-
Notifications
You must be signed in to change notification settings - Fork 145
Conference call notes 20250423
Kenneth Hoste edited this page Apr 23, 2025
·
1 revision
(back to Conference calls)
Notes on the 268th EasyBuild conference call, Wednesday 23 April 2025 (08:00 UTC / 10:00 CEST)
List of attendees (17):
- Alex Domingo (Vrije Universiteit Brussel, Belgium
- Jasper Grimm (University of York, UK)
- Alexander Grund (ZIH, Dresden, Germany)
- Leonardo Honfi Camilo (Wageningen University, The Netherlands)
- Kenneth Hoste (HPC-UGent, Belgium)
- Georgios Kafanas (University of Luxembourg)
- Kurt Lust (UAntwerpen / LUMI)
- Sam Moors (Vrije Universiteit Brussel, Belgium)
- Alan O'Cais (CECAM)
- Mikael Öhman (Chalmers University of Technology, Sweden)
- Jure Pečar (EMBL, Germany)
- Andrea Piserchia (E4)
- Jan Reuter (JSC, Germany)
- Jörg Saßmannshausen (Imperial College London, UK)
- Roberto Scipioni (Red Oak Consulting, UK)
- Alain van Hoof (TU Eindhoven, Netherlands)
- Cintia Willemyns (Vrije Universiteit Brussel, Belgium)
- overview of recent developments
- 2025a update of common toolchains
- Q&A
- latest EasyBuild release: EasyBuild v5.0.0 (18 March 2025) 🥳
- next (stable) EasyBuild release: EasyBuid v5.0.1 (soon, days-not-weeks)
- structural merge sprints to be planned (for easyconfigs repo)
- once a month, fixed day?
- pairing up of maintainer with frequent contributor?
- "
release after 5.0.0
" milestones are being renamed to "5.0.1
" + some issues/PRs are being moved to new "release after 5.0.1
" milestone
(changed made in PRs marked with *
are included latest EasyBuild stable release)
-
docs (merged PRs)
- ...
-
framework (merged PRs)
-
easyblocks (merged PRs)
- bug fixes
-
enhancements
- Add option to impi easyblock to rebuild Fortran 2008 bindings (PR #3678)
- used for impi 2021.15.0, see easyconfigs PR #22657
- Add option to impi easyblock to rebuild Fortran 2008 bindings (PR #3678)
-
updates
- Update easyblock for AEDT 2025R1 (PR #3644)
-
changes
- ...
-
new easyblocks
- ...
-
code cleanup
- ...
-
CI
- ...
-
easyconfigs (merged PRs)
- XXX easyconfig PRs were merged since last conf call
- bug fixes/reports
-
enhancements
- ...
-
(noteworthy) new software
- ...
-
noteworthy software updates
- ...
-
cleanup
- ...
-
changes
- ...
-
docs (open PRs + issues)
- there still are some gaps in documentation that covers changed in EasyBuild 5.0.0
-
framework (open PRs + issues)
-
bug fixes
- Regression:
CMAKE_LIBRARY_PATH=lib
added to modulefiles (issue #4844 )
- Regression:
- enhancements
-
code cleanup
- ...
-
changes
- Deprecation of checksum types should be done on the EasyConfig level instead of during computation (issue #4849)
- replace deprecation warning about reproducible tarballs on Python older than 3.9 to a regular warning (PR #4852)
-
bug fixes
-
easyblocks (open PRs + issues)
- bug fixes/reports
- run
pip check
only once for PythonBundle (PR #3432) - Avoid failure in Python package installation and sanity check when
$PIP_REQUIRE_VIRTUALENV
is set (PR #3460) - Fix access to unassigned variable when installing Python < 3.9.1 with using sysroot (PR #3686)
- fix using LLVM easyblock for bundle component + don't build GO tests by default for LLVM versions < 16 (PR #3690)
- also other PR for LLVM easyblock by Jan (PR #3675)
- mostly relevant for LLVM 19.x + 20.x (no merged easyconfigs yet)
- also other PR for LLVM easyblock by Jan (PR #3675)
- Set Rust channel to stable to prevent usage of rust-lld as default linker (PR #3691)
- run
- enhancements
-
updates
- ...
-
changes
- ...
-
code cleanup
- Remove Python2 style
super()
-calls and fixExtensionEasyBlock
(PR #3681)
- Remove Python2 style
-
new easyblocks
- ...
- bug fixes/reports
-
easyconfigs (open PRs + issues)
-
~1.1k open easyconfig PRs
- we're (over)due for a cleanup of backlog of old/stale easyconfigs PRs
- closing PRs that are only using unsupported/deprecated toolchains, see https://docs.easybuild.io/policies/toolchains/#current-situation
-
bug fixes/reports
- ...
-
enhancements
- ...
-
(noteworthy) new software
- ...
- software updates
- ...
-
changes
- add easybuild package to EasyBuild/5.0.0 (PR #22774)
- see also change to
EasyBuildMeta
easyblock in easyblocks PR #3700
- see also change to
- add easybuild package to EasyBuild/5.0.0 (PR #22774)
-
~1.1k open easyconfig PRs
- (
2024b
is skipped to catch up with original schedule for defining common toolchain versions)- EasyBuild v5.0 is ideal excuse for that break in continuity...
- GCC 14.2 as a base (see easyconfigs PR #21114)
- we should define candidate toolchains using latest version of all components
-
foss/2025a
:- easyconfigs using
GCCcore/14.2.0
toolchain available for Python 3.13.1, Perl + Perl-bundle-CPAN 5.40.0 - easyconfigs PR #22125 merged for FlexiBLAS, OpenBLAS, BLIS
- (easyconfigs PR #22349) merged for
cryptography
, required forPython-bundle-PyPI
, after fixing issue with RPATH linking of Rust crates -
gfbf/2025.02
+ FFTW merged in easyconfigs PR #22126
- easyconfigs using
-
intel/2025a
:- updates for
intel-compilers
andimpi
merged in PR #22657 - TODO: imkl
- separate package for
intel-shmem
, doesn't really belong inintel
toolchain
- updates for
- Alex: issue opened for NVHPC-based toolchain (framework issue #4853)
- options like
-Mscalapack
assume that NVHPC is installed with the OpenMPI that comes with it - adopting NVHPC "batteries included" as proper toolchain EasyBuild opens up a lot of questions
- unclear how OpenMPI that comes with was configured
- HPC-X is customized fork of OpenMPI (see https://developer.nvidia.com/networking/hpc-x)
- current NVHPC easyconfigs should be renamed to
nvidia-compilers
, since they don't provide full NVHPC SDK- makes sense to only do this for new versions of NVHPC, and require sufficiently recent version in toolchain support for NVHPC
- Alex will look into full NVHPC toolchain
- main motivator currently is VASP
- Jörg has also been looking into this
- shadowing self-provided (ABI-compatible) OpenMPI could be an option?
- trick we pull in EESSI to allow injecting MPI libraries could be helpful here
- performance tests done by Alexander @ TU Dresden show that self-compiled OpenMPI on top of NVHPC was a bit faster than the one included with NVHPC
- unclear if that was for OpenMPI or HPC-X
- tweaks done by NVIDIA may only be relevant for GPU-to-GPU communication?
- options like