September 2021 - petsc-dev - mailman3.cels.anl.gov

Re: [petsc-dev] Cannot locate file: share/petsc/datafiles/matrices/small
by Satish Balay 12 Sep '21

12 Sep '21

> you either need to copy the files from the datafiles repository Actually clone https://gitlab.com/petsc/datafiles and use its location with DATAFILESPATH option But yeah - its not really needed. If DATAFILESPATH is missing - the corresponding tests won't be run. Satish On Sun, 12 Sep 2021, Pierre Jolivet wrote: > Hello, > Did you copy the files from https://gitlab.com/petsc/datafiles <https://gitlab.com/petsc/datafiles> before the make check? > cfd.1.10 is not part of https://gitlab.com/petsc/petsc <https://gitlab.com/petsc/petsc>, so you either need to copy the files from the datafiles repository, or remove the --DATAFILESPATH=/home/sagitter/rpmbuild/BUILD/petsc-3.15.4/petsc-3.15.4/share/petsc/datafiles configure option. > > Thanks, > Pierre > > > On 12 Sep 2021, at 6:02 PM, Antonio T. sagitter <sagitter(a)fedoraproject.org> wrote: > > > > Hi all. > > > > This error is repeated when the tests are compiled in PETSc-3.15.4 compilation: > > https://paste.in/MhdNjf > > > > This happens in Fedora 34 x86_64 with gcc-11.2.1 > > > > make.log and configure.log are attached > > -- > > --- > > Antonio Trande > > Fedora Project > > mailto: sagitter(a)fedoraproject.org > > GPG key: 0x29FBC85D7A51CC2F > > GPG key server: https://keyserver1.pgp.com/ > > <make_configure.zip><OpenPGP_0x29FBC85D7A51CC2F.asc> > >

1 0

Re: [petsc-dev] Cannot locate file: share/petsc/datafiles/matrices/small
by Pierre Jolivet 12 Sep '21

12 Sep '21

Hello, Did you copy the files from https://gitlab.com/petsc/datafiles <https://gitlab.com/petsc/datafiles> before the make check? cfd.1.10 is not part of https://gitlab.com/petsc/petsc <https://gitlab.com/petsc/petsc>, so you either need to copy the files from the datafiles repository, or remove the --DATAFILESPATH=/home/sagitter/rpmbuild/BUILD/petsc-3.15.4/petsc-3.15.4/share/petsc/datafiles configure option. Thanks, Pierre > On 12 Sep 2021, at 6:02 PM, Antonio T. sagitter <sagitter(a)fedoraproject.org> wrote: > > Hi all. > > This error is repeated when the tests are compiled in PETSc-3.15.4 compilation: > https://paste.in/MhdNjf > > This happens in Fedora 34 x86_64 with gcc-11.2.1 > > make.log and configure.log are attached > -- > --- > Antonio Trande > Fedora Project > mailto: sagitter(a)fedoraproject.org > GPG key: 0x29FBC85D7A51CC2F > GPG key server: https://keyserver1.pgp.com/ > <make_configure.zip><OpenPGP_0x29FBC85D7A51CC2F.asc>

1 0

Cannot locate file: share/petsc/datafiles/matrices/small
by Antonio T. sagitter 12 Sep '21

12 Sep '21

Hi all. This error is repeated when the tests are compiled in PETSc-3.15.4 compilation: https://paste.in/MhdNjf This happens in Fedora 34 x86_64 with gcc-11.2.1 make.log and configure.log are attached -- --- Antonio Trande Fedora Project mailto: sagitter(a)fedoraproject.org GPG key: 0x29FBC85D7A51CC2F GPG key server: https://keyserver1.pgp.com/

