Re: [patch] A fix for PR40386

[Vladimir Makarov <vmakarov at redhat dot com>]

 On 09/08/2010 04:20 PM, Bernd Schmidt wrote:
> On 09/08/2010 09:21 PM, Vladimir Makarov wrote:
> >   The following patch fixes PR40386.  The problem in details is described
> > on http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40386.
> According to that,
>
> > The problem is in that pseudos (r121 in our case) spilled by IRA are
> > not added to live_throughout of reload chain.
> Is there a good reason to do this, or should we change that instead?
We should add all pseudos (I guess it was my typo).  The patch actually 
results in adding all pseudos (spilled or not) to live_throughout 
independently from value of flag_ira_share_spill_slots.  For the old RA 
it was done only for non-spilled pseudos.  IRA needs this because it can 
assign hard registers in reload to pseudos spilled by IRA which global.c 
never did.