On Wed, 27 Oct 2004 11:46:07 +0200, Cdric Bastoul
<cedric.bastoul@prism.uvsq.fr> wrote:

>> Hi Michael,
>> here is CLooG GMP (a light version of the future CLooG 1.0, with still
>> some memory leaks and bugs). You should get and build the GMP library
>> (http://www.swox.com/gmp/ ) and build the PolyLib GMP
>> (http://icps.u-strasbg.fr/polylib/ then use the option
>> --with-libgmp=/gmp_install_directory for the configure script). For
>> CLooG you have to use the --with-bits=gmp and
>> --with-libgmp=/gmp_install_directory for the configure script. There is
>> no -cpp option in this CLooG version, I hope you don't use it. Be aware
>> that your example needs a lot of memory (more than 256MB), I did not
>> succeed to find the result on my laptop, but at least there were no
>> other problems.
>> Good luck, please feel free to send me some feedbacks about this
>> version, and please, please, do not distribute it.
>> Cheers,
>> 
>> Cedric


Hello Cedric,

I'd like to thank you for your very quick response to my problem. I
tried to build CLooG-GMP with our version of Polylib5.11.1 with
gmp-support enabled. The only problem that occured to me was that I
had to include "-lgmp" to the default library-flags in CLooG's
makefile.in.

Also, I always have a little trouble, because we use polylib as a
statically linked library, but there is no target in the CLooG
makefile for building the executable using static linking. It's no big
deal, I just add a little "-static" to the linking options. I don't
know if it would be usefull to others, but as you provide a static and
shared option for the CLooG library targets, could you think about
providing a similar option for the executable, too? Again, that is
only a very minor issue, I don' want to waste your time on things like
that... 

One more important thing: there is that "MAX_RAYS" constant that is
hard-coded to 200 in the domain.h file. This is simply too low for
most of my bigger examples. I think it would be quite usefull in the
future to give an additional option for CLooG to set the number of
MAX_RAYS without having to re-compile CLooG.

At the moment, I try to get my example done, CLooG has been running
now for about 50min on my Athlon-64 with 2GB physical and 4GB swap
memory (of which CLooG is using 2.5GB at the moment). 
I doubt that it will finish with a result, but that is simply because
the example seems to be too difficult (although I still don't see how
I could reduce it...).

Thanks again for the GMP-version! 

greetings,
Michael


