CLooG hell

Vladimir Kraljević vlaad@beotel.net
Mon Jul 16 17:03:00 GMT 2012


Dear Sir,

I find the tone of your letter inappropriate. CAPITALS, "Clear enough 
yet", "Do you understand that yet?" and similar phrases could be 
offensive to some people. If you directed it toward a young man, it 
could have devastating results for that young spirit.

However, I'm not here to argue with you, I'm here to get some help - I 
believe that is the reason that some wise guy named this mailing list 
gcc-help, but it seems that nobody recalls why these days.

Now, I'd like to write something to the people willing to actually help, 
if there are such.

I'm using help from http://gcc.gnu.org/install/ when trying to build gcc 
from files I got with the following command (and my intention was to get 
graphite, framework for HL mem optimizations, maybe I got something 
wrong, I don't know):

svn co svn://gcc.gnu.org/svn/gcc/branches/graphite graphite

I've also tried and successfully built gcc 4.6.3, gcc 4.7.1 and gcc 
4.8.0 without graphite support. Now I wanted the graphite framework 
included in the build, in order to test optimizations provided.

I'm stuck with something I don't have time to investigate, so I'm kindly 
asking for help, if there is competent person, willing to hear and track 
the issue with me. My ultimate goal is working gcc (with latest 
libraries if possible, if not - say so - don't write "libppl 0.11 or 
later" if it actually requires "libppl 0.11 ONLY", I believe that we 
need clarification on this - google and you'll find that I'm not the 
only one having trouble with this issue) and g++ supporting these new 
optimization methods.

Kind regards,
Vladimir


On 07/16/2012 02:26 PM, Jonathan Wakely wrote:
> On 16 July 2012 13:13, Vladimir Kraljević wrote:
>> I'm refering to these instructions: http://gcc.gnu.org/install/
>> If there is some better source, please tell me.
> For the third time, USE THE INSTRUCTIONS INCLUDED IN THE 4.7.1 RELEASE SOURCES
>
> The instructions at http://gcc.gnu.org/install/ refer to the SVN
> trunk, which was recently changed to use CLooG-ISL, without PPL.
> Those instructions don't even mention PPL.
>
> If you want to build 4.7.1 use the instructions for 4.7.1, not the
> instructions for trunk. Clear enough yet?
>
>> I know that 4.6.3 and 4.7.2 work with PPL 0.11 (using provided
>> instructions). however, configure should check if later version is present,
>> IMHO, as it states.
> Your opinion is noted.
>
>> What I'm trying to highlight is that builders are enforced to use old
>> library (PPL 0.11) for some reason (then again, the question is why using
>> PPL when it is instructed to use ISL? I allow the possibility that I didn't
>> understand CLooG dependencies well).
> Because you're trying to use the instructions for GCC trunk to build GCC 4.7.1
>
> THOSE ARE THE WRONG INSTRUCTIONS.
>
> Do you understand that yet?
>
>> I'm going to test this statement after PPL 0.12 build, I'll write about this
>> issue once more.
> Hmm, or you could use the version recommended in the 4.7.1 installation docs.
>
>> But seriously guys, I think that it should be nice to plan for the future.
>> New versions of libraries are sometimes faster and sometimes provide bug
>> fixes. Just allow the possibility of choosing the libraries to the builders.
> As documented:
> "While any sufficiently new version of required tools usually work,
> library requirements are generally stricter. Newer versions may work
> in some cases, but it's safer to use the exact versions documented. We
> appreciate bug reports about problems with newer versions, though."
>
> Since you're failing repeatedly to get it working, maybe you should
> just stick to the exact versions documented and leave it to others to
> test the newer versions.  Follow the instructions and it will work.
>
> GCC trunk no longer uses PPL at all, that dependency has been replaced
> by ISL, but for GCC 4.7.1 just try following the instructions. The
> right ones.
>
> You're not saying anything new or useful to anyone, you're just
> failing to follow instructions and complaining about a known situation
> which is described in the docs, so I'm done with this thread. Good
> luck.





More information about the Gcc-help mailing list