LLVM/Clang Replacing GCC In FreeBSD Base

[phoronix]

Phoronix: LLVM/Clang Replacing GCC In FreeBSD Base

In the quarterly report for FreeBSD, we learned something interesting: the FreeBSD developers intend to replace GCC with LLVM/Clang. The FreeBSD project wants to replace the GNU Compiler Collection with the Apple-backed Clang front-end compiler from LLVM...

http://www.phoronix.com/vr.php?view=NzI1Ng

[Pfanne]

im not much of a compiler guy, but would opencl speed up compiling?
i dont really know much about it, but asking can never hurt

[Kreed]

Is this going to happen in FreeBSD 8.0 or FreeBSD 9.0?

[DragonSA]

Is this going to happen in FreeBSD 8.0 or FreeBSD 9.0?

Short answer: Definitely not FreeBSD 8, probably FreeBSD 9 (although no commitment has been made).

There is a wiki that keeps track of clang's progress with FreeBSD and details how to compile FreeBSD with clang (for those interested...): http://wiki.freebsd.org/BuildingFreeBSDWithClang

Long answer: There is some C++ code in src (`find . -name '*.cpp'` shows groff and openssl). Although the C++ gcc frontend could be used the ideal would be for clang to support enough of C++ to compile these components. Furthermore a solution for ports need to be found, since there are bound to be quite a few ports that require special gcc features that clang doesn't support yet. These issues and the various bugs within clang will need to be sorted out (and direction provided by the developers).

[geoffp]

I, for one, would love to see some head-to-head GCC4 vs. LLVM+Clang benchmarking. This new compiler battle could be really interesting and beneficial.

[hax0r]

Dunno what's the point of making new compiler or branching off from gcc or whatever they're doing; 2x more work, 2x less developers on each project, why not work together?

[Ex-Cyber]

Dunno what's the point of making new compiler or branching off from gcc or whatever they're doing; 2x more work, 2x less developers on each project, why not work together?

They don't have the same goals. It's hard to "work together" if upstream has no interest in the changes you want to make (which is actually the reason that the current incarnation of GCC exists; a bunch of developers couldn't get their changes into GCC 2 and forked what eventually became GCC 2.95). Also, there's no reason to assume that it will be "2x more work, 2x less developers"; a fork or a new project can draw frustrated developers from the existing project and/or new developers who have no interest in contributing to the existing project for ideological, policy, or technical reasons.

[hax0r]

They don't have the same goals. It's hard to "work together" if upstream has no interest in the changes you want to make (which is actually the reason that the current incarnation of GCC exists; a bunch of developers couldn't get their changes into GCC 2 and forked what eventually became GCC 2.95). Also, there's no reason to assume that it will be "2x more work, 2x less developers"; a fork or a new project can draw frustrated developers from the existing project and/or new developers who have no interest in contributing to the existing project for ideological, policy, or technical reasons.

Cool, thanks .

[rohcQaH]

still, the big question remains unanswered: why the change to LLVM? What's wrong with GCC, where are the tangible benefits of LLVM?

It might be technical superiority (which I doubt for now), it may be because LLVM's license is more permissive (and BSD folks seem to avoid GPL whenever possible), it may be different reasons. Funny enough, the linked wiki page doesn't tell.

If anyone could shed some light on that, it'd be appreciated

[RealNC]

No, it's really the license that interests them. They have some hate of sorts against GPL stuff. I highly doubt this LLVM will offer any real advantages over GCC. Ever.