Conference call notes 20200902 - easybuilders/easybuild GitHub Wiki

(back to Conference calls)

Notes on the 155th EasyBuild conference call, Wednesday September 2nd 2020 (08:00 UTC - 10:00 CEST)

Attendees

Alphabetical list of attendees (8):

  • Simon Branford (Univ. of Birmingham, UK)
  • Kenneth Hoste (HPC-UGent, Belgium)
  • Alex Domingo (VUB, Belgium)
  • Terje Kvernes (University of Oslo, Norway)
  • Alan O'Cais (Jülich Supercomputing Centre, Germany)
  • Jörg Saßmannshausen (NIHR Biomedical Research Centre, UK)
  • Åke Sandgren (Umeå University, Sweden)
  • Lars Viklund (Umeå University, Sweden)

Agenda

  • outlook to next EasyBuild release (v4.3.0)
  • 2020b common toolchains
  • Q&A

Recent developments

  • next release will be EasyBuild v4.3.0
  • recent changes
    • framework
      • bug fix for combination of --include-easyblocks and --include-easyblocks-from-pr (PR #3311)
      • add templates for CUDA compute capabilities (PR #3382)
        • %(cuda_compute_capabilities)s, %(cuda_sm_comma_sep)s, %(cuda_sm_space_sep)s
        • value is determined by --cuda-compute-capabilities EasyBuild configuration option or cuda_compute_capabilities easyconfig parameter value (in that order)
      • code style fixes to make all code pass flake8 check (PR #3416, PR #3417)
      • limit Travis to only test with Python 2.6 + Lmod 7 (PR #3418)
      • use one argument 'module swap' statements in Tcl modulefiles (PR #3397)
        • required to support Environment Modules 4.2.3+
        • mostly relevant for Cray* toolchain module files generated by EasyBuild
      • add --sysroot EasyBuild configuration option (PR #3419)
        • useful when installing software in a Gentoo Prefix environment (context: EESSI project)
      • fixes for RPATH wrappers
        • make RPATH wrapper script more robust by using 'python -E -s -S' to run rpath_args.py (PR #3422)
        • don't inject -Wl,-rpath options when '-x c++-header' compiler option is used (PR #)
      • don't use -mno-recip on Arm (aarch64) when precise toolchain option is enabled (PR #3425)
    • easyblocks
    • easyconfigs
      • add patch in h5py v2.8.0 + v2.9.0 easyconfigs to avoid MPI_Init on import h5py (PR #11118, PR #11119)
      • Kent_tools built from source (PR #11121)
      • use pip to install protobuf-python in 2019b toolchain (PR #11143)
        • requires re-downloading source tarball!
      • rename sdsl-lite to SDSL (to use one single name) (PR #11084)
      • enhance Java/11 wrapper to also support for aarch64/Arm (PR #)
      • RPATH related fixes:
        • update Meson build dep to 0.55.1 for GLib, X11, Mesa & co to fix aggressive RPATH stripping (PR #11178)
        • add fontconfig and bzip2 as direct dependencies for Qt5 to fix installation with --rpath (PR #11183)
        • fix failing 'make check' for MPFR 4.x when installing with 'eb --rpath' (PR #11187)
      • significant software updates
  • to merge soon:

2020b update of common toolchains

  • base: GCC 10.2 + binutils 2.35 (see PR #11056
  • foss:
    • OpenMPI 4.1.0 (RC is available, final release soon)
      • still no final 4.1.0 release yet
    • OpenBLAS 0.3.10 (updated from 0.3.9 in 2020a)
    • FFTW 3.3.8 (no updates)
  • intel:
    • compilers: 2020 update 2
    • MPI: 2019 update 8
    • MKL: 2020 update 2
    • is GCC 10.x supported officially as base compiler?
      • strictly according to release notes: no => "gcc 4.8.5 to gcc 9.0.1" and "binutils 2.27 to binutils 2.31.1", but that seems very conservative
      • Kenneth has contacted Intel guy, no reply yet...
    • PR with updated Perl easyconfig with GCCcore/10.2.0 coming soon

Q&A

  • Terje: segfault with yasm on aarch64: https://github.com/easybuilders/easybuild-easyconfigs/issues/11190
  • discussion: bringing down Boost, HDF5, netCDF to compiler-only level
    • we need a separate discussion on this, and involve Maxime
    • splitting things up will be confusing for users
      • hide serial versions by default (only used as deps)
      • "regular" named installations for users
    • Boost could a bundle of Boost.serial and Boost.MPI
    • we should work on this in a different branch