[Bug tree-optimization/69556] [6 Regression] forwprop4/match.pd undoing work from recip

["rguenther at suse dot de" <gcc-bugzilla at gcc dot gnu dot org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69556

--- Comment #6 from rguenther at suse dot de <rguenther at suse dot de> ---
On January 29, 2016 10:45:12 PM GMT+01:00, "glisse at gcc dot gnu.org"
<gcc-bugzilla@gcc.gnu.org> wrote:
>https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69556
>
>--- Comment #5 from Marc Glisse <glisse at gcc dot gnu.org> ---
>(In reply to rguenther@suse.de from comment #4)
>> I think there is a misunderstanding.  A replacement is still allowed
>if it
>> is a single operation as that replaces at least one other (the one we
>are
>> simplifying).  This assumes equal cost of course which for divide vs.
>Mult
>> is not the case.  So an explicit && single_use as in the patch below
>is
>> needed.
>
>The number of patterns that have to use an explicit single_use is
>growing,
>maybe we need a syntax like :S for "single_use, and I mean it, not like
>:s".

Heh.  I hoped to avoid this and find some better way to cater to the different
users of the machinery in the gcc 7 timeframe.