Re: RFA: enable LRA for rs6000 [patch for WRF]

[Vladimir Makarov <vmakarov at redhat dot com>]

On 13-04-22 3:31 PM, Michael Meissner wrote:
> On Mon, Apr 22, 2013 at 03:26:45PM -0400, Vladimir Makarov wrote:
> > On 13-04-22 12:35 AM, Alan Modra wrote:
> > > On Fri, Apr 19, 2013 at 05:16:43PM -0400, Vladimir Makarov wrote:
> > > >    I don't understand what this check means and what comments ??? means too.
> > > A lo_sum mem is only valid if you know it won't be offset (or that
> > > offsetting will never cross a 64k+32k boundary).  If the access is
> > > smaller than a word then the load or store can be done in one insn.
> > > No offset required.  If the access is a DFmode *and* you are loading
> > > or storing a floating point reg, then the access is also done in one
> > > insn.  The ??? comment is referring to the fact that you don't know
> > > for sure that the DFmode is in a floating point reg.  It usually is,
> > > but may be in two general purpose regs.  Which then need an offset to
> > > load/store the second reg.
> > Alan, thanks for the explanation. I'll search for another solution.
> I'm suspecting secondary_reload needs more tuning for TF/TD modes.
I've fixed dealII crash and commited the patch into the branch as rev. 
198169.

The dealII crash itself can be cured by treatment of lo_sum for LRA the 
same way as for reload (please see code checking modes for addressing 
more one word).  But in this case a few tests fail which is cured in LRA 
itself by trying to load address into pseudo using lo_sum.

The patch was successfully bootstrapped (--with-cpu=power7) and tested 
on GCC testsuite.

2013-04-22  Vladimir Makarov <vmakarov@redhat.com>

        * lra-constraints.c (process_address): Try to put lo_sum into
        register.
        * config/rs6000/rs6000.c (legitimate_lo_sum_address_p): Remove
        lra_in_progress guard for addressing something bigger than word.

Attachment: deal.patch
Description: Text document