1 0

Re: [petsc-dev] NVIDIA licensing
by Barry Smith 07 Sep '21

07 Sep '21

There is no way to do do it consistently for all contributors, it would be a big mess plus it would miss all the history. If NVIDIA requires it to contribute then I see no harm in having those two lines in certain source files. It is not tracking provenance in any way, it is just a way to allow NVIDIA to check its check marks. Barry > On Sep 6, 2021, at 11:23 PM, Satish Balay <balay(a)mcs.anl.gov> wrote: > >>> 2) NVIDIA's lawyers have some thoughts that we will need to address > > When I saw this is, my thought was: ANL lawyers would be involved here [as generally employees couldn't respond to legal stuff] > >> git grep Copyright locates some files that are copyrighted by other groups (but fit within the BSD license) so I think it is fine to include the NVIDIA copyright information in appropriate files. > > My impression here was we are bundling some external stuff [yaml, abi-compliance-checker, khash, valgrid] - so in some sense they are still external stuff - not petsc native stuff - where we are to retain their copyright - as per terms of use. > > So would 'SPDX' be similar module? It wasn't clear to me if - say any current petsc sources say src/ksp/pc/interface/pcregis.c gets updated as part of this work - would it also need this copyright update? Perhaps thats ok? > > And if this a policy (for current petsc sources) - we should have to do this for all contributions to be consistent? > > Satish > > On Mon, 6 Sep 2021, Barry Smith wrote: > >> >> git grep Copyright locates some files that are copyrighted by other groups (but fit within the BSD license) so I think it is fine to include the NVIDIA copyright information in appropriate files. >> >> We should also add some text to LICENSE indicating certain files have additional copyrights indicated at the top of the individual file. >> >> Barry >> >> >>> On Sep 3, 2021, at 7:48 AM, Mark Adams <mfadams(a)lbl.gov> wrote: >>> >>> All, >>> >>> Matthew (cc'ed) and I are going to start working on an AMGx interface in PETSc, PCAMGX, and I suggested that he may want to think about moving the core AMG PC code into PETSC as a built-in PC. >>> >>> 1) We would have to decide of we want the maintenance burden of moving the whole thing in, but we will probably start with a 3rd party library unless this move starts looking like a really good idea. We have not started to really dig into this. >>> >>> 2) NVIDIA's lawyers have some thoughts that we will need to address even with a 3rd party library (I know Hypre has this license notification and LLNL seems happy with what we did). >>> >>> First, Mathew says that this process "can open the door to us helping with PETSc more generally." >>> >>> I will just quote the rest of Mathews' message: >>> >>> Thanks, >>> Mark >>> >>> I can see you don't add license notifications to the headers of the source files, but this tends to be the approach we take for copyright. I'm also not sure there is a very good alternative? >>> >>> Instead of the monolithic copying of the full BSD-2 license with the NVIDIA copyright (which I appreciate might not be desirable), would it be a suitable compromise if we used the SPDX identifier, which looks like: >>> >>> // Copyright (c) 1991-2021, NVIDIA. All rights reserved. >>> // Copyright (c) 1991-2021, UChicago Argonne, LLC and the PETSc Development Team. All rights reserved. >>> // SPDX-License-Identifier: BSD-2-clause >>> >>> This would be for the main source files that we make any significant contributions to or author. I would be interested to hear your thoughts / ideas on this. >>> >> >> >

