Re: Improved scanner performance

Tom Lane writes:

> Peter Eisentraut <peter_e(at)gmx(dot)net> writes:
> > Tom Lane writes:
> >> I had the idea that -CF would enlarge the lexer tables quite a bit ---
> >> what's the change in executable size?)
>
> > +150 kB
>
> > I've also looked at -CFe, which is supposedly the next slowest level, but
> > it doesn't do nearly as well.
>
> Ouch; that sounds like about a ten percent increase in the size of
> the backend executable. That's enough to reach my threshold of pain;
> is the long-literal issue worth that much?

Here's a breakdown of the postmaster file sizes and the wall-clock run
time of the long-literal test:

no options 1749912 1m58.688s
-CFe 1754315 1m49.223s
-CF 1817621 1m43.780s
-CFa 1890197 1m45.600s

(These numbers are different than yesterday's because they don't have
profiling and debugging overhead.)

Seeing this, I think -CF should be OK space and time-wise.

> How much of your reported improvement is due to -CFa, and how much to
> the coding improvements you made?

As I recall it, probably a third of the overall improvement came from
using -CF[a].

--
Peter Eisentraut peter_e(at)gmx(dot)net