Dear Mateusz,
if you already have a system-installed HDF5 then just use that. Otherwise I am confused why the HDF5_C_INCLUDE_DIR variable should be relevant for your Fortran codes. One way to work around that seems to be to add the proper include path to the Fortran_FLAGS variable. That variable comes before the other includes in the command line.
Matthias
On 21 Dec 2011, at 22:20, Mateusz Łącki wrote:
Dear Matthias, I set "hdf5 C include directory" to one containing the proper hdf5.mod but somehow /usr/local/include got priority. The problem is that the module in /usr/local/lib is in wrong version and compiling it with gfortran gives "wrong module version error".
Technically i used "ccmake .", but this is the same as editing CMakeCache.txt if I understand correctly. For sure I have:
HDF5_C_INCLUDE_DIR:PATH=/opt/hdf5-1.8.8-gnu/include
in the file, but still /usr/local/lib had priority.
To succesfully compile I edited the filesthat I mentioned:
applications/dmrg/tebd/CMakeFiles/CMakeDirectoryInformation.cmake
(butthis one does not really matter If I understand correctly)
but more important are *.make files. I found a file flags.make - it containt whole thing: gfortran with proper -I 's and I set the correct order there - I moved -I /usr/local/lib to last position (originally it was at the very beginning).
This lets the code compile and pass all tests. But this was extremely ugly.
I just got an idea: perhaps it is possible to compile with gfortran 4.6 and gcc 4.4... I did not check that.
Regards, Mateusz
Have you tried editing the CMakeCache.txt file? Or to use the version in ust/local/include?
On 21 Dec 2011, at 19:37, Mateusz Łącki wrote:
Dear All I have fond in the file " applications/dmrg/tebd/CMakeFiles/CMakeDirectoryInformation.cmake "
SET(CMAKE_C_INCLUDE_PATH "/usr/local/include" "/home/lacki/opt/hdf5-1.8.8-gnu/include"
and this is a little bit troublesome as gfortran generated modules are version specific. Therefore "/usr/local/include/hdf5.mod" shadows "/home/lacki/opt/hdf5-1.8.8-gnu/include/hdf5.mod" and "wrong module version" error appears.
This file is automatically generated by cmake.
Is there some option that can be set so that the order in the file was reversed:
SET(CMAKE_C_INCLUDE_PATH "/home/lacki/opt/hdf5-1.8.8-gnu/include" "/usr/local/include"
???
Regards, Mateusz Łącki
Hello Mattias,
Yes, I need basis and all eigenvectors classified by magnetization.
Olya.
2011/12/21 Matthias Troyer troyer@phys.ethz.ch Hi Olya,
You will probably also need the basis since the vectors without basis are useless. And do you need all eigenvectors at each magnetization or just a few?
Matthias
On 21.12.2011, at 18:10, Shevchuk Olya oluska2004@gmail.com wrote:
Dear Mattias,
I need them exactly for each magnetization to apply to my transfer matrix in order to get the eigenvalues of transfer matrix. I don't need any other quantity just eigenstates. I am not using too big systems, L=12 is the biggest.
Olya
2011/12/21 Matthias Troyer troyer@phys.ethz.ch Hi Olya,
The eigenvectors would be huge and typically don't help you much. Just measure whatever you want to measure in the eigenstates. What do you need the vectors for?
Matthias
On Dec 21, 2011, at 5:41 PM, Shevchuk Olya wrote:
Hello,
I am trying to run fulldiag for Heisenberg model. What I need is eigenstates for each Sz. I read in the parameter description that by default fulldiag or sparsearray does not store eigenvectors. Can you please help me to figure it out how to store them?
thanks in advance, Olya.