LLVM 2.6 Released, Clang Is Now Production Ready

[XorEaxEax]

I'm using -march=native -O3 -fomit-frame-pointer in my makepkg.conf on Arch Linux and I've had no problems in recent year. True I don't recompile everything (perhaps unlike many Gentoo users) but mostly packages that actually benefit from speed optimization. This includes compression, video codecs, emulators... Also I often use other flags like PGO (-fprofile-generate, -fprofile-use), -ffast-math on stuff I recompile and it's been rock solid. Also I'm going to try out the new link time optimizations in the current gcc 4.5 snapshots soon, although they are likely to break certain things since they are still in development.

Are there any specific packages or threads that you can point to that fails to build correctly with -O3 using gcc 4.x?

[XorEaxEax]

If there's nothing wrong with GCC's PGO support, then why is GCC being blamed here:
https://bugzilla.mozilla.org/show_bug.cgi?id=418866

Yes there were problems compiling the firefox 3.6 beta, atleast that's what I read in the Arch forum. But gcc 4.5 seemed to build it fine. Firefox 3.5.x compiles fine with PGO.

[StringCheesian]

Are there any specific packages or threads that you can point to that fails to build correctly with -O3 using gcc 4.x?

Hmm... maybe you're right. I did some forum searches and googling and couldn't find anything about -O3 on 4.4.x.

There was this for -ftracer

http://forums.gentoo.org/viewtopic-p-5846744.html#5846744

[XorEaxEax]

There was this for -ftracer
http://forums.gentoo.org/viewtopic-p...4.html#5846744

Well, -ftracer has to be manually set and isn't part of any -O levels. It is however added during the -fprofile-use stage from gcc 4.4 atleast, which would indicate that it is stable from that version. And since I've done alot of profiled builds on a variety of packages it certainly seems stable from my end.

[Wyatt]

I like the other features (other than faster compile time, and hopefully faster programs) even more, but I'm not sure what languages can take full advantage of it...

say, a REAL meta-programming facility in C++...

I think you're looking for D...

Are there any specific packages or threads that you can point to that fails to build correctly with -O3 using gcc 4.x?

Seems like Boost usually fails to compile with anything beyond -O2, but that may be a design feature that attempts to prevent you from using it.

[kraftman]

ALL projects die eventually. Linux itself will die someday. Maybe not anytime soon, but it'll happen.

In what contest? I don't see any Linux fork out there. Are you a prophet or something? :P

Heck, Windows already died twice, too.

When?

Clang could replace GCC, and it can only be an improvement (because if it isn't, it won't ever replace GCC in the first place).

Clang can't even compile Linux kernel, GCC supports far more architectures, it's solid, mature and it's noticeably faster in some things, so right now, it doesn't seem GCC will die before Clang.

Btw. GCC devs can always take LLVM code :> If it will be hard to adapt, why not to do llvm flag which will instruct GCC to use LLVM (to use it in situations where it makes faster code) instead: gcc -llvm [...] GCC will have all benefits then ;P just joking...