Re: [easybuild] Dependency over modaltsoftname

Hi, I figured it out. There is a strict naming in easyconfig names for dependencies we have to adhere to: name-version-toolchain-versionsuffix. So the dep is of the form: ('HDF5', '1.8.18', '-serial', ('iccifort', '2016.4.258')) (the toolchain can be optional here though) and in the correspond

Re: [easybuild] GCC, Haswell and RHEL 6 pain

Hi, just to note there is a typo here: eb GCC-4.8.2.eb --try-amend="dependencies=,('binutituls', '2.25')" this should be eb GCC-4.8.2.eb --try-amend="dependencies=,('binutils', '2.25')" I think at the moment it is easiest to use a checkout of the develop branch of the easybuild-easyconfigs git

[easybuild] Question about lmod files, and unloading.

Hi, I am trying to set up a hierarchical module structure with Easybuild 2.3.0 and Lmod 6.0.9. I do however observe that module load foss/2015b followed by module unload foss/2015b only unloads the foss module but the dependencies are still there. I locally built an iomkl/2015b toolchain and

Re: [easybuild] Question about lmod files, and unloading.

Hi Alan, Kenneth, great, the combination of module-naming-scheme=ToolchainMNS (and listing subtoolchains as dependencies in toolchains) and recursive-module-unload = 1 makes for a system that I like much better and will be easier to explain to our users. Thanks a lot! Note: seeing that the other

Re: [easybuild] GNU toolchain

Just my 2 cents: This is really about a bootstrap procedure for what is presently called the GNU/4.9.3-2.25 module (proposed to be just GCC/4.9.3-binutils-2.25). I think one issue is now that this procedure is user-exposed: all the bits and pieces used in the bootstrap stay there by default, and

Re: [easybuild] foss module not cleaning up

Hi Joachim, I had the same question, and was answered here: https://lists.ugent.be/wws/arc/easybuild/2015-09/msg00055.html The easiest is to set recursive-module-unload = 1 in your ~/.config/easybuild/config.cfg file. (I wish the mailing list archives were more easily searched, e.g. by google,

Re: [easybuild] How to promote a package to a toolchain?

Hi Joachim, (disclaimer: I contributed the PGI easyconfig and easyblock). Kenneth decided to ship the PGI easyconfigs/block just in time for 2.6.0, but left out the toolchain with a milestone of 2.7.0. You can find it here: https://github.com/hpcugent/easybuild-framework/pull/1342 Once that's

Re: [easybuild] configure flag for openmpi

Hi Yann, you'll end up creating your own stack of easyconfig files in a seperate location instead of editing the ones shipped with easybuild directly. Remember that the shipped easyconfigs are examples, if you can use them as is (or with ---try-toolchain=xxx), great, but often you end up needing

Re: [easybuild] next EasyBuild conf call: Wed April 13th 2016

Hallo Kenneth, as I see PR 1342 mentioned I'd like to join the conf call tomorrow. Regards, Bart On 12 April 2016 at 09:00, Kenneth Hoste wrote: > Dear EasyBuilders, > > The next EasyBuild conf call is planned for tomorrow, Wednesday April 12th > 2016, 5pm CET, > see

Re: [easybuild] Compiling with Intel Xeon Phi (via EasyBuild)

Hello Adrian, this is on my to do list to check out (we have 50 nodes with 2 Xeon Phis each here). I am not sure who else has done it so far though. A bare minimum would be to get the Intel compiler modules and MKL to set MIC_LD_LIBRARY_PATH and SINK_LD_LIBRARY_PATH in addition to

Re: [easybuild] R 3.3.1 with Rmpi

HI Yann, the orte_run message (Error obtaining unique transport key from ORTE (orte_precondition_transports not present in the environment) is normal with OpenMPI on QLogic/Intel/PSM fabrics. There are two ways to avoid it: 1) disable PSM (obviously not recommended for parallel jobs): export

Re: [easybuild] Automatic detection of hidden modules

Also you can put that list in a configuration file, e.g. I have a config.cfg file with hide-deps = icc,ifort and use EASYBUILD_CONFIGFILES to point to that file. That is good for the environment. On 18 October 2016 at 13:26, Alan O'Cais wrote: > hidden = True > is now

Re: [easybuild] Has anyone created EB files for the Intel 2017 compilers?

