Re: Re: Abbreviated keys for Datum tuplesort

On Fri, Feb 20, 2015 at 3:57 PM, Tomas Vondra
<tomas.vondra@2ndquadrant.com> wrote:
> I've spent a fair amount of testing this today, and when using the
> simple percentile_disc example mentioned above, I see this pattern:
>
>                                  master   patched   speedup
>    ---------------------------------------------------------
>     generate_series(1,1000000)      4.2       0.7      6
>     generate_series(1,2000000)      9.2       9.8      0.93
>     generate_series(1,3000000)     14.5      15.3      0.95
>
>
> so for a small dataset the speedup is very nice, but for larger sets
> there's ~5% slowdown. Is this expected?

I had a look at this patch today with a view to committing it, but it
seems that nobody's commented on this point, which seems like an
important one.  Any thoughts?

For what it's worth, and without wishing to provoke another flamewar,
I am inclined to use Andrew's version of this patch rather than
Peter's.  I have not undertaken an exhaustive comparison, nor do I
intend to.  It is the reviewer's responsibility to justify the changes
they think the author needs to make, and that wasn't done here.  On
the points of difference Andrew highlighted, I think his version is
fine.

-- 
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company