[Patch RFC IA64] Fixing predication & autoinc interaction

Alexander Monakov amonakov@ispras.ru
Fri Sep 25 14:39:00 GMT 2009



On Fri, 25 Sep 2009, Steven Bosscher wrote:

> On Fri, Sep 25, 2009 at 3:04 PM, Alexander Monakov <amonakov@ispras.ru> wrote:
>> If I understand the situation correctly, the problem is not in the machine
>> description, but rather in the optimization passes (combine in this case).
>
> Or the fact that we have a COND_EXEC before reload?  No passes before
> reload expect to see COND_EXEC, so I'm surprised that you have a
> COND_EXEC in combine.

Oh, sorry, I should have been more specific.  The combiner combines the 
following to patterns:

(set (reg <r1>) (mem (post_inc ...)))
(set (reg <r3>) (if_then_else (reg <r1>) (reg <r2>)))

So the conditional code is wrapped into IF_THEN_ELSE, not COND_EXEC.


>
> Ciao!
> Steven
>



More information about the Gcc-patches mailing list