Re: new cwm release

Ouch.  Well, it is a re-write using objects for things which were tuples
before, and using method calls instead of a wallpaper strip of code.
So it I would expect it to be slower ... but four times is tough.

The good news  is I haven't profiled it at all, whereas the old on I had
tweaked as much as I could.  Also, I plan to halve the number of indexes,
which
should reduce the load time into the store - I don't think half of them are
used.
I do indeed need some benchmarks.

(Is your data confidential or could you email it as a public performance
test?)

Sigh... each change has made it slower!

(I need to clone it in C or machine code, but that's another matter)

TIm

----- Original Message -----
From: "Norman Walsh" <ndw@nwalsh.com>
To: "Tim Berners-Lee" <timbl@w3.org>
Cc: <www-rdf-interest@w3.org>
Sent: Friday, February 22, 2002 7:36 AM
Subject: Re: new cwm release


> / "Tim Berners-Lee" <timbl@w3.org> was heard to say:
> | I have just made a new release of cwm.  Functionalty is basically the
same,
> | but code is better inside.  List handling should be improved.  Formulae
> | are interned and can therefore be compared and matched in queries.
> | Hmmm. must add tests for that.
>
> Have you run any performance tests? The new release seems slower by at
> least a factor of four, at least for my one big document (about 7
> minutes with the 1.82 release and I gave up at about 28 minutes with
> the latest release from CVS).
>
>                                         Be seeing you,
>                                           norm
>
> --
> Norman Walsh <ndw@nwalsh.com> | Simplicity is always a virtue.--Edward
> http://nwalsh.com/            | Abbey
>

Received on Friday, 22 February 2002 09:33:36 UTC