On Thursday, August 19, 2021 12:28 PM, Michael Paquier
wrote
>On Wed, Aug 18, 2021 at 10:04:04AM +0900, Michael Paquier wrote:
>> Yes, that does not seem wise on performance grounds. The case of
>> !zero_is_valid is never reached, so it seems like this code was just a
>> copy-paste from the floa
On Wed, Aug 18, 2021 at 10:04:04AM +0900, Michael Paquier wrote:
> Yes, that does not seem wise on performance grounds. The case of
> !zero_is_valid is never reached, so it seems like this code was just a
> copy-paste from the float code in the backend. Your patch looks right
> to me.
Applied.
-
On Tue, Aug 17, 2021 at 11:08:59AM +, tanghy.f...@fujitsu.com wrote:
> On Friday, August 6, 2021 11:14 PM, tanghy.f...@fujitsu.com
> wrote:
>> Commit 6bf0bc842 replaced float.c's CHECKFLOATVAL() macro with static inline
>> subroutines,
>> but the fix introduced a performance regression as T
On Friday, August 6, 2021 11:14 PM, tanghy.f...@fujitsu.com
wrote:
>Commit 6bf0bc842 replaced float.c's CHECKFLOATVAL() macro with static inline
>subroutines,
>but the fix introduced a performance regression as Tom Lane pointed out and
>fixed at 607f8ce74.
>
>Found obsolete CHECKFLOATVAL usage