[cfe-dev] Switching http://llvm.org/demo to clang or providing a separate clang demo

[David Blaikie]

On Mon, Aug 29, 2011 at 6:14 PM, Chris Lattner <clattner at apple.com> wrote:

>
> On Aug 29, 2011, at 2:37 PM, David Blaikie wrote:
>
> > http://lists.cs.uiuc.edu/pipermail/llvmdev/2011-March/038828.html
>>> > http://llvm.org/bugs/show_bug.cgi?id=1440
>>>
>> So, now to figure out what to do with the page. The bug/patch I mentioned
>> above seems to still have some outstanding issues (given the discussion in
>> the bug itself) - cross compiling on the demo machine/website won't come for
>> free. Any ideas on what we should do there? Should we actually get the demo
>> machine setup with all the necessary libs/things to do real cross
>> compilation?
>>
>> I wonder what the real benefit is here though versus the cost? If we are
>> just trying to give a quick demo of LLVM/Clang, then how much value is in
>> supporting multiple targets?
>>
>> Also, whats the benefit in demoing targets that are not officially
>> supported by releases?
>>
>
> For myself I'm not sure how the cost/benefit ratio comes out - both are
> fairly unknown to me. Mostly I was bringing this up to make sure the
> existing outstanding work, patch, and bug were resolved rather than left to
> rot some more. More or less I'm asking Chris if, given the points raised in
> the bug regarding multi-architecture targeting, he still felt it was
> something worth pursuing or not.
>
>
>
> Are you interested in showing the -S output?  I think it would be nice to
> be able to get (for example) ARMv7 and X86 output.  I don't think that going
> farther than that is a good idea though.
>

 Personally, I don't have much interest in the back end side of things - for
myself I'll be more interested in improving the front end (C++0x, libc++,
the static analyzer, maybe post-fixit code, etc). But I don't want to work
on lots more changes to the page while there's still some pending work here
that would be easier to use now than later.

Since Duncan brought up the major objection to producing incorrect assembly
(due to not actually having a full cross-compiling story on the demo
machine) I've included him on this mail.

I guess your take on it, Chris, was that what Bjarke put together is all you
were interested in - demoing the ability for LLVM to produce multiple
architecture assembly & it's not too important to you if the libraries don't
match the target platform, etc. (I'm sure a lot of experiments on the demo
page like that might not even use any libraries explicitly anyway)?

I'll take another look at Bjarke's patch & see about getting it up to date
with the current code, maybe stripping out the extra architectures &
bringing it back down to x86 and ARM so we have a talking point.

- David
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20110829/59369140/attachment.html>