On 12/13/2017 07:39 AM, Jakub Jelinek wrote:
> Hi!
>
> While the PRED_NEGATIVE_RETURN heuristics generally works quite well, for
> qsort comparison functions and similar, including the planned C++
> spaceship operator <=> where typically negative and positive are
> approximately even it doesn't work that well. This patch is an attempt
> to at least detect some of these cases. It won't catch functions where
> also other values are returned (e.g. a - b or similar), but then it would be
> even harder to make a distinction.
>
> Bootstrapped/regtested on {x86_64,i686,powerpc64le}-linux, regtest on
> powerpc64-linux pending. Honza, if it doesn't look completely bogus to you,
> could you give it a spin on SPEC (which I don't have easy access to)?
Note you can have access to SPEC. Red Hat is appropriately licensed.
Contact Vlad to get suitable bits.
>
> 2017-12-13 Jakub Jelinek <ja...@redhat.com>
>
> PR middle-end/81914
> * predict.c (zero_one_minusone): New function.
> (apply_return_prediction): Avoid return prediction for functions
> returning only -1, 0 and 1 values, unless they only return -1 and 0
> or 0 and 1.
Seems reasonable to me. You call on how long to wait for Honza.
Jeff
