Re: In PG12, query with float calculations is slower than PG11

Hi,

On 2020-02-12 14:18:30 -0500, Tom Lane wrote:
> Andres Freund <andres(at)anarazel(dot)de> writes:
> > I do wonder if we're just punching ourselves in the face with the
> > signature of these checks. Part of the problem here really comes from
> > using the same function to handle a number of different checks.
>
> Yeah, I've thought that too. It's *far* from clear that this thing
> is a win at all, other than your point about the number of copies of
> the ereport call. It's bulky, it's hard to optimize, and I have
> never thought it was more readable than the direct tests it replaced.
>
> > For most places it'd probably end up being easier to read and to
> > optimize if we just wrote them as
> > if (unlikely(isinf(result)) && !isinf(arg))
> > float_overflow_error();
> > and when needed added a
> > else if (unlikely(result == 0) && arg1 != 0.0)
> > float_underflow_error();
>
> +1

Cool. Emre, any chance you could write a patch along those lines?

I'm inclined that we should backpatch that, and just leave the inline
function (without in core callers) in place in 12?

Greetings,

Andres Freund