Martin Rohde wrote:
It was indeed something here and not the PPL library. Some problems arised because we installed GMP at an 'unusual' location. We did the following for configuring:
export CFLAGS="-Wl,-rpath /opt/gmp-4.1/lib" export CXXFLAGS="-Wl,-rpath /opt/gmp-4.1/lib" ./configure \ --prefix=/opt/ppl-0.4pre2 \ --with-gmp-includes=/opt/gmp-4.1/include \ --with-gmp-dir=/opt/gmp-4.1/lib
For including the ppl AND gmp, we used the following 10-liner:
g++ test_program.cc \ -I /opt/ppl-0.4pre2/include \ -I /opt/gmp-4.1/include \ -L /opt/gmp-4.1/lib \ -lgmp \ -lgmpxx \ -L /opt/ppl-0.4pre2/lib \ -lppl \ -Wl,-rpath /opt/gmp-4.1/lib \ -Wl,-rpath /opt/ppl-0.4pre2/lib
I don't know whether this is a general problem for people who will use the library in the future, or a problem that only relates to our installation/configuration. If this is a more general problem, maybe the above can be included in the INSTALL (or README) file.
Dear Martin,
sorry for replying so late. We have been very busy finishing version 0.4 of the library (released on July 1st).
Yes, I believe it is worth to explain users how the place where they install libraries can affect the way they use the libraries. It is, in fact, a common source of headaches. However, I don't understand why you assign the environment variables before configuration
export CFLAGS="-Wl,-rpath /opt/gmp-4.1/lib" export CXXFLAGS="-Wl,-rpath /opt/gmp-4.1/lib"
What happens if you don't do that? Would you mind trying with the newly released version 0.4?
Right now, I'm going to study the manual. The tests I performed yesterday looked very promising. I mainly use the library to calculate the intersection of two polyhedra. The polyhedra are defined in R^3 (so real numbers), but the constraints and generators can easily be transformed into large integers without losing accuracy (this is the reason why using version 0.3, using input gmp numbers):
Please let us know more about your application. We may list it on our web pages (you are already listed ;-) All the best
Roberto
P.S. I don't know if you already did that but, if you plan to use the library, I suggest you subscribe to the ppl-announce mailing list. As you will see from the archives it is very quiet and 100% useful contents. The URI is http://www.cs.unipr.it/mailman/listinfo/ppl-announce