Modularizing GCC Gets Discussed, Again

[phoronix]

Phoronix: Modularizing GCC Gets Discussed, Again

There's yet another conversation taking place within the GCC camp about modularizing the GNU Compiler Collection...

http://www.phoronix.com/scan.php?page=news_item&px=Modular-GCC-Discussion-GSoC

[wizard69]

Nice!

The only problem with this is that it will take years, maybe a decade, to fully modernize GCC. That is if there is little resistance from entrenched stake holders.

[hubicka]

Phoronix: Modularizing GCC Gets Discussed, Again

There's yet another conversation taking place within the GCC camp about modularizing the GNU Compiler Collection...

http://www.phoronix.com/scan.php?pag...iscussion-GSoC

Hehe, there is not only conversation, but real work going on. GCC 5.0 contains great deal of changes cleaning up interfaces and adding better APIs. Changes are easier with specific goals in mind. LTO and libjit are two important projects bringing real motivation to gets things cleaner and more flexible.

[Luke_Wolf]

The only problem with this is that it will take years, maybe a decade, to fully modernize GCC. That is if there is little resistance from entrenched stake holders.

Which there's been significant resistance by Stallman, going by the GNU lists...

[tuxd3v]

Which there's been significant resistance by Stallman, going by the GNU lists...

Maybe, but the problem is...gcc is the bigger compiler out there...
It will take a decade or so, to transform...

They can start by little parts, starting standardizing concepts, but in the end, they will have to deal with a bunch of code, that seems to be too much...I don't even know if it is possible at all...

[Kemosabe]

Maybe, but the problem is...gcc is the bigger compiler out there...
It will take a decade or so, to transform...

They can start by little parts, starting standardizing concepts, but in the end, they will have to deal with a bunch of code, that seems to be too much...I don't even know if it is possible at all...

Perhaps a complete rewrite would help? Yeah, let's rewrite and underline they shiny newiness with a brand new name! Let's call it c-language ... too long ... let's call it clang! Oooops

[justmy2cents]

deja vu anyone?

https://www.youtube.com/watch?v=RIctzAQOe44&t=861

[sobkas]

deja vu anyone?

https://www.youtube.com/watch?v=RIctzAQOe44&t=861

So you don't remember the time in with egcs was a thing?
Maybe it's a time for an egcs++?
More info:

https://gcc.gnu.org/news/announcement.html

https://gcc.gnu.org/wiki/History

[wizard69]

Which there's been significant resistance by Stallman, going by the GNU lists...

Exactly!

Stallman is more than resistance though as he out right undermines anything that might be seen as a new approach. It is a shame that he turned GPL3 into such an unworkable license for companies to work with as there is little reason to support the development of GCC anymore.


The other big gotcha here is mind share that LLVM attracts. Not so much for the C/C++ compilers but rather for all the new language development that has taken place on top of LLVM. Just today I noticed on Phoronix that another LUA derivative is in the works built upon LLVM. This is just one of many languages now working on top of LLVM. This means there is a considerable amount of talent being expended on interaction with LLVM which is helps with LLVM development as well as the language in question.

The head start LLVM has is huge. This is why I suggested that the GCC team has a decades worth of work ahead of them. That is work to make the GCC suite as flexible and as broadly supported as LLVM is now.

[duby229]

While I do agree that LLVM is awesome. Clang still has issues with the GCC entrenchment. Right now gcc compiles my entire gentoo system without any problems really. Clang only works with specific packages. It seems like clang needs a gcc compatibility mode or something so that it deals with gcc entrenchment better.