Re: Alternatives to gcc (was Re: gcc 4.3: when will it become standard compiler?)

On Wed, Jan 14, 2009 at 11:21:00AM -0500, Kevin Wilcox wrote:
> 2009/1/14 Michel Talon <talon_at_lpthe.jussieu.fr>:
> 
> > Apparently the FreeBSD project doesn't want to include any GPL V3 because
> > there are industrial partners who have banned the GPl V3, out of purely
> > ideological position, without any rational basis. I wonder why the FreeBSD
> > project has any reason to follow them.
> 
> If you make the claim that banning GPL v3 is being done by commercial
> entities that rely on FreeBSD (and other BSD licensed code) for
> reasons that are purely ideological and have no rational basis then
> you either do not fully understand the significance of GPL v3 versus
> GPL v2, you haven't sufficiently worked with a commercial entity with
> regards to GPL/LGPL/BSD code that you (or others) have licensed to
> them or you are just trying to troll the FreeBSD community. I would
> wager the first two are the case here?
> 

Yes, i make the claim that, as far as the compiler tool chain is considered,
there is no difference between GPL V2 and GPL V3 because both licences make no
restriction on the software compiled with the tool chain.  Of course i make no
claim considering other type of software. I have no problem that FreeBSD
produces an operating system under BSD licence, unpolluted by viral clauses.
And, yes, i have an idea of the difference between V2 and V3.

Dimitry Andric says that "dependence on GNU gcc has always been bad", but
gcc has been the only "game in town" for all these years and will perhaps
remain for a lot of years. The FreeBSD project is not a project on compiler
construction. Having done computations in the past, i have seen times when
gcc produced code twice slower than the Intel compiler. I have also seen
that it took around 10 years to the gcc folks, who are very competent compiler
developers, to get more or less at the same level of perfomance than the 
Intel compiler. So i hope, and i am quite sure, that the FreeBSD developers
will not abandon something that works rather well for some very hypothetical
advantage. If it appears at some moment that llvm works well and produces code
as fast as gcc, for all the platforms of interest for the FreeBSD people, 
i have no doubt that they will switch immediately. But one of the aims of 
FreeBSD, distinguishing it from other BSDs, is performance. Having a good
compiler is of paramount importance for performance, not only of the base
operating system, but of all the ports running on it. People wanting BSD
purity, above everything, don't have any difficulty finding it.



-- 

Michel TALON