This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [tree-ssa] Patch ping
> "Zack Weinberg" <zack@codesourcery.com> writes:
>
> > Representing control flow only via CFG, during RTL expansion and
> > beyond - all the way until final, in fact - would get rid of a number
> > of issues with RTL. (*cough* tablejumps)
>
> I'm pretty sure this wouldn't solve tablejumps. You still have to
> associate a particular case value with a label.
The cfglayout mode kind-of solves tablejumps.
The major evil is that ADDR_VEC tables ends up in between basic blocks
after tablejump is elliminated. This is moved away from instruction
stream and if we didn't need to re-insert it later, we would live all
happy :)
I hope to send out code I have for discussion really soon (only major
stopper is the reverted tail call removal patch that will hopefully get
in) and then we can discuss in more specific way.
There are still some hacks in the code (wrt nested functions, broken
fold in gimple and such) and I don't feel 100% motivated to do the hard
work on it before we decide whether we want it or not, but the concept
works, bootstrap and seems to be robust enought to catch many latent
bugs and not introduce too many of new.
Honza
>
> --
> - Geoffrey Keating <geoffk@geoffk.org>