Re: Progress on fast path sorting, btree index creation time

On 15 February 2012 15:27, Robert Haas <robertmhaas@gmail.com> wrote:
>> I am inclined to agree that given that we already use Perl to generate
>> source code like this, it seems natural that we should prefer to do
>> that, if only to avoid paranoia about the inclusion of a dial-a-bloat
>> knob. I am at a disadvantage here, since I've never written a line of
>> Perl.
>
> I think it's still dial-a-bloat, but I feel pretty comfortable about
> how we've got that knob adjusted in this version.  It's almost as much
> improvement as any previous version, it applies to more cases, and the
> code footprint is the least of any version I've measured.

I'm happy that the principle that a dial-a-bloat knob isn't
necessarily a bad thing has been accepted, though that term is kind of
pejorative in that it implies that the knob necessarily adds bloat to
the binary.

I define bloat here as the addition of dead instructions to the
binary, or at least code that doesn't pull its weight. Clearly, that
isn't the case here, and I suspect that we will find that it isn't the
case in other places too.

--
Peter Geoghegan       http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training and Services