[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: ocaml 3.04-5 in incoming ...



On Thu, Jan 24, 2002 at 07:32:09AM +0100, Christian Perrier wrote:
> Quoting Sven (luther@dpt-info.u-strasbg.fr):
> 
> > Changes:
> >  ocaml (3.04-5) unstable; urgency=low
> >  .
> >    * Split ocamlc.opt, ocamlopt.opt and ocamllex.opt into the
> >      ocaml-native-compilers package.
> >    * The ocaml-best-compilers virtual package is provided by
> >      ocaml-native-compilers if it is built and by ocaml if not.
> >    * Added an ocaml-source package containing the source to ocaml for packages
> >      which need them as build-depends.
> >    * Applied Ian Zimmerman's caml.el patch. (Closes:#129650,#130301)
> 
> I have followed the discussion about this problem from time to time.

:)))

> Are my conclusions OK : I should modify the Build-Depends of my package from 
> ocaml to ocaml-best-compilers, am I wrong? Then, if the optimised compiler
> exist for the platform the package is built on, it will be used (my rules
> files already handles this) and the normal compiler will be used if the
> optimised one does not exist.

Well, what packages do you maintain, are they very big sources, like coq,
where there is a sensible difference, or smaller stuff, where there is no
noticeable difference.

Also notice, that it was said by the caml team, that there is a tradeoff
between the faster compilation time of the native compiler and the longer
loading time of the bigger executable. So the slower the processor, the more
benefit one would gain from using the native compilers.

Finally, i think it will mostly make a difference for the arm autobuilder (if
there is such a thing), since the other opt plateforms are fast enough to not
have real performance gain on small files. That said, i have not done any kind
of testing, so the above may not be exact.

In the long run, i think it is only needed to build depend on the
ocaml-best-compilers if your source is huge. Where exactly the barrier is,
this i don't know.

Friendly,

Sven Luther



Reply to: