RFA: version 2 patch to fix PR37514

Vladimir Makarov vmakarov@redhat.com
Mon Dec 1 18:12:00 GMT 2008


H.J. Lu wrote:
> On Mon, Dec 1, 2008 at 8:29 AM, Jeff Law <law@redhat.com> wrote:
>   
>> Vladimir Makarov wrote:
>>     
>>> The following patch solves a latent reload bug (in reload inheritance)
>>> triggered by IRA.  It is a second version (less pessimistic in reload
>>> inheritance optimization) of the patch
>>>
>>> The problem is described on
>>>
>>> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37514
>>>
>>>
>>> The patch invalidates reg_last_reload_reg set in previous insns for
>>> INC/DEC if reg_last_reload_reg set is not set for the current insn for some
>>> reasons (e.g. the hard register is used in other insn reloads besides reload
>>> for INC/DEC).
>>>
>>> The patch was tested on SH (with -m4 -ml -O3 -fomit-frame-pointer) and
>>> successfully bootstrapped on itanium (another port using INC/DEC heavily).
>>>
>>>
>>> Ok to commit?
>>>
>>> 2008-11-25  Vladimir Makarov  <vmakarov@redhat.com>
>>>
>>>   PR rtl-optimization/37514
>>>   * reload1.c (reload_as_needed): Invalidate reg_last_reload
>>>   from previous insns.
>>>
>>>
>>>       
>> This is fine.  However, at this stage we should be more focused on the
>> simpler, safer fix rather than picking up these uncommon
>> micro-optimizations.
>> Jeff
>>
>>     
>
> I need this patch.  Otherwise, bootstrap will fail on Linux/x86 with:
>
>   
Thanks, H.J.   I'll add it to my patch when  I'll be committing it to 
the trunk.



More information about the Gcc-patches mailing list