Hello all.
I followed the instructions in
http://alps.comp-phys.org/mediawiki/index.php/Download_and_install_ALPS_2
to compile and install ALPS2 on a Fedora 14 x86_64 machine, and did manage to install it, but there were a few caveats and 3 tests failed in the end, so I thought I'd share my experience and ask for help here.
Basically, what I did was get the tar.gz with boost (because f14 still has 1.44), unpacked, created the build dir and then installed missing dependencies (numpy, hdf5, lapack-devel, blas-devel, fftw-devel, sqlite-devel).
Some where not documented, like lpsolve and szip (from hdf site), and I had to look in CMakeCache.txt for the variable names, so if I didn't miss it, maybe this could be included in the docs?
After trying cmake a lot of times, I had to delete the build directory and create a fresh one, and run
cmake -D SZIP_INCLUDE_DIR:PATH=~/Downloads/szip-2.1/szip/include -D LPSolve_INCLUDE_DIR:PATH=/usr/include/lpsolve -D Boost_ROOT_DIR:PATH=~/Downloads/alps-2.0.0-r5363-src-with-boost/boost ~/Downloads/alps-2.0.0-r5363-src-with-boost/alps 2>&1 | tee cmake.log make make test
and torwards the end I got
90/92 Test #90: python_hlist_test ................***Failed 1.84 sec Start 91: python_mcdata 91/92 Test #91: python_mcdata ....................***Failed 0.19 sec Start 92: python_pyhdf5 92/92 Test #92: python_pyhdf5 ....................***Failed 0.18 sec
97% tests passed, 3 tests failed out of 92
I did "make install", but has anyone else had this problems? Could it be because Fedora 14 comes with python 2.7 ?
Thanks in advance, Cheers, Miguel
These tests can also fail if ALPs does not set up the Python test environment correctly. I would not worry about it. Just try to use the Python packages.
Matthias
On Mar 10, 2011, at 12:58 PM, Miguel Costa wrote:
Hello all.
I followed the instructions in
http://alps.comp-phys.org/mediawiki/index.php/Download_and_install_ALPS_2
to compile and install ALPS2 on a Fedora 14 x86_64 machine, and did manage to install it, but there were a few caveats and 3 tests failed in the end, so I thought I'd share my experience and ask for help here.
Basically, what I did was get the tar.gz with boost (because f14 still has 1.44), unpacked, created the build dir and then installed missing dependencies (numpy, hdf5, lapack-devel, blas-devel, fftw-devel, sqlite-devel).
Some where not documented, like lpsolve and szip (from hdf site), and I had to look in CMakeCache.txt for the variable names, so if I didn't miss it, maybe this could be included in the docs?
After trying cmake a lot of times, I had to delete the build directory and create a fresh one, and run
cmake -D SZIP_INCLUDE_DIR:PATH=~/Downloads/szip-2.1/szip/include -D LPSolve_INCLUDE_DIR:PATH=/usr/include/lpsolve -D Boost_ROOT_DIR:PATH=~/Downloads/alps-2.0.0-r5363-src-with-boost/boost ~/Downloads/alps-2.0.0-r5363-src-with-boost/alps 2>&1 | tee cmake.log make make test
and torwards the end I got
90/92 Test #90: python_hlist_test ................***Failed 1.84 sec Start 91: python_mcdata 91/92 Test #91: python_mcdata ....................***Failed 0.19 sec Start 92: python_pyhdf5 92/92 Test #92: python_pyhdf5 ....................***Failed 0.18 sec 97% tests passed, 3 tests failed out of 92
I did "make install", but has anyone else had this problems? Could it be because Fedora 14 comes with python 2.7 ?
Thanks in advance, Cheers, Miguel
On 03/10/2011 01:23 PM, Matthias Troyer wrote:
These tests can also fail if ALPs does not set up the Python test environment correctly. I would not worry about it. Just try to use the Python packages.
You're right - I had to set PYTHONPATH=$PYTHONPATH:/opt/alps/lib, and now all tests run fine, and I'm able to launch simulations from python.
I also wrote a small python class to parse the results of "loadEigenstateMeasurements" after a full diag and calculate thermodynamical averages at different temperatures from them, which in retrospect seems to be a very general use case - does Alps 2 have this functionality elsewhere?
Cheers, Miguel
On 03/16/2011 09:19 AM, Miguel Dias Costa wrote:
On 03/10/2011 01:23 PM, Matthias Troyer wrote:
These tests can also fail if ALPs does not set up the Python test environment correctly. I would not worry about it. Just try to use the Python packages.
You're right - I had to set PYTHONPATH=$PYTHONPATH:/opt/alps/lib, and now all tests run fine, and I'm able to launch simulations from python.
I also wrote a small python class to parse the results of "loadEigenstateMeasurements" after a full diag and calculate thermodynamical averages at different temperatures from them, which in retrospect seems to be a very general use case - does Alps 2 have this functionality elsewhere?
Please disregard this last part, I saw now that it's available as "pyalps.evaluateFulldiagVersusT".
Cheers, Miguel
comp-phys-alps-users@lists.phys.ethz.ch