Richard Stallman Calls LLVM A "Terrible Setback"

[boltronics]

BSD: Free, but doesn't require derivatives to be Free
GPL: Free, but requires derivatives to be Free

Or to put it another way:

BSD: Usually free, but allows developers to restrict users through proprietary forks.
GPL: Free, and ensures the code stays that way.

Which is more free is obvious.

[dee.]

Seriously, it feels like there's a thread right now in 4chan or some other shithole, where kids are laughing how easy it is to troll phoronix forums - just go there, make an account, and start talking way over the top shit about GPL or BSD... shitstorm ensues.

[Delgarde]

LLVM is a tool.

On a completely unrelated note, I'd suggest that there are a lot of those in this thread...

[a user]

The way RMS speaks you'd think that writing closed-source code was a crime against humanity.

no! what he says is that taking others code and contributions as base for closed source without giving anything back is a crime against contributors.

you may or may not share his opinion but at least try to understand what he is saying and don't twist his words around to make it look more dramatic than it is.

[a user]

Or to put it another way:

BSD: Usually free, but allows developers to restrict users through proprietary forks.
GPL: Free, and ensures the code stays that way.

more or less correct.

Which is more free is obvious.

obviously not. otherwise we wouldn't have this never ending discussion. the understanding of "free" quite differs among the people.
don't forget that some think anarchy is highest level of freedom.

[raineee]

So while the Clang C/C++ compiler is nearly running at the same speed as GCC and has also spawned many interesting projects with its interesting modular compiler infrastructure design like Gallium3D LLVMpipe, other LLVM GPU back-ends, various language front-ends, OpenCL support, EmScripten for compiling to the web, disassembly/decompilers, and countless other interesting open-source projects that build upon LLVM, RMS says "the existence of LLVM is a terrible setback for our community precisely because it is not copylefted and can be used as the basis for nonfree compilers."

You seemed to have missed the point. As RMS mentions, technical superiority is not the purpose of free software people.

[raineee]

Stallman is wrong to this one. He has a limited vision for the word freedom. A closed program that runs on all platforms is more free than an open one that runs only on one (i speak for hardware and software). In our case GCC did take the Intel road for statically linking compilation as MS compilers did. So they benefit mostly Intel and x86 monopoly and Instruction_set based computing, and they stall technology. We have the same computing for 40 years. I prefer a BSD LLVM because i am thinking than if consoles like PS4 use it, and some graphic engines use it as well, there is a huge change to get rid of x86 processors, in favor of better processors or free processors and easy emulation. I don't understand something: What is the importance to have something free (copy left), that doesn't work or can't be used with a new vision of what is beneficial for as in our time.

No, you are are wrong on this one.
GCC is a FSF project, and was created for the singular purpose of being a free compiler suite. Motivations aside, it's FSF's code (and all contributers agree to GPL before their patches are accepted). Its not others' place to dictate how FSF should publish their project.

And luckily, unlike you try to depict, GCC does work.

[pasaulais]

no! what he says is that taking others code and contributions as base for closed source without giving anything back is a crime against contributors.

you may or may not share his opinion but at least try to understand what he is saying and don't twist his words around to make it look more dramatic than it is.

Companies are giving back contributions to LLVM. The reason is simple: maintaining your own fork requires a lot of work that could be spent doing something useful.

Some of these companies just take LLVM/Clang and make changes anywhere in the source tree (without contributing these changes back) because it's the easiest thing to do for them. Then months/years down the road they realise how much work it is to merge these changes with new releases. That's when they start tracking tip and pushing back changes upstream in order not to waste so much time merging and not to be months/years behind everybody else in terms of features, optimizations, etc.

Of course what they contribute back is often a small part of their work. But that's usually the code that is most useful to the community, e.g. changes to the shared, architecture-independent code.

[-MacNuke-]

I think it is pretty shortsighted to believe that companies use LLVM only to "steal" code. Most companies give their code (like backends for processors) back to LLVM. Just read the LLVM mailing list. Other parts like OpenMP from Intel are also an example.

I think the reason is that it is far more easy to add stuff to LLVM than to GCC. You only need to inherit some classes and fill in the gaps, like converting the LLVM IR to you machine code. Than everything works.

Before LLVM most companies invented their own compilers. Like NVidia, AMD, Intel, many scriptlanguages, ... now they are "all" (sure, not all) working together on LLVM. Sure, all working on their own backend (PTX, AMDIL, ...), but everyone can use it because the code was given back.

And... because of that, the Khronos Group was able to standardize a OpenCL binary format, based on LLVM IR. This leads to better compatibility for OpenCL code on ALL hardware. One OpenCL frontend for all hardware out there... this as an example.

So, whatever one mght believe about GPL vs. BSD, GGC vs. LLVM... for whatever reasons thing are possible with LLVM that are not possible with GCC, and all these things are OpenSource in LLVM and as long as these things are not possible with GCC, LLVM will be used. So GCC has to move on, not LLVM (like standardize a IR / Bitcode one can use to standardize a binary format).

So my opinion on LLVM is that it has genereated more contributions to the OpenSource world (also to GCC, like better diagnostics), and so, it is welcome.

[aphirst]

Of course what they contribute back is often a small part of their work. But that's usually the code that is most useful to the community, e.g. changes to the shared, architecture-independent code.

A situation to worry about is when architecture-dependent code is what gets withheld from the public codebase.

Companies who manufacture microprocessors are in the best position to provide this kind of low-level optimisation knowledge. The fear is that if LLVM were dominant, in 10 years' time the public version would only be able to perform very naive optimisations on the then all-but-secret architectures of AMD/Intel/Arm/etc., who provide their own proprietary fork of LLVM which optimises properly for their CPU type. There's a lot of incentive to do this, and it would be pretty crippling for those who wanted to live their lives using only Free software.