Hi, I wonder if the QM-espresso builds are ok with 2017 update 1 (released yesterday, Nov 1)? Bart On 2 November 2016 at 06:27, Åke Sandgren wrote: > Yes, i have a working set of files. > And i think there was a PR for them too. > > Although 2017 suite needs a

[easybuild] Why is MKL linked statically and other BLAS not?

Hi, in framework intelmkl.py it says: BLAS_LIB_STATIC = True (hardcoded, not an option) I did some "git blame"ing to find out the reason and found this commit introduced it: commit 65c5e3ea01f8a175244115629090ace41fa30fdc Author: stdweird Date: Tue Sep 18 10:22:07

Re: [easybuild] Mesa with SWR on non-AVX processor

That phoronix article is quite old. In the end swrast will use llvmpipe if you also use --enable-gallium-llvm ( http://www.paraview.org/Wiki/ParaView/ParaView_And_Mesa_3D) Software rendering without either llvmpipe or swr is very slow indeed, but as you can see on the above page software

Re: [easybuild] Recipe for Torch ?

Hi, I have some notes here about Torch installation: Installation by my ex-colleague Dan Mazur (28 Aug 2015) Source location: /software/CentOS-6/applications/torch Install location : /software/CentOS-6/applications/torch/install Modules loaded for built: git/2.4.1 gnuplot/4.6.4

Re: [easybuild] Boost-1.64.0-intel-2017a.eb fails in eb 3.2.1

Hi Joachim, this is Boost bug https://svn.boost.org/trac/boost/ticket/12723 We installed Boost 1.60 as the default module to avoid any issue with this. On 18 May 2017 at 12:12, Joachim Hein wrote: > Hi, > > I have an issue in EB 3.2.1 when tyring to build

Re: [easybuild] Boost-1.64.0-intel-2017a.eb fails in eb 3.2.1

Also, https://svn.boost.org/trac/boost/ticket/12516 is another reason for not doing anything newer than 1.60 for Boost-mpi. Bart On 18 May 2017 at 12:46, Bart Oldeman <bart.olde...@mcgill.ca> wrote: > Hi Joachim, > > this is Boost bug https://svn.boost.org/trac/boost/tick

Re: [easybuild] Installing python bundles in a virtual environment

Hi, in the end I (as I work with Maxime) found a solution that seems to work but needs some more testing: * instead of setting PYTHONPATH, eb-generated modules set EBPYTHONPATH * then in the default site-packages directory (or in a location pointed to by a single-entry PYTHONPATH) we have a file

Re: [easybuild] custom glibc

Hello Juri, yes in Compute Canada we use Nix to install glibc, and install easybuild on top of a particular (read-only for regular users!) Nix profile. A regular user will use very little outside that stack (exceptions are e.g. the scheduler and suid-needing applications such as Singularity and

Re: [easybuild] Supporting many python versions with the same module

Hi, just to elaborate because a few things are missing from Maxime's email. 1. We still need PYTHONPATH to find sitecustomize.py itself. We put that in the main Python module but alternatively it could be put in $EBROOTPYTHON/lib/pythonx.y/site-packages in which case you don't need to set

Re: [easybuild] Python on GCCcore

Hi Markus, our solution is to make python-bare a little less bare by installing a bare minimum of packages, and also allow users to pick and choose pip-installed packages in virtualenvs where we provide pre-compiled wheels in common cases (e.g. numpy, ...) so this installs just setuptools, pip,

Re: [easybuild] initial proposal for 2018b update of common toolchains

On Sat, 26 May 2018 at 08:21, Jure Pečar wrote: > Btw, what is the status of CUDA with recent gcc versions? > Brand new CUDA 9.2 supports GCC 7. CUDA 9.1 and CUDA 9.0 only GCC 6. https://docs.nvidia.com/cuda/cuda-toolkit-release-notes/index.html From:

Re: Re: [easybuild] Minimal vs full toolchain, Qt, CUDA etc.

Hi Maik, there is already an in-between solution for Python that sits at the compiler level using the "golf" toolchain which is a subset of "foss" (everything except MPI, so GCC + OpenBLAS + LAPACK + FFTW, but no ScaLAPACK and no FFTW with MPI).

