[Prev][Next][Index][Thread]
Re: beyond java
> ...
> I'd be interested to know the relative speeds of your Dylan and Java IDL
> compilers. Perhaps your readers would too. Or do you want to leave
> them with the impression that Dylan programs are shorter and nicer than
> Java ones, but they're probably even slower than Java because otherwise
> the author would certainly have mentioned the speed advantage?
Sorry for not replying to this sooner. I just wanted to say thanks for the
comments. I might try to produce a revision incorporating them, but I don't
seem to have much spare time at the moment.
On the subject of speed I can't say that I've benchmarked either one or that
they have been tuned for speed particularly (*) -- so comparisons on that
basis may be unfair. I don't even currently have the two compilers on the
same machines. Subjectively I think that the Java one is quite slow and part
of the reason is that it has a lot of classes (some unnecessary) that have
to be loaded on startup. This initialization time is highlighted in our
current build system (Perforce's JAM) which we use to drive the IDL compiler
one file at a time. There are other problems are well, such are repeatedly
reprocessing #included trees of background IDL, but that particular issue is
common between the compilers.
__Jason
(*) Of more concern to us with the Java IDL compiler is the quality of the
emitted code -- particularly in terms of VM consumed.