have about 5 or 6 patches for cselib itself that hit latent bug in
schedule-ebbs using released datastructures.  I didn't listed these in
the mail. They are important (causing 50% savings on combine.c), so I
really hope this will get resolved soon.

The two patches I mentioned goes other way around and cuts the cost by
reducing amount of times cselib is used, this is safe.  However I just
noticed that the second patch wasn't approved for mainline yet and
without it the first patch may cause code quality regression (it does
not reproduce on SPECs tought), so perhaps I shall wait until the second
patch is accepted for mainline anyway.

Thanks
Honza
> 
> Thanks,
> 
> -- 
> Mark Mitchell
> CodeSourcery, LLC
> mark@codesourcery.com