Re: [easybuild] EB 3.6.0: Build of iomkl-2018.02.eb fails on Intel Omni-Path systems

Hello Ole, the issue is here: ld: cannot find -lpciaccess ld: cannot find -lxml2 to solve this you can either install OS packages (.rpm, .deb, etc.) for libpciaccess and libxml2, and perhaps add them as OS dependencies. Alternatively you can put in dependencies for ('libxml2', '2.9.8') and

Re: [easybuild] EB 3.6.0: Build of iomkl-2018.02.eb fails on Intel Omni-Path systems

Hello Ole, you'd need the "-devel" packages too, e.g. libpciaccess-devel-0.13.4-3.1.el7_4.x86_64 you can install them as modules using EB, perhaps as hidden modules, it's your choice. Regards, Bart -- Dr. Bart E. Oldeman | bart.olde...@mcgill.ca | bart.olde...@calculquebec.ca Scientific

Re: [easybuild] iomkl-2018b toolchain availability?

* - Intel® MPI Library - Intel® Cluster Checker the first 4 you will need to buy licenses for (for typical HPC use that is, it is free for certain other use cases), but the last 8 not any more. Bart On Mon, 16 Jul 2018 at 11:59, Ole Holm Nielsen wrote: > Hi Bart, > On 07/16/2018 05:47 PM

Re: [easybuild] iomkl-2018b toolchain availability?

Hello Ole, as far as I can see Intel MPI is free beer now (unless you want priority support of course), since 2018 update 2. That doesn't mean iomkl is useless of course! >From what I can see 5805 was worked around by disabling ucx and the upstream bug was fixed in OpenMPI 3.1.1: commit

Re: [easybuild] Loading toolchain module and CC/CXX env vars

Note that there is also a complimentary approach using --dump-env-script, e.g.: $ eb --dump-env-script HPL-2.2-foss-2018b.eb == temporary log file in case of crash /localscratch/eb-GygDhc/easybuild-HYqbH4.log Script to set up build environment for HPL-2.2-foss-2018b.eb dumped to

Re: [easybuild] PyPi legacy is gone.... and ideas on source_urls for Python extensions

Hi Jack, I think what Kenneth was trying to say is that you need to use: 'source_urls': ['https://pypi.python.org/packages/source/i/ipython/'], instead since PYPI_LOWER_SOURCE does not work for extensions. -- Dr. Bart E. Oldeman | bart.olde...@mcgill.ca | bart.olde...@calculquebec.ca Scientific

Re: [easybuild] Re: iomkl-2018.02 toolchain. What OpenMPI version?

Hi Kenneth, On 28 March 2018 at 03:55, Kenneth Hoste wrote: > > I *think* that the main difference between OpenMPI 2.x and 3.x is support > for MPI-3, so the major version bump doesn't need to be a signal of > potential trouble... > OpenMPI has supported MPI 3.0 fully

Re: [easybuild] Re: iomkl-2018.02 toolchain. What OpenMPI version?

Hi Pablo, OpenMPI 3.0.1 and 3.1.0 are supposed to be out real soon now (TM). Supposed to be March 19 but I guess some bugs propped up: https://www.mail-archive.com/devel@lists.open-mpi.org//msg20529.html https://www.mail-archive.com/devel@lists.open-mpi.org//msg20530.html maybe 3.0.1rc4 would be

Re: [easybuild] Cuda supporting toolchains

You can mitigate this issue by installing as much as possible at the compiler level -- we do that at Compute Canada. I have some pending work on the framework that could make that possible for Python too. The major incompatibility between goolfc and goolf=foss is in the MPI libraries, one with and

[easybuild] Re: next EasyBuild conf call: Wed 3 Oct 2018, 17:00 CEST

at 11:18, Bart Oldeman wrote: > Dear EasyBuilders, > > The next, 111th, EasyBuild conf call is planned for tomorrow, Wed 3 Oct > 2018, > at 17:00 CEST, see > https://plus.google.com/events/u/1/cdh4nm4jf5obt7bfhtpekk2vgqc > <https://plus.google.com/events/cdh4nm4jf5obt7bfht

[easybuild] next EasyBuild conf call: Wed 3 Oct 2018, 17:00 CEST

Dear EasyBuilders, The next, 111th, EasyBuild conf call is planned for tomorrow, Wed 3 Oct 2018, at 17:00 CEST, see https://plus.google.com/events/u/1/cdh4nm4jf5obt7bfhtpekk2vgqc . Current agenda: * update on upcoming EasyBuild v3.7.1

Re: [easybuild] Future of Python at GCCcore level?