1 0

Re: [petsc-dev] NVIDIA licensing
by Satish Balay 06 Sep '21

06 Sep '21

> > 2) NVIDIA's lawyers have some thoughts that we will need to address When I saw this is, my thought was: ANL lawyers would be involved here [as generally employees couldn't respond to legal stuff] > git grep Copyright locates some files that are copyrighted by other groups (but fit within the BSD license) so I think it is fine to include the NVIDIA copyright information in appropriate files. My impression here was we are bundling some external stuff [yaml, abi-compliance-checker, khash, valgrid] - so in some sense they are still external stuff - not petsc native stuff - where we are to retain their copyright - as per terms of use. So would 'SPDX' be similar module? It wasn't clear to me if - say any current petsc sources say src/ksp/pc/interface/pcregis.c gets updated as part of this work - would it also need this copyright update? Perhaps thats ok? And if this a policy (for current petsc sources) - we should have to do this for all contributions to be consistent? Satish On Mon, 6 Sep 2021, Barry Smith wrote: > > git grep Copyright locates some files that are copyrighted by other groups (but fit within the BSD license) so I think it is fine to include the NVIDIA copyright information in appropriate files. > > We should also add some text to LICENSE indicating certain files have additional copyrights indicated at the top of the individual file. > > Barry > > > > On Sep 3, 2021, at 7:48 AM, Mark Adams <mfadams(a)lbl.gov> wrote: > > > > All, > > > > Matthew (cc'ed) and I are going to start working on an AMGx interface in PETSc, PCAMGX, and I suggested that he may want to think about moving the core AMG PC code into PETSC as a built-in PC. > > > > 1) We would have to decide of we want the maintenance burden of moving the whole thing in, but we will probably start with a 3rd party library unless this move starts looking like a really good idea. We have not started to really dig into this. > > > > 2) NVIDIA's lawyers have some thoughts that we will need to address even with a 3rd party library (I know Hypre has this license notification and LLNL seems happy with what we did). > > > > First, Mathew says that this process "can open the door to us helping with PETSc more generally." > > > > I will just quote the rest of Mathews' message: > > > > Thanks, > > Mark > > > > I can see you don't add license notifications to the headers of the source files, but this tends to be the approach we take for copyright. I'm also not sure there is a very good alternative? > > > > Instead of the monolithic copying of the full BSD-2 license with the NVIDIA copyright (which I appreciate might not be desirable), would it be a suitable compromise if we used the SPDX identifier, which looks like: > > > > // Copyright (c) 1991-2021, NVIDIA. All rights reserved. > > // Copyright (c) 1991-2021, UChicago Argonne, LLC and the PETSc Development Team. All rights reserved. > > // SPDX-License-Identifier: BSD-2-clause > > > > This would be for the main source files that we make any significant contributions to or author. I would be interested to hear your thoughts / ideas on this. > > > >

1 0

Re: [petsc-dev] NVIDIA licensing
by Jed Brown 06 Sep '21

06 Sep '21

On Mon, Sep 6, 2021, at 7:16 PM, Barry Smith wrote: > > * We should also add some text to LICENSE indicating certain files have additional copyrights indicated at the top of the individual file.* > ** Copyright is held by hundreds of individuals and organizations. I'm not a fan of having it in each file as though that were an effective way to track provenance. I'm not saying we refuse to put it there for Nvidia, but we should make it clear that this is not the expected way to track provenance.

1 0

Re: [petsc-dev] NVIDIA licensing
by Barry Smith 06 Sep '21

06 Sep '21

git grep Copyright locates some files that are copyrighted by other groups (but fit within the BSD license) so I think it is fine to include the NVIDIA copyright information in appropriate files. We should also add some text to LICENSE indicating certain files have additional copyrights indicated at the top of the individual file. Barry > On Sep 3, 2021, at 7:48 AM, Mark Adams <mfadams(a)lbl.gov> wrote: > > All, > > Matthew (cc'ed) and I are going to start working on an AMGx interface in PETSc, PCAMGX, and I suggested that he may want to think about moving the core AMG PC code into PETSC as a built-in PC. > > 1) We would have to decide of we want the maintenance burden of moving the whole thing in, but we will probably start with a 3rd party library unless this move starts looking like a really good idea. We have not started to really dig into this. > > 2) NVIDIA's lawyers have some thoughts that we will need to address even with a 3rd party library (I know Hypre has this license notification and LLNL seems happy with what we did). > > First, Mathew says that this process "can open the door to us helping with PETSc more generally." > > I will just quote the rest of Mathews' message: > > Thanks, > Mark > > I can see you don't add license notifications to the headers of the source files, but this tends to be the approach we take for copyright. I'm also not sure there is a very good alternative? > > Instead of the monolithic copying of the full BSD-2 license with the NVIDIA copyright (which I appreciate might not be desirable), would it be a suitable compromise if we used the SPDX identifier, which looks like: > > // Copyright (c) 1991-2021, NVIDIA. All rights reserved. > // Copyright (c) 1991-2021, UChicago Argonne, LLC and the PETSc Development Team. All rights reserved. > // SPDX-License-Identifier: BSD-2-clause > > This would be for the main source files that we make any significant contributions to or author. I would be interested to hear your thoughts / ideas on this. >

1 0

Re: [petsc-dev] Remove PetscDevice initialization from main before 3.16 is released
by Jacob Faibussowitsch 03 Sep '21

