Re: Binary search in fmgr_isbuiltin() is a bottleneck.

On Mon, Sep 25, 2017 at 8:42 AM, Jeevan Ladhe
<jeevan(dot)ladhe(at)enterprisedb(dot)com> wrote:
> As Andres has already pointed, may be we want to move above code in a
> separate
> function, and just call that function here in case the hash is not already
> built.

No, I think what Andres is saying is that we ought to build the hash
table before we ever reach this function, so that we don't have to
have a branch here to check whether it's been done. I don't see why
that's particularly hard -- it can be jammed into the startup sequence
someplace early, I assume. In EXEC_BACKEND builds it will have to be
redone in each child, but that's just a matter of sticking a call into
SubPostmasterMain() as well as PostMasterMain().

Aside from that issue, this seems like a pretty boring patch. If a
hash table is faster than a binary search, then it is. Using
simplehash makes sense for this application, I think, and I'm not
really sure what else there is to say.

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