Hi Mikael, you could actually go one step further and also compile OpenBLAS/LAPACK or install MKL at GCCcore level. FZ Julich does something like that now: https://apps.fz-juelich.de/jsc/llview/jureca_modules/Core/gcccoremkl/7.3.0-2019.0.117.xhtml (they use a seperate SciPy-Stack module at the

Re: [easybuild] 2019a update of common toolchains

Hi Damian, I have no idea if it helps but a new version of libfabric has just been released: https://github.com/ofiwg/libfabric/releases/tag/v1.7.0 whereas according to https://software.intel.com/en-us/articles/intel-mpi-library-release-notes-linux Intel MPI 2019.1 ships with a customized 1.7.0

Re: [easybuild] Openblas(foss) matrix issue

As far as I can see recent versions of OpenBLAS up to 0.3.5 have issues with a) AVX512 kernels, specifically DGEMM has issues. b) Newer compilers such as GCC 8 and in particular GCC 9 which have uncovered bugs in the inline assembler. I don't think that means the code is doomed but it does

[easybuild] Re: next EasyBuild conf call: Wed Oct 30th 2019, 5pm CEST

Notes on today's conf call are available at https://github.com/easybuilders/easybuild/wiki/Conference-call-notes -20191030. Next conf call is planned for Wed Nov 13th 2019, 5pm CET; join via https://tiny.cc/eb_conf_call . regards, Bart On Tue, 29 Oct 2019 at 18:00, Bart Oldeman wrote

[easybuild] next EasyBuild conf call: Wed Oct 16nd 2019, 5pm CEST

Dear EasyBuilders, The next EasyBuild conf call is planned for Wed Oct 16nd 2019 (today or tomorrow, depending on your time zone), at 5pm CEST. You can join the conf call via https://tiny.cc/eb_conf_call . Current agenda: * update on next release (EasyBuild v4.0.1) * update on recently merged

[easybuild] next EasyBuild conf call: Wed Oct 30th 2019, 5pm CEST

Dear EasyBuilders, The next EasyBuild conf call is planned for Wed Oct 30th 2019 (today or tomorrow, depending on your time zone), at 5pm CEST. Note that this is e.g. 12 noon EDT in North America, since daylight savings ends a week later there. You can join the conf call via

Re: [easybuild] difficulties installing FFTW-3.3.8-gompi-2018b.eb

Hi, this smells like a framework bug in the parsing of depends_on in Tcl module files to figure out the dependencies. I don't have much time to look at it today, but best to open an issue here: https://github.com/easybuilders/easybuild-framework/issues Bart

[easybuild] Re: next EasyBuild conf call: Wed Oct 16nd 2019, 5pm CEST

Notes on today's conf call are available at https://github.com/easybuilders/easybuild/wiki/Conference-call-notes-20191016 . Next conf call is planned for Wed Oct 30th 2019, 5pm CEST; join via https://tiny.cc/eb_conf_call . regards, Bart On Tue, 15 Oct 2019 at 21:23, Bart Oldeman wrote

Re: [easybuild] Replacing Intel compiler option -ftz with -no-ftz

Hello Loris, unfortunately it's not so straightforward to easily override a few flags to my knowledge (maybe somebody else knows?) but you can try toolchainopts = {'usempi': True, 'precise': True} which sets -fp-model precise, which may be what you want anyway if you care about denormals. you

[easybuild] Re: next EasyBuild conf call: Wed Mar 4th 2020 at 17:00 CET

Notes for today's conf call are available at https://github.com/easybuilders/easybuild/wiki/Conference-call-notes-20200304 . Next conf call is planned for Wed March 18th 2020, at 10am CET. regards, Bart On Tue, 3 Mar 2020 at 21:15, Bart Oldeman wrote: > > Dear EasyBuilders, > &

[easybuild] next EasyBuild conf call: Wed Mar 4th 2020 at 17:00 CET

Dear EasyBuilders, The next EasyBuild conf call is planned for Wed Feb 19th 2020, at 17:00 CET, alternating from the 10:00 time from the last conf call on Feb 19th. You can join the conf call via https://tiny.cc/eb_conf_call . Current agenda: * overview of recent changes, outlook to next

Re: [easybuild] A problem in the postprocessing stage of GCC on Debian Buster

Hi Samet, On 20/11/2020 16:45, Samet Demir wrote: > Exactly one 'include-fixed' directory expected, found 2: > ['/opt/easybuild/software/GCCcore/10.2.0/lib/gcc/i686-pc-linux-gnu/10.2.0/include-fixed', > >