03 Sep '21

https://gitlab.com/petsc/petsc/-/merge_requests/4284 <https://gitlab.com/petsc/petsc/-/merge_requests/4284> Best regards, Jacob Faibussowitsch (Jacob Fai - booss - oh - vitch) > On Sep 3, 2021, at 03:34, Stefano Zampini <stefano.zampini(a)gmail.com> wrote: > > Jacob > > Just got a report about the error below > PetscDevice code is still experimental and I believe it should not be active (in any way) in the next release. > It can also conflict with the current way we initialize devices. > Can we comment out these two extra lines in PetscInitialize in main and 3.16? > It seems it is currently the only point of entry to this new stuff. > > [129]PETSC ERROR: --------------------- Error Message -------------------------------------------------------------- > [129]PETSC ERROR: GPU error > [129]PETSC ERROR: cuda error 100 (cudaErrorNoDevice) : no CUDA-capable device is detected > [129]PETSC ERROR: See https://petsc.org/release/faq/ <https://petsc.org/release/faq/> for trouble shooting. > [129]PETSC ERROR: Petsc Development GIT revision: v3.15.3-732-ge2a280ad64 GIT Date: 2021-08-30 16:02:52 -0400 > [129]PETSC ERROR: ../viscousDriver3d.Linux.64.mpiCC.mpifort.OPT.MPI.PETSC.ex on a named a13n04 by Unknown Thu Sep 2 18:38:04 2021 > [129]PETSC ERROR: Configure options --COPTFLAGS="-O0 -mcpu=power9" --CUDAOPTFLAGS="-O0 -Xcompiler -mcpu=power9" --CXXOPTFLAGS="-O0 -mcpu=power9" --FOPTFLAGS="-O0 -mcpu=power9" --download-hypre-commit=origin/hypre_petsc --download-hypre=1 --download-metis --downl\ > oad-parmetis --with-cuda-gencodearch=70 --with-cuda=1 --with-cxx-dialect=C++14 --with-debugging=1 --with-fortran-bindings=0 --prefix=/gpfs/alpine/world-shared/geo127/petsc/arch-dbg-gcc9.1.0-cuda11.0.3 PETSC_ARCH=arch-summit-hypre-cuda-dbg2 PETSC_DIR=/gpfs/alpine\ > /csc314/scratch/adams/petsc2 > [129]PETSC ERROR: #1 __initialize() at /gpfs/alpine/csc314/scratch/adams/petsc2/src/sys/objects/device/impls/cupm/cupmdevice.cxx:40 > [129]PETSC ERROR: #2 getDevice() at /gpfs/alpine/csc314/scratch/adams/petsc2/src/sys/objects/device/impls/cupm/cupmdevice.cxx:59 > [129]PETSC ERROR: #3 PetscDeviceCreate() at /gpfs/alpine/csc314/scratch/adams/petsc2/src/sys/objects/device/interface/device.cxx:49 > [129]PETSC ERROR: #4 InitializeDeviceHelper_Private() at /gpfs/alpine/csc314/scratch/adams/petsc2/src/sys/objects/device/interface/device.cxx:143 > [129]PETSC ERROR: #5 PetscDeviceInitializeDefaultDevices_Internal() at /gpfs/alpine/csc314/scratch/adams/petsc2/src/sys/objects/device/interface/device.cxx:179 > [129]PETSC ERROR: #6 PetscInitialize() at /gpfs/alpine/csc314/scratch/adams/petsc2/src/sys/objects/pinit.c:1178 > [129]PETSC ERROR: #7 main() at viscousDriver.cpp:241

