Hello,
I would like to update this page:
http://clang.llvm.org/features.html
In particular, I would like to remove most of the information relative
to performances. This for various reasons:
* It is better if we leave third parties to do "independent" benchmark.
* We compare Clang (version unspecified) with gcc 4.0 or 4.2.
* The graphs have not been updated for a while.
* Clang is well known now. I don't think we still need to explain why
Clang is great.
Any objection if I apply the attached patch?
Cheers,
Sylvestre
update-features.diff (4.58 KB)
imho it's good to remove outdated performance comparisons or at least update them against current versions. If not publishing any data is a goal(?) then a reference to the trusted "online benchmarks" should be made. There are tons of passionate people who or may not mean well, but publish worthless garbage.
To avoid trolls - It may be best to ping the community leaders on irc or via email and coordinate with them.
This seems like a good idea to me; this old marketing material doesn't
really belong on the website any more, but I'd like input from others.
Hello,
I would like to update this page:
http://clang.llvm.org/features.html
In particular, I would like to remove most of the information relative
to performances. This for various reasons:
* It is better if we leave third parties to do "independent" benchmark.
* We compare Clang (version unspecified) with gcc 4.0 or 4.2.
* The graphs have not been updated for a while.
* Clang is well known now. I don't think we still need to explain why
Clang is great.
I disagree about the latter. Clang has a long-standing reputation of compiler
which is faster than GCC, but generating slower code than GCC. On the other hand,
recent releases of GCC greatly improved compilation speed, memory consumption,
and diagnostics.
So it would be useful to explain why Clang is still great when we have GCC 4.9.
Perhaps so, but the text being removed is not that explanation. It explains
why Clang 2.9 was better than GCC 4.0, which is no longer relevant to
anyone.
--- end quoted text ---
Hi,
I agree clang needs to document why it is a good choice compared to g++-4.9.
I believe g++-4.9.x will experience wide adoption in 2014 and 2015.
One issue likely is strongly in clang's favor: g++ tends to violate the standards
frequently as a matter of practicality. My understanding is it is deliberate
in the context of performance and implementation details. But there is a real
cost for folks who are not limiting their work to the g++ tool chain. I suspect
a lot of folks would want to be informed of these differences.
I speak of my experiences with g++-4.8.x and clang 3.3-xubuntu but have no idea how
these issues play out with current releases.
enjoy,
Karen
> Hello,
>
> I would like to update this page:
> http://clang.llvm.org/features.html
>
> In particular, I would like to remove most of the information relative
> to performances. This for various reasons:
> * It is better if we leave third parties to do "independent" benchmark.
> * We compare Clang (version unspecified) with gcc 4.0 or 4.2.
> * The graphs have not been updated for a while.
> * Clang is well known now. I don't think we still need to explain why
> Clang is great.
I disagree about the latter. Clang has a long-standing reputation of
compiler
which is faster than GCC, but generating slower code than GCC. On the
other hand,
recent releases of GCC greatly improved compilation speed, memory
consumption,
and diagnostics.
Maybe one of our "features" could be "stimulate improvements in GCC" 
-- Sean Silva
Committed. Thanks for the feedback.
If you disagree, feel free to update the page or send me comments.
Cheers,
Sylvestre
g++-4.8.x -std=c++11 also includes numerous standard violations where they provide
support for a few c++14 features. I personally don't agree with it. IMO, folks who want to
use those features have the -std=c++1y available. It seems reasonable to think, once
-std=c++14 is released, they are likely going to leave those standard violations in
-std=c++11. I believe it is an unwise policy myself. But apparently a lot of folks
using g++ disagree with my view on that issue.
My understanding is clang provides much better support for the standards in this
context.
enjoy,
Karen