3rd EasyBuild hackathon meeting minutes day 2 - easybuilders/easybuild GitHub Wiki

(Tuesday Mar. 12th 2013, 10am-6pm)

The second day of the 3rd EasyBuild hackathon was the first day of the actual 2-day hackathon. The attendees got to know EasyBuild better by installing it and getting hands-on experience on building software. Some already even got to adding support for new software by producing their own easyconfig files, or started fixing outdated easyblocks/easyconfigs for software they care about.

These notes were mainly taken by Kenneth and Jens, with contributions by Fotis.

Attendees

  • Kenneth Hoste (HPC-UGent, EasyBuild developer and release manager)
  • Jens Timmerman (HPC-UGent, EasyBuild developer)
  • Fotis Georgatos (University of Luxembourg, HPC sysadmin and active contributor)
  • Thekla Loizou (The Cyprus Institute, HPC user support)
  • George Tsouloupas (The Cyprus Institute, HPC sysadmin and user support)
  • Mohamed Gafaar (Bibliotheca Alexandrina, HPC sysadmin/user support)
  • Dina Mahmoud Ibrahim (Cairo University, HPC sysadmin/user support)
  • Alan O'Cais (Jülich Supercomputing Centre, HPC user support & LinkSCEEM)
  • Alexander Schnurpfeil (Jülich Supercomputing Center, HPC user suppot)
  • Nicolas Kanaris (The Cyprus Institute, HPC user (OpenFOAM))
  • George Fanourgakis (The Cyprus Institute, HPC user, molecular dynamics)
  • Stelios Erotokritou (The Cyprus Institute, HPC user support/PRACE)

Program

  • [10am-6pm] 1st day of actual hackathon
  • [6pm-8.00pm] aftermath: discussathon with George T., Fotis, Jens T. and Kenneth

Discussion notes

  • [Fotis]
  • discuss a way to make modules conflict for compilers, MPI, BLAS/LAPACK, ...(?)
  • ie. it makes not much sense to have 2 different MPI stacks loaded together, better spit a warning
  • KH?: one way would be to add ghost modules being loaded as e.g. a compiler is needed
  • COMPILER/GCC-4.6.3
  • add extra conflict COMPILER line in GCC, icc, ifort modules
  • problems: * shows up in module list * icc vs ifort needs special attention, they're both compilers
    • C_COMPILER vs FORTRAN_COMPILER? * no way of easily combining GCC and Intel compilers anymore if you wanted to

Hackathon notes

  • George F.: Ferret
  • ported outdated easyblock and easyconfig for Ferret
  • patch out hardcoded from *.mk file used by Ferret
  • add missing dependencies in easyconfig (zlib, ncurses, cURL, ...)
  • Nicolas
  • OpenFOAM / pyFOAM
  • look into OpenFOAM 'extensions' (commercial ThirdParty stuff)
  • Alexander
  • getting EasyBuild to work on JUROPA (SLES/x86-64)
  • with Tcl environment modules
  • interested in building NWChem
  • Dina
  • work on training exercises and Getting Started wiki page (gzip, ...)
  • build GCC in VM on her laptop
  • build OpenMPI with dummy toolchain during waiting
  • Alan
  • building basic libraries
  • look into UNITE and its dependencies
  • George & Fotis
  • CUDA toolkit 5.0.35; DONE.
  • GROMACS v4.6 (w/o CUDA); DONE.
  • FFTW 3.3.3; DONE.
  • OpenMPI v1.7 for DMA to GPU support
  • only 1.7rc8 is available; DONE.
  • Fortran libraries were renamed in this version * libmpi_f77 -> libmpi_mpifh * libmpi_f90 -> libmpi_usempi
  • requires GCC newer than v4.8 (which hasn't been released yet) * see http://www.open-mpi.org/software/ompi/versions/ (bottom of page)
The incorrect interface was removed in Open MPI v1.7.

To be clear: applications that use the old/incorrect MPI_SCATTERV
binding will no longer be able to compile properly (*).  Developers
must fix their applications or use an older version of Open MPI.

(*) Note that using this incorrect MPI_SCATTERV interface will not be
    recongized in v1.7 if you are using gfortran (as of gfortran
    v4.8).

    This is because gfortran <=v4.8 does not (yet) have the support
    Open MPI needs for its new, full-featured "mpi" and "mpi_f08"
    modules.  Hence, Open MPI falls back to the same "mpi" module from
    the v1.6 series, but the "large" size of that module -- which
    contains the MPI_SCATTERV interface -- been disabled because it is
    broken.  Further, this "large" sized (old) "mpi" module has been
    deemed unworthy of fixing because it has been wholly replaced by a
    new, full-featured "mpi" module.  We anticipate supporting
    gfortran in the new, full-featured module in the future.
  • OpenMPI v1.6.3 & v1.6.4 (Fotis); DONE.
  • with slightly more relaxed sanity check paths
  • goolf 1.4.10; ~DONE (deviations: FFTW is the default version, so far)
  • GCC 4.7.2
  • OpenMPI 1.6.4
  • OpenBLAS 0.2.6
  • LAPACK 3.4.2
  • FFTW 3.3.3 (single/double)
  • ScaLAPACK 2.0.2
  • goolf 1.5.10; same as above, except with OpenMPI v1.7rc8; existing just as a prototype really
  • gmolf: no much progress on that yet
  • goolfc
  • CUDA toolkit 5.0.35
  • GROMACS v4.6 w/ CUDA support
  • Jens T.
  • help out with several people
  • Kenneth
  • write easyconfig for OpenBLAS
  • help out with several people