Re: [Caml-list] -thread and ocamlc vs ocamlc.opt

[Pixel <pixel@mandrakesoft.com>]

Xavier Leroy <xavier.leroy@inria.fr> writes:

[...]

> > I've been trying to reduce the size of the ocaml rpm which is getting quite
> > huge (27MB), and doing
> >   for i in /usr/bin/*.opt ; do
> >     ln -s `basename $i` `echo $i | sed "s/.opt$//"`
> >   done
> > helps!
> 
> Frankly, I would rather remove the .opt compilers from the RPM, rather
> than remove the non-.opt compilers.  

?? switching from non-.opt to .opt makes some impressive speed-ups (x2) when
compiling.

> 
> Concerning the RPM size, several things should be noted (with tongue
> in cheek):
> 
> - If you had one package with gcc, gdb, libc and assorted goodies, it
> would be even larger

the main difference is that not many people would remark if ocaml was not
included ;p

> - Mandrake contains even larger packages of even more dubious utility
> than OCaml

trolls, trolls, vive les trolls :)

> - It is certainly possible to split it into different packages, if you
> think not all users will need everything

done

> - Mandrake could always switch to DVDs as their distribution format :-)

expensive, hard to test, many packages to maintain is costly...

> 
> > - should threads.cmxa be included somehow/somewhere? The doc says
> >     ocamlopt -thread other options threads.cmxa other files
> >   is that wrong?
> 
> It also says you have to configure with POSIX threads for this to work.
> The reason this is not the default configuration is that you need to
> trust your POSIX thread implementation and have a good understanding
> of its limitations (slower thread creation, fewer max threads allowed).
> My experience as the initial developer of LinuxThreads suggest that
> these conditions are not often met...

i don't know much about this. Are these conditions met under linux?


thanks!

--
Pixel
programming languages addict      http://merd.net/pixel/language-study/
-------------------
Bug reports: http://caml.inria.fr/bin/caml-bugs  FAQ: http://caml.inria.fr/FAQ/
To unsubscribe, mail caml-list-request@inria.fr  Archives: http://caml.inria.fr