This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: [new-regalloc] What's the status on the new allocator?
- To: Peter Bergner <bergner at borg dot umn dot edu>
- Subject: Re: [new-regalloc] What's the status on the new allocator?
- From: Dan Nicolaescu <dann at godzilla dot ICS dot UCI dot EDU>
- Date: Tue, 27 Mar 2001 09:35:37 -0800
- Cc: gcc at gcc dot gnu dot org
You wrote:
> I just joined the mailing list yesterday due to someone telling me
> that work was being done on rewriting gcc's register allocator
> using a Chaitin/Brigg's style graph coloring algorithm. This topic
> is near and dear to me (I developed Interference Region Spilling using
> gcc 2.7.2 for my thesis work) and I was contemplating reimplementing
> it (targeting PPC) to test out some new spill code minimization and
> coalescing ideas I have. However, I see that you've beaten me to it! :-)
>
> Scanning through the mailing list archive, I see that the allocator
> is somewhat implemented. Can someone please give a short status
> update on the new allocator? Also, I haven't seen a detailed list
> of what will be implemented (ie, Iterated Coalescing, Optimistic Coalescing,
> live range spilling,...). I'd appreciate it if someone could describe
> what will be going into the new allocator too. Thanks!
As nobody else answered...
There is a branch in CVS with the tag new-regalloc-branch that
contains the new register allocator. There's a TODO list in the file
ra.c
That branch is known to bootstrap on x86-linux (but you need to
configure with --enable-checking=misc,gc ) and I heard it bootstraps
on ppc-linux.
At http://www.cygnus.com/~dnovillo/SPEC/ you can find SPEC95 results
for the new-regalloc-branch ...
Hope this helps.