On Fri, Jun 21, 2013 at 5:19 PM, Jakub Jelinek <ja...@redhat.com> wrote:
> On Fri, Jun 21, 2013 at 05:11:39PM +0200, Uros Bizjak wrote:
>> > Is it ok to install?
>>
>> This is OK for mainline.
>>
>> BTW: There are many other single-underscore prefixed intrinsics [1]
>> besides _bextr_*. Perhaps you should also add these to the intrinsics
>> header, then the complete header could be backported to other release
>> branches.
>
> I actually don't understand how this can work, bmi_bextr_{si,di} expanders
> have just 3 operands (one target, 2 arguments), so just by giving it
> 4 operands instead just means the last one is dropped on the floor.
> Why do you need a builtin for this at all?
> I was expecting that _bextr_u{32,64} would be implemented either using
> __bextr_u{32,64} or at least using it's underlying builtin, by constructing
> the combined len/start argument with shifts/ands first.
>
> But I admit I haven't applied the patch and looked how it works.OK, I assumed that the patch was tested according to established standards, and that it doesn't need to be reviewed for its most basic functionality. If the patch was not tested appropriately, I will simply ignore future submissions from the submitters that want to bend the rules. Sorry. Uros.
