[RFC][PATCH,reload] PR middle-end/42431, fix latent bug in gen_reload_chain_without_interm_reg_p()

[Richard Guenther]

On Tue, Mar 2, 2010 at 1:51 AM, Jeff Law <law@redhat.com> wrote:
> On 03/01/10 10:02, Peter Bergner wrote:
>>
>> On Mon, 2010-03-01 at 09:45 -0700, Jeff Law wrote:
>>
>>>
>>> Yea.  I'm going to twiddle the patch to avoid the copy_rtx completely
>>> and add a suitable comment.  If you could bootstrap it on ppc after I
>>> post it, it'd be greatly appreciated.
>>>
>>
>> Will do.
>>
>>
>
> Attached is the patch I'm officially submitting.   All that changed from the
> previous iteration was the removal of the copy_rtx call and we free the
> vector when we're done.
>
> Bootstrapped and regression tested on x86_64-linux-gnu.

!   VEC_free (rtx_p , heap, substitute_stack);
!   return result;

not freeing the stack should speed up further calls for the cost of
a very small leak.

Otherwise ok.

Thanks,
Richard.


>
>