-
Notifications
You must be signed in to change notification settings - Fork 144
Conference call notes 20220928
(back to Conference calls)
Notes on the 206th EasyBuild conference call, Wednesday 28 September 2022 (15:00 UTC)
List of attendees (15):
-
Mikael Öhman (Chalmers University of Technology, Sweden)
-
Bart Oldeman (Digital Research Alliance of Canada)
-
Jörg Saßmannshausen (Imperial College London, UK)
-
Kurt
-
Thomas Hayward-Schneider (Max Planck Institute for Plasma Physics, Germany)
-
Sam Moors (Vrije Universiteit Brussel, Belgium
-
Sebastien Moretti (SIB, Switzerland)
-
Jasper Grimm (University of York, UK)
-
Adam
-
Alan
-
Sebastian Achilles
-
Casper
-
Simon Branford (Univ. of Birmingham, UK)
-
Em Dragowsky (Case Western Reserve University, Ohio, US)
-
Kenneth Hoste (HPC-UGent, Belgium)
-
Abhishek Mukherjee (New Jersey Institute of Technology, US)
-
Åke Sandgren (Umeå University, Sweden)
-
Alexandre Strube (JSC, Germany)
-
Davide Vanzo (Microsoft Azure)
-
Matthew West (University of Exeter, UK)
- overview of recent developments + outlook to next EasyBuild release
- how to communicate that some software packages have been replaced going forward (pkg-config, GTK+, ...)
- NVHPC toolchains
- FFmpeg and GLEW dep conflicts
- backwards-incompatible changes being considered for EasyBuild v5.0
- Q&A
- release timeline
- latest release: EasyBuild v4.6.0 (8 July 2022)
- ETA next release (EasyBuild v4.6.1): ASAP
- progress on problems with older Intel MPI versions on RHEL 8.6 (see issue #15651)
- Should be fixed with updating to
kernel-4.18.0-372.26.1.el8_6.x86_64.rpm
- Should be fixed with updating to
- recent changes
-
framework
- bug fixes
- enhancements
-
changes
- ...
-
easyblocks
-
bug fixes
- Fix non-cuda builds of suitesparse PR #2790
- don't enable building of
ld.gold
when installing binutils on a RISC-V system + don't configure GCC to use gold as default linker on a RISC-V system (PR #2780) - add and fix patches for PyTorch 1.9.0 on POWER (PR #15919)
- use versioned syms in system level ncurses (PR #16064)
- fix checksum for Stacks v2.62 (due to silent re-release without version bump) (PR #16134)
- fix libsanitzer for glibc 2.36 to build GCCcore 11.3.0 (PR #16145)
- ...
-
updates
- Add python support to EMSF PR R2789
-
changes
- ...
- new software
- ...
-
bug fixes
-
easyconfigs
- Probably a whole bunch of merge easyconfigs since last conf call
- bug fixes
-
enhancements
- use
det_cmake_version
function to determineCMake
version inCMakeMake
generic easyblock (PR 2772)
- use
-
(noteworthy) new software
- see also overview at https://github.com/easybuilders/easybuild-easyconfigs/issues?q=is%3Amerged+label%3Anew
- ...
-
noteworthy software updates
- ...
- changes
-
framework
- work in progress
-
framework
-
reported bugs / bug fixes
-
- issues with -xHOST/-march=native on AMD issue #3793
-
-
enhancements
- add support for easystack file that contains easyconfig filenames + implement parsing of configuration options (PR #4021)
- adding support for the use of
--from-pr
_and other options) in easystack files (WIP PR #4057)
- changes
-
reported bugs / bug fixes
-
easyblocks
- bug reports/fixes
- ...
- enhancements
- ...
-
updates
- ...
- new software
- ...
-
changes
- ...
- bug reports/fixes
-
easyconfigs
- close to 700 open easyconfig PRs...
- bug fixes
- new software
- noteworthy software updates
- changes
- inconsistent use of versionsuffix for Java (issue #16038)
-
framework
- SYSTEM variable has been usable for some time, just the CI tests that were not happy with it. Should we now start enforcing it?
How to communicate that some software packages have been replaced going forward (pkg-config, GTK+, ...)
- make easyconfigs test suite complain in PRs
- also add comment in easyconfigs?
- README in easyconfigs directory for that software
- also in documentation? (and refer to that from everywhere)
- other cases: util-linux, texinfo (makeinfo), numpy/scipy/pandas now in SciPy-bundle, ...
- cfr. https://github.com/easybuilders/easybuild-easyconfigs/issues/16066
- Add CUDA as versionsuffix to NVHPC easyconfigs, to make it clear which CUDA is being used
- drop
nvompic
in favor ofnvompi
+CUDA
(similar to recentfoss
+ CUDA) - should current
NVHPC
be renamed (tonv-compilers
) so we can usenvhpc
as toolchain name for a full toolchain? - or use
nvoff
as full toolchain name?-
nv
for NVHPC,o
forOpenMPI
,ff
for FlexiBLAS + FFTW
-
- should
NVHPC
that is used in a toolchain be namednv-compilers
, while also keepingNVHPC
modules stand-alone?
- ...