From: RaeNye <raenye_at_netvision.net.il>
Date: Sun, 18 Mar 2007 23:02:16 +0200

>So I'm finally ready to present a patch for the WPS tokenizer.
>The basic idea is to store the WPS as an array of tokens.

>Please tell me what you think of the concept and the code.
>IMHO this has the advantage of being much clearer, readable and
maintainable than the current code.
>As of performance, I don't have the confirmation that it's faster, but I
think it should be, as it was designed
>to make the parser do all the work. Binsize isa bit of a disappointment. I
thought it could be smaller,
>but instead I'm seeing a (small) increase on some targets (I haven't done
much checking, though).

I haven't yet looked at the code, but I think we could parse the WPS in a
preprocessor plugin
(don't call it a WPS compiler, it sounds intimidating), creating some
"bytecode" that would be interpreted by the WPS viewer
(really don't call it a WPS-VM, that's just sick).

We can surely save core size as well as CPU cycles.
All this can be made in a user-transparent way by having the preprocessing
done by a viewer plugin, caching the result with a WPS timestamp/filehash to
save future work.

The real question is whether we can come up with a platform independent
intermediate representation that captures all current (and hopefully future
WPS aspects).

My 2 cents,
R.
Received on 2007-03-18