Re: Speeding up inform


10 Jul 95 17:03:01 BST

In article <3tqu60$lgs@flood.xnet.com>, mblaheta@flood.xnet.com (Michael Blaheta) writes:
> On Sat, 8 Jul 95 22:03:00 -0500, Tim Middleton (tim.middleton@canrem.com) said:
>> CC= Has anyone tried speeding up the inform compiler ? It's very slow on
>> CC= my system, so I'd welcome some ideas.
...
>
> I do like the idea of precompiled headers, though... really, they
> shouldn't need to be changed ever, what with "extend"s and so on.
>
> Graham, are you listening? Do you see precompiled headers in your great
> big Inform crystal ball?
>

Yes, over here in the spirit world (a.k.a. Oxford University) I do try to
listen to the plaintive cries of the souls of the living.

I have considered something like this twice or so. I may yet consider
again. The problems are twofold: one, the format is very complicated
in terms of cross-references, which makes linking a non-trivial proposition;
two, it would be a major, major destabilisation of a currently reasonably
sound compiler. There are lots of subsidiary problems too (the library
compiles differently according to constant settings; some sort of "make"
utility would need to be built in, and making it work portably could be
a real nightmare, etc.).

Also, it wouldn't save quite as much time as appears. It would make the
compilation of tiny games very quick; but the time saving on really
substantial games would be nowhere near as dramatic. I do have to wait
2 minutes for "Jigsaw" to compile, but that's not the end of the world,
and it is pretty huge.

Well... a pusillanimous sort of post. I may yet reconsider.

Graham Nelson