Re: [easybuild] Re: UCX ibv_create_cq and UCP worker errors on nodes with EL8 OS and Omni-Path fabric

ursue this approach any > further. > > Do you have any other ideas about OMPI_* variables that we could try? > Since I'm not an MPI expert, complete commands and variables would be > appreciated :-) > > I would like to remind you that we're running AlmaLinux 8.5 with new > versi

[easybuild] Re: next EasyBuild conf call: Wed Nov 10th 2021 (*today*) at 16:00 UTC (17:00 CEST)

/Conference-call-notes-2020 On Tue, 9 Nov 2021 at 14:36, Bart Oldeman wrote: > Dear EasyBuilders, > > The next EasyBuild conf call is planned for Wed Nov 10th 2021 at 16:00 > UTC (17:00 CEST). > > You can join the conf call via https://tiny.cc/eb_conf_call . > > Curr

[easybuild] next EasyBuild conf call: Wed Nov 10th 2021 (*today*) at 16:00 UTC (17:00 CEST)

Dear EasyBuilders, The next EasyBuild conf call is planned for Wed Nov 10th 2021 at 16:00 UTC (17:00 CEST). You can join the conf call via https://tiny.cc/eb_conf_call . Current agenda: * update of recent developments * Q Suggestions for additional topics are always welcome! More

Re: [easybuild] Re: UCX ibv_create_cq and UCP worker errors on nodes with EL8 OS and Omni-Path fabric

Hi Ole, we found that UCX isn't very useful not performant on OmniPath, so if your compiled isn't used on both InfiniBand and OmniPath you can compile OpenMPI using "eb --filter-deps=UCX ..." Open MPI works well there either using libpsm2 directly (using the "cm" pml and "psm2" mtl), or via

[easybuild] Re: next EasyBuild conf call: Wed 25 Oct 2023 at 15:00 UTC (17:00 CEST)

, 24 Oct 2023 at 23:21, Bart Oldeman wrote: > > Dear EasyBuilders, > > The next EasyBuild conf call is planned for Wed 25 Oct 2023 at > 15:00 UTC (17:00 CEST). > > You can join the conf call via https://tiny.cc/eb_conf_call . > > Current agenda: > > * overview

[easybuild] next EasyBuild conf call: Wed 25 Oct 2023 at 15:00 UTC (17:00 CEST)

Dear EasyBuilders, The next EasyBuild conf call is planned for Wed 25 Oct 2023 at 15:00 UTC (17:00 CEST). You can join the conf call via https://tiny.cc/eb_conf_call . Current agenda: * overview of recent developments + outlook to next EasyBuild release; * quick status update on development

[easybuild] next EasyBuild conf call: Wed 28 Sep 2022 (*today*) at 15:00 UTC (17:00 CEST)

Dear EasyBuilders, The next EasyBuild conf call is planned for Wed 28 Sep 2022 (*today*) at 15:00 UTC (17:00 CEST). You can join the conf call via https://tiny.cc/eb_conf_call . Current agenda: * overview of recent developments + outlook to next EasyBuild release * NVHPC toolchains *

[easybuild] Re: next EasyBuild conf call: Wed 28 Sep 2022 (*today*) at 15:00 UTC (17:00 CEST)

. regards, Bart On Wed, 28 Sept 2022 at 07:18, Bart Oldeman wrote: > > Dear EasyBuilders, > > The next EasyBuild conf call is planned for Wed 28 Sep 2022 (*today*) at > 15:00 UTC (17:00 CEST). > > You can join the conf call via https://tiny.cc/eb_conf_call . > > Cur

Re: [easybuild] OpenBLAS-0.3.20-GCC-11.3.0.eb: Too many LAPACK tests failed

Hi Loris, there are some LAPACK tests that are very sensitive to floating point rounding, in particular those involving complex numbers, vectorization, and fused-multiply-addition (FMA). Most of these have been addressed by GCC patches, but it depends a bit on the compilation flags you are using

Re: [easybuild] Failure in UCX-1.14.1-GCCcore-12.3.0.eb when installing foss-2023a.eb

Hi, On 7.9 it's defined here: $ rpm -qf /usr/include/infiniband/verbs.h rdma-core-devel-22.4-6.el7_9.x86_64 $ grep IBV_ACCESS_ON_DEMAND /usr/include/infiniband/verbs.h IBV_ACCESS_ON_DEMAND= (1<<6), if you have MOFED it would be somewhere else, not sure. Le jeu. 14 mars