This is the mail archive of the
mailing list for the GCC project.
Re: [RFC] Fix PR19401: always completely peel loops
- From: Paul Schlie <schlie at comcast dot net>
- To: Andrew Pinski <pinskia at physics dot uc dot edu>
- Cc: <gcc-patches at gcc dot gnu dot org>
- Date: Thu, 13 Jan 2005 14:22:02 -0500
- Subject: Re: [RFC] Fix PR19401: always completely peel loops
> From: Paul Schlie <firstname.lastname@example.org>
>> From: Andrew Pinski <email@example.com>
>>> On Jan 13, 2005, at 1:51 PM, Paul Schlie wrote:
>>>> Richard Guenther writes:
>>>> This minimal patch unconditionally enables complete loop peeling
>>>> at the tree level.
>>> Isn't it rather presumptuous to assume that such a thing is universally
>>> beneficial to all targets?
>>> (as would expect it to predominantly result in bloated code for smaller
>>> lightly pipelined targets, with little if any tangible benefit; unless
>> Yes, which is why there should be constraint on when this should be
>> done, only do it for -O2/-Os when it reduces the number of
>> instructions (usually also size but at this point we don't
>> know the size) and do it fully at -O3.
> And there are no other optimizations enabled at -O3 which may be desired,
> without unconditionally also fully peeling loops?
Surely some constraint is better than none: like no more than double it's
un-peeled size possibly; as otherwise detrimental effects likely begin to
manifest themselves even on larger machines, as caches may begin to become
needless invalidated etc. with likely negligible local loop performance