On 2017-05-12 09:40:01 -0700, Josh Berkus wrote:
> Josh, Justin, Andres, Gunnar, Peter:
>
> Replying in one long-ass email:
>
> On 05/11/2017 12:11 PM, Andres Freund wrote:
> > On 2017-05-11 12:08:04 -0700, Josh Berkus wrote:
> >> This has been updated with fixes per Andres, Debra, and Petr.
> >>
> >> Please look and suggest any additional improvements.
> >
> > There's not JIT compilation in v10 - I wish ;). I tried to point in the
> > right direction with the list of commits, and the relevant slide
> > numbers...
If we had that, I'd put it as a major feature ;)
> Can you suggest a different 1-line description?
It's like four relatively large patches (the last being 8868
insertions(+), 6531 deletions(-)), so it's not necessarily easy to break
it down to a 1-line summary, without loosing meaning. Therefore I'd
suggest something general "Significantly improved performance for
analytics queries", unless we want to break those into multiple ones
("faster hashtable for bitmapscan and aggregation", "faster aggregation
when multiple aggregates are used", "faster expression and filter
evaluation"), but that seems a bit excessive...
> Because that's what I got from your slides. What are we actually
> doing differently?
That's why I'd pointed to specific slides and the commit hashes... Part
of that work is committed in v10, part is not.
- Andres