PR target/16482: first scheduling pass on SH4

tm_gccmail@kloo.net tm_gccmail@kloo.net
Fri Oct 29 14:36:00 GMT 2004


On Fri, 29 Oct 2004, Kaz Kojima wrote:

> > I'm a bit concerned with this approach.  Consider, for example, a (mem
> > (reg)), in which reload finds this reg to be equivalent to a (plus
> > (reg) (reg)), or a (plus (reg) (const_int BIG)).  Both might end up
> > needing r0, and I don't quite see how you could prevent reload from
> > trying such a substitution.
> 
> Agreed.
> 
> > That said, your patch definitely fixes a real problem.  I'm just
> > concerned that it doesn't fix it completely.  The approach feels a bit
> > like overkill, but I could live with it.  Jörn, any concerns?  Anyone,
> > any further thoughts?
> 
> Jörn has pointed out in #4 of PR description thread that the real
> problem is in lcm.c:optimize_mode_switching.
> 
> Regards,
> 	kaz

Eventually, we should keep track of R0 register pressure separately, and
force instructions which free R0 to be emitted first.

But I guess this needs to stay for now.

Toshi





More information about the Gcc mailing list