Apple Originally Tried To Give GPL'ed LLVM To GCC

[Annabel]

The modularity makes it easier, AFAIK. You only need to care about the lexer and the parser, everything else is already done and can be plugged.

why isn't the same with GCC?

[storm_st]

The modularity makes it easier, AFAIK. You only need to care about the lexer and the parser, everything else is already done and can be plugged.

So, what exactly prevent LLVM developers all that 8 years from "plugging". I see only plug Apples Objective-C and lately few years attempts to do some C/C++ clang things. No OpenMP plugging (actually only related talk every article, it almost there, all that years, Intel, shmintel, in fact you cannot compile average OpenMP code yet, early experiments), no endless GCC arch plugged. Deadly GPL monster curse that plug sockets ? When that sacred god blessed "better architecture" will actually shine? In 2020? 2050?.

[doom_Oo7]

I would like to know why this is easier with llvm, someone can do the same for GCC. What's so much easier in LLVM?

You just need to make something that parses your language to LLVM's very readable and easy "super-asm".

As a matter of fact, in my engineering school, we have compilation classes in second year, and we must make a compiler from a random language (in my year it was Ruby, this year I think they have something not very far from VHDL) to LLVM assembly.

[Annabel]

You just need to make something that parses your language to LLVM's very readable and easy "super-asm".

As a matter of fact, in my engineering school, we have compilation classes in second year, and we must make a compiler from a random language (in my year it was Ruby, this year I think they have something not very far from VHDL) to LLVM assembly.

GCC have a IR too

[zanny]

So, what exactly prevent LLVM developers all that 8 years from "plugging". I see only plug Apples Objective-C and lately few years attempts to do some C/C++ clang things. No OpenMP plugging (actually only related talk every article, it almost there, all that years, Intel, shmintel, in fact you cannot compile average OpenMP code yet, early experiments), no endless GCC arch plugged. Deadly GPL monster curse that plug sockets ? When that sacred god blessed "better architecture" will actually shine? In 2020? 2050?.

Uh, just copy / pasting from wikipedia here, but: ActionScript, Ada, D, Fortran, GLSL, Haskell, Java bytecode, Julia, Objective-C, Python, Ruby, Rust, Scala[2] and C#.

That is a lot of languages. More than GCC. And it is much easier to integrate a new language into llvm. You usually don't have to write asm backends for every architecture in gcc though, but you have to write to their internal codepaths for language structures if you want to implement a new language for it, which is more cumbersome than just translating to a common asm surrogate.

[Del_]

Interesting information on the events leading to the current state, thanks Michael. I have spent a couple of hours digging through follow-up mails to understand what happened. The sentiment from GCC developers did seem constructive and positive. The ones that I found shedding some new light (except the replies to the already posted links in the article) were these:

http://gcc.gnu.org/ml/gcc/2006-03/msg00706.html

demonstrating that four months later there was no copyright transfer, and no re-licensing. After that I only found release notices for LLVM with no replies and no mention of copyright or licensing:

http://gcc.gnu.org/ml/gcc/2006-04/msg00408.html

http://gcc.gnu.org/ml/gcc/2006-08/msg00126.html

http://gcc.gnu.org/ml/gcc/2006-11/msg00697.html

Then finally (half year after the introduction of GPL version 3) one short mail making it clear that LLVM would never be part of GCC:

http://gcc.gnu.org/ml/gcc/2007-12/msg00395.html

I may have missed something, but it seems to me that the copyright transfer policy of FSF was the main obstacle. Very grateful if others have additional information that can shed light on this. Personally, I believe FSF should abolish the practise of copyright transfer, requiring the + after the GPL version automatically gives them the right to re-license anyway. Of course, copyright also empowers the FSF to enforce compliance with the license, but I am leaning towards that not weighing up for the draw-backs of having copyright transfer agreements.

[Annabel]

More than GCC. And it is much easier to integrate a new language into llvm.

Source please?

[Nobu]

http://phoronix.com/forums/showthrea...170#post392170 ?

Or, the GCC docs, looks like 6 or 7 officially supported (including C/C++). Don't know about 3rd party.
There are a number of LLVM-based projects here, not all programming languages. Don't really feel like digging through, but feel free to do so at your leisure.

[mrugiero]

Personally, I believe FSF should abolish the practise of copyright transfer, requiring the + after the GPL version automatically gives them the right to re-license anyway.

Are you sure the + suffices? Remember what they try to do with the relicensing is avoid the use under GPLv2, which allows patent trolling and tivoisation. I don't know if the + allows to start distributing as 3+ later, as far as I understand, they have no right (without the copyright assignment) to restrict the use of following versions of GCC to GPLv3 and above.

[Annabel]

http://phoronix.com/forums/showthrea...170#post392170 ?

Or, the GCC docs, looks like 6 or 7 officially supported (including C/C++). Don't know about 3rd party.
There are a number of LLVM-based projects here, not all programming languages. Don't really feel like digging through, but feel free to do so at your leisure.

this does not answer what I asked