[Xapian-discuss] xapian performance

[Olly Betts]

On Thu, Nov 23, 2006 at 10:58:08AM +0100, Arjen van der Meijden wrote:
> The first file (no-patch) contains the results (the 3rd in three runs) 
> with no patches at all. And the second the results of all the patches 
> from this thread and the converted database.
> I see no cases of slower queries, and some are even 30-40% faster.

Indeed - a quick bit of scripting on the two logs shows that the speedup
ranges from 7-34%.

> If you're interested I can also execute the same queries with your 
> display-blocks-read patch, but that'll have to wait untill next week.

It would definitely be interesting to have an idea just how many disk
blocks we're reading for these bad cases.

Cheers,
    Olly