1 0

NVIDIA licensing
by Mark Adams 03 Sep '21

03 Sep '21

All, Matthew (cc'ed) and I are going to start working on an AMGx interface in PETSc, PCAMGX, and I suggested that he may want to think about moving the core AMG PC code into PETSC as a built-in PC. 1) We would have to decide of we want the maintenance burden of moving the whole thing in, but we will probably start with a 3rd party library unless this move starts looking like a really good idea. We have not started to really dig into this. 2) NVIDIA's lawyers have some thoughts that we will need to address even with a 3rd party library (I know Hypre has this license notification and LLNL seems happy with what we did). First, Mathew says that this process "can open the door to us helping with PETSc more generally." I will just quote the rest of Mathews' message: Thanks, Mark *I can see you don't add license notifications to the headers of the source files, but this tends to be the approach we take for copyright. I'm also not sure there is a very good alternative?* *Instead of the monolithic copying of the full BSD-2 license with the NVIDIA copyright (which I appreciate might not be desirable), would it be a suitable compromise if we used the SPDX identifier, which looks like:* *// Copyright (c) 1991-2021, NVIDIA. All rights reserved.* *// Copyright (c) 1991-2021, UChicago Argonne, LLC and the PETSc Development Team. All rights reserved.* *// SPDX-License-Identifier: BSD-2-clause* *This would be for the main source files that we make any significant contributions to or author. I would be interested to hear your thoughts / ideas on this.*

1 0

Remove PetscDevice initialization from main before 3.16 is released
by Stefano Zampini 03 Sep '21

03 Sep '21

Jacob Just got a report about the error below PetscDevice code is still experimental and I believe it should not be active (in any way) in the next release. It can also conflict with the current way we initialize devices. Can we comment out these two extra lines in PetscInitialize in main and 3.16? It seems it is currently the only point of entry to this new stuff. [129]PETSC ERROR: --------------------- Error Message -------------------------------------------------------------- [129]PETSC ERROR: GPU error [129]PETSC ERROR: cuda error 100 (cudaErrorNoDevice) : no CUDA-capable device is detected [129]PETSC ERROR: See https://petsc.org/release/faq/ <https://petsc.org/release/faq/> for trouble shooting. [129]PETSC ERROR: Petsc Development GIT revision: v3.15.3-732-ge2a280ad64 GIT Date: 2021-08-30 16:02:52 -0400 [129]PETSC ERROR: ../viscousDriver3d.Linux.64.mpiCC.mpifort.OPT.MPI.PETSC.ex on a named a13n04 by Unknown Thu Sep 2 18:38:04 2021 [129]PETSC ERROR: Configure options --COPTFLAGS="-O0 -mcpu=power9" --CUDAOPTFLAGS="-O0 -Xcompiler -mcpu=power9" --CXXOPTFLAGS="-O0 -mcpu=power9" --FOPTFLAGS="-O0 -mcpu=power9" --download-hypre-commit=origin/hypre_petsc --download-hypre=1 --download-metis --downl\ oad-parmetis --with-cuda-gencodearch=70 --with-cuda=1 --with-cxx-dialect=C++14 --with-debugging=1 --with-fortran-bindings=0 --prefix=/gpfs/alpine/world-shared/geo127/petsc/arch-dbg-gcc9.1.0-cuda11.0.3 PETSC_ARCH=arch-summit-hypre-cuda-dbg2 PETSC_DIR=/gpfs/alpine\ /csc314/scratch/adams/petsc2 [129]PETSC ERROR: #1 __initialize() at /gpfs/alpine/csc314/scratch/adams/petsc2/src/sys/objects/device/impls/cupm/cupmdevice.cxx:40 [129]PETSC ERROR: #2 getDevice() at /gpfs/alpine/csc314/scratch/adams/petsc2/src/sys/objects/device/impls/cupm/cupmdevice.cxx:59 [129]PETSC ERROR: #3 PetscDeviceCreate() at /gpfs/alpine/csc314/scratch/adams/petsc2/src/sys/objects/device/interface/device.cxx:49 [129]PETSC ERROR: #4 InitializeDeviceHelper_Private() at /gpfs/alpine/csc314/scratch/adams/petsc2/src/sys/objects/device/interface/device.cxx:143 [129]PETSC ERROR: #5 PetscDeviceInitializeDefaultDevices_Internal() at /gpfs/alpine/csc314/scratch/adams/petsc2/src/sys/objects/device/interface/device.cxx:179 [129]PETSC ERROR: #6 PetscInitialize() at /gpfs/alpine/csc314/scratch/adams/petsc2/src/sys/objects/pinit.c:1178 [129]PETSC ERROR: #7 main() at viscousDriver.cpp:241

1 0