-
Notifications
You must be signed in to change notification settings - Fork 144
Conference call notes 20201125
Kenneth Hoste edited this page Nov 25, 2020
·
3 revisions
(back to Conference calls)
Notes on the 161st EasyBuild conference call, Wednesday November 25th 2020 (9:00 UTC - 10:00 CET)
Alphabetical list of attendees (6):
- Sebatian Achilles (Jülich Supercomputing Centre, Germany)
- Simon Branford (University of Birmingham, UK)
- Kenneth Hoste (HPC-UGent, Belgium)
- Alan O'Cais (Jülich Supercomputing Centre, Germany)
- Jörg Saßmannshausen (NIHR Biomedical Research Centre, UK)
- Adam Huffman (Big Data Institute, Oxford, UK)
- recent developments
- easyconfigs for 2020b generation
- virtual EasyBuild User Meeting
- Q&A
- recent changes
- framework
-
easyblocks
-
bug fixes
- fix two bugs in GROMACS easyblock when using GCC & MKL for FFT and BLAS/LAPACK (PR #2212)
- always define
$FCCPP
in QuantumESPRESSO easyblock (PR #2221) - only configure Python with
--enable-optimizations
when compiling Python with (recent) GCC compiler PR #2228 - Don't try to patch newer versions of Bazel where the patches won't apply (PR #2249)
- fix sanity check for Boost with multithreading support (PR #2231)
-
enhancements
- enhance extension filter for Python packages to ignore user site dir when for checking Python package extensions (PR #2224)
- add support for building Clang with OpenMP offload support (PR #2229)
- add support for building GCC with GPU offloading support via
nvptx
(PR #2235) - evaluate
numpy
test results (rather than ignoring them) (PR #2238)
-
bug fixes
- easyconfigs
- to merge soon
-
framework (v4.3.2 milestone)
-
bug fixes
- fix combination of
--copy-eb
and--from-pr
(PRs #3473 + #3482)- needs more love after other PRs that were split off have been merged...
- crash in
apply_regex_substitutions
when passing list of paths to patch (issue #3493) - check if PR is already merged in
--merge-pr
(PR #3502)
- fix combination of
- enhancements
-
bug fixes
-
easyblocks (v4.3.2 milestone)
- bug fixes
-
enhancements
- enhance OpenBLAS easyblock to make it aware of
optarch
(PR #1946) - improve config checks for Boost.MPI and Boost.Python and fix sanity checks of MT libs (PR #2167)
- run
motorBike
tutorial case for recent (community) OpenFOAM versions (PR #2201)- needs testing...
- enhance OpenMPI easyblock to catch any
--with-ucx*
configure options (PR #2230
- enhance OpenBLAS easyblock to make it aware of
-
new software
- custom easyblock for
Metagenome-Atlas
(PR #2219)
- custom easyblock for
-
easyconfigs (v4.3.2 milestone)
- bug fixes
- enhancements
- new software *
- software updates
-
framework (v4.3.2 milestone)
- done (to be included in EasyBuild v4.3.2)
-
TODO:
-
intelcuda/2020b
(PR #11730) - Boost 1.74.0 (without MPI) (PR #11642)
- CP2K 7.1
- HDF5, netCDF, netCDF-Fortran (PR #11639)
- FFmpeg (PR #11671)
- FFTW 3.3.8 with
intel/2020b
(PR #11664) - GSL 2.6 (PR #11650)
- Libint 2.6.0 (PR #11665)
- libxc 4.3.4 (PR #11666)
- libxsmm 1.16.1 (PR #11667)
- METIS 5.1.0 (PR #11672)
- OpenFOAM 8 (PR pending)
- OSU-Micro-Benchmarks 5.6.3 (PR #11674)
- Qt5 5.15.1 (PR #11555)
- requires OpenSSL 1.1.x (CentOS 7 has OpenSSL 1.0.x)...
- PyQt5 5.15.5 (PR #11563)
- ParaView 5.8.1 (PR pending)
- PLUMED 2.6.2 (PR #11668)
- R 4.0.3 (PR #11663)
- failing test reports?!
- SCOTCH 6.1.0 (PR #11673)
-
SciPy-bundle
2020.11 (PR #11629)- blocked by Mikael's failing tests?
- WRF 3.9.1.1 (PR #11675)
- failing test reports on
generoso
(CentOS 8)?!
- failing test reports on
-
- next sprint (still) being planned...
- see https://github.com/easybuilders/easybuild/wiki/6th-EasyBuild-User-Meeting
- CernVM-FS tutorial
- ReFrame tutorial
- Singularity demo or workshop by Jörg
- talk to introduce Singularity (Mon)
- multiple short workshops
- Alan: should
rstudio
be renamed?- only includes server component, shouldn't be lowercase
- probably
RStudio-Server
- Simon: fairly recent Intel MPI on CentOS 8: segfaults...
- at HPC-UGent the
impi
inintel/2019b
was bumped to the one used inintel/2020a
- at HPC-UGent the
- Adam: should we start thinking about
2021a
toolchains - Adam: should we promote adding of sanity check commands more?
- require it in tests?
- Sebastian: which math libraries do people use on AMD Rome systems?
- magic
I_MKL_DEBUG
environment variable, but only works with olderimkl
(2019.x version)- also doesn't give best performance when compared to BLIS
- toolchain with BLIS
- ~10% faster compared to imkl
- pure
dgemm
was ~40% faster - only double precision kernels are well optimized for AMD Rome in recent imkl
- in latest imkl version also provides fast-ish single precision kernels (but not as good as BLIS)
- we should follow up on this in a dedicated conf call on EasyBuild toolchains in AMD systems
- magic