Talk Of GCC 5.0 To Be Modular, More Like LLVM
David Malcolm, a Red Hat developer, initially forked a GCC development list thread and re-titled it to GCC 5? and laid out his views. In his lengthy message he talked about the ongoing work for moving more of the GNU Compiler Collection from being written in plain C to being in C++, the need for a proper GCC plug-in API, and moving GCC towards being a collection of libraries.
These GCC 5 libraries could then be embedded into applications easily, just like LLVM. "What I'm really hoping for from GCC is a move towards a collection of libraries that can be embedded in (license-compatible) apps: LLVM is gaining ground for the use case of programs that need JIT-compilation (e.g. the X server, or a JVM). I appreciate that JIT compilation has different characteristics to a classic ahead-of-time compiler...GCC becomes a family of shared libraries that can be dynamically and statically linked. The executables become a thin wrapper that invoke these libraries (perhaps statically linked to minimize startup costs)." Of course, those are just his views.
Malcolm also brought up some existing GCC architectural issues like lots of global state, memory management issues, no name-spacing, threading, and more. He ended up proposing a GCC API as well in his e-mail. This e-mail sparked a host of replies and then another thread: GCC 5 & modularity.
Many developers it seems are in strong support of the GNU Compiler Collection being properly modularized to the point that GCC could be released as modules. This seems to be the general theme towards GCC 5.0.
Of course, properly modularizing this aging code-base is a huge undertaking. "And making the current GCC core base modular is not easy (it might be impossible), because we cannot decide at a glance to what module a given current code should belong. An important thing if we want to go modular is to define a road-map and probably to accept the fact that, if GCC 5 is a modular GCC, is will have less functionalities, less optimizations, less power than the current GCC 4.7: we probably would have to temporarily drop some feature or power from GCC to make it modular (otherwise, that would be too gigantic an effort). I have no idea if we can do that (this is why I am sometimes pessimistic); probably most GCC contributors are paid by companies which might not afford that."
Besides the huge effort in restructuring GCC, hindering a modularized LLVM-like GCC once complete would be its license. GCC is licensed under the GNU GPLv3 and that's very unlikely to change where as LLVM/Clang is made available under the more liberally licensed under a BSD-style license. More organizations like dealing with LLVM's license than the GPLv3 with GCC and is license-compatible with more software.
GCC would also need to come forward with some strong feature-set in order for developers to re-think their usage of LLVM rather than GCC within their projects. LLVM is already widely deployed within projects from Mesa/Gallium3D, OpenCL to Mono for its JIT engine to other interesting cases like converting C/C++ to JavaScript.
Even if all the developers agree to GCC 5.0's focus on modularity, it will be a while before the release materializes. This would be a significant undertaking and right now the development focus is on GCC 4.8 with GCC 4.7 coming in a few days.
20 Comments
