This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Graph coloring for register allocation?
- To: Daniel Berlin <dberlin at redhat dot com>
- Subject: Re: Graph coloring for register allocation?
- From: Michael Matz <matzmich at cs dot tu-berlin dot de>
- Date: Wed, 24 Jan 2001 20:21:11 +0100 (MET)
- cc: Zack Weinberg <zackw at stanford dot edu>, gcc at gcc dot gnu dot org
Hi,
On Wed, 24 Jan 2001, Daniel Berlin wrote:
> > SMALL_REGISTER_CLASSES (and therefore all kinds of cruft all over the place).
> > Inability to run the scheduler before register allocation on some machines.
>
> Unless i'm mistaken, all the papers i've looked at(save one or two, i
> think Chow and Hennessy didn't, it's hard to recall) require that
> scheduling be run before register allocation.
There is still no good knowledge how the connection between RA and
scheduling is. Either the papers develop a merged algorithm for RA and
scheduling, or they run scheduling before and after RA. Exclusively
running it before wouldn't make sense, as spill insertions would destroy
(or change) dependence relationship and therefore scheduling assumbptions,
and if there was no spill insertion, running scheduling after RA would
make no difference. So, we would still have sched and sched2.
Ciao,
Michael.