Hello Yann, I got this email last week just before submiting correctly my email to the easybuild mailing list. I mention correctly because before that try I sent an email directly to the owners and they told me I was not subscribed in the mailing list. After that, I sent a proper email to the mailing list that I will attach here again, but before I wanted to ask about the mail you sent to me, because I do not know if it is an answer for my previous email or is something related to another problem. The thing that also is maybe a coincidence is that the name of the person you sent the email to first, Loris Bennet, is the same name of the person that is tha administrator of the HPC cluster where I am working. So now I attach the email with my doubt about installing a program in the HPC:
Hello dear easybuild staff, My name is Jorge Torres and I am phD student at the FU Berlin in the physics department. I am trying to install some software in an HPC cluster which works under the easybuild plattform. I have subscribed to the mailing list but I am not sure how it works so I send this email to you and also attach it to the owners to see if I can get some advice. The software is called SHARC and is a free software that works in linux. Its repository page is https://github.com/sharc-md/sharc/blob/master/INSTALL Also, I got a message from the administrator from the HPC center, which says : "The instructions to install the program are a bit vague, e.g. edit the Makefile by adjusting the USE_PYSHARC USE_COMPILER USE_LIBS ANACONDA variables. EasyBuild is about automation, but if there is not obvious how to install the software by hand, it will not be easy to automate. The program also uses Anaconda to provide dependencies such as HDF5, but with EasyBuild such dependencies are usually packaged as separate modules. This will also make it harder to integrate into EasyBuild. In fact, I would assume that a separate EasyBlock would be necessary, which is a part of the EasyBuild framework which encapsulates the specific build mechanism for a piece of software. Writing such an EasyBlock is not trivial. It is much better if software uses a standard installation procedure, since EasyBlock for such standard mechanisms already exist." I have installed this program in my computer but since the program run in parallel with the quantum chemistry software ORCA, the calculations are very slow. I was wondering if you have received any question on this software before or on any similar program which needs to be installed by hand. If yes, is there any instruction for easy configuration? I tried to install it too but at some point I need to have some administrator permission to install some program by sudo. This is the main issue I got: > error: cannot find -lblas > error: cannot find -llapack This error I also got on my computer but then I could add the libraries by sudo apt-get install libblas-dev liblapack-dev Then I got an answer of the administrator: "No users on the HPC system can run 'sudo' and the 'apt-get' would be wrong anyway, because the operating system, CentOS, is not Debian-based, but RedHat-based, and so the package manager is not 'apt' but rather 'yum'." Then there is a makefile whit a command line that needs to be changed from: LD= -lblas -llapack -lfftw3 -lgfortran $(NETCDF_LIB) to LD= -lopenblas -lscalapack -lfftw3 -lgfortran $(NETCDF_LIB) since the 'foss' toolchain contains OpenBLAS rather than BLAS and ScaLAPACK rather than LAPACK. Also, the administrator told me that he needs to install the Ancaconda packages hdf5, hdf5_hl, netcdf, mfhdf, df, jpeg, gcc Some of these packages are also available via EasyBuild, namely - HDF5 - netCDF - GCC (included in foss) Via conda, the administrator was able to install - hdf - jpeg but he wasn't able to install - hdf5_hl - netcdf - mfhdf - df via conda. He said that he don't know what channel or channels has/have to be set for this. Thus, he cannot say whether installing this way will work. For the moment this is my query and I will be very greatful if you can reply to me if there is any solution. For your attention thank you very much. best Jorge Torres > Hi, > > If I'm not mistaken the meaning is that now EB infere the Python version > and as you are requiring the "default" Python vrersion with this > compiler, there is no need to specify it. > > I have no clue if CPLEX needs Python. > > versionsuffix could be used for example if you compile a variant of a > software. > > Best > > Yann > > Le 09.08.2022 à 11:46, Loris Bennett a écrit : >> Hi, >> >> For my PR #1003 I am getting the error: >> >> AssertionError: '-Python-%(pyver)s' should no longer be included in >> versionsuffix in CPLEX-22.10-GCCcore-11.2.0-Python-3.9.6.eb >> >> Following the model of the last version >> >> CPLEX-12.10-GCCcore-8.3.0-Python-3.7.4.eb >> >> the new EC is named: >> >> CPLEX-22.10-GCCcore-11.2.0-Python-3.9.6.eb >> >> In the EC I have >> >> versionsuffix = '-Python-%(pyver)s' >> >> Is the error telling me that Python versions are not what >> 'versionsuffix' should be used for? >> >> In addition it is not clear to me whether CPLEX does depend on Python. >> It certain has a Python API (and one for C/C++, Java) but I could find >> no mention of a specific version. >> >> So my questions are: >> >> 1. What should 'versionsuffix' be used for? >> 2. Does CPLEX really need Python as a dependency? >> >> Cheers, >> >> Loris >> > -- > > Logo UNIGE Yann Sagon > Référent HPC > > Division du système et des technologies de l'information et de la > communication > Université de Genève | 24 rue Général-Dufour > Tél 022 379 77 37 | Bureau 151 > > www.unige.ch/stic <http://www.unige.ch/stic> >
