Re: I/O on select count(*)

Heikki Linnakangas escribió:
> Alvaro Herrera wrote:

>> The problem is that the bgwriter does not understand about the content
>> of the pages it is writing -- they're opaque pages for all it knows.  So
>> it cannot touch the hint bits.
>
> We know what kind of a relation we're dealing with in ReadBuffer, so we
> could add a flag to BufferDesc to mark heap pages.

Hmm, I was thinking that it would need access to the catalogs to know
where the tuples are, but that's certainly not true, so perhaps it could
be made to work.

>> If we had the bitmask in a separate map fork, this could be cheap.
>
> I don't buy that. The point of a hint bit is that it's right there along
> with the tuple you're looking at. If you have to look at a separate
> buffer, you might as well just look at clog.

True -- I was confusing this with the idea of having the tuple MVCC
header (xmin, xmax, etc) in a separate fork, which would make the idea
of index-only scans more feasible at the expense of seqscans.

--
Alvaro Herrera                                http://www.CommandPrompt.com/
PostgreSQL Replication, Consulting, Custom Development, 24x7 support