{numlib}[foss/2021a] PETSc v3.17.0 w/ Python 3.9.5

[sassy-crick]

(created using eb --new-pr)

Requires:

• PR {math}[gompi/2021a] Mmg v5.6.0 #15515
• PR {math}[GCCcore/10.3.0] muParser v2.3.3 #15514

[jfgrimm]

this could do with cleaning up, probably something along the lines of:

Suggested change

	configopts += '--LIBS="-lm -lpthread -lrt" --with-mpi4py=0 '
	configopts += '--CFLAGS="-O2 -ftree-vectorize -march=native -fno-math-errno -fopenmp -fPIC" '
	configopts += '--CXXFLAGS="-O2 -ftree-vectorize -march=native -fno-math-errno -fopenmp -fPIC '
	configopts += '-DOMPI_SKIP_MPICXX -DMPICH_SKIP_MPICXX" --with-mpi=1 --with-build-step-np=4 '
	configopts += '--FFLAGS="-O2 -ftree-vectorize -march=native -fno-math-errno -fopenmp -fPIC" '
	configopts += '--with-shared-libraries=1 --with-debugging=0 --with-pic=1 --with-x=0 '
	configopts += '--with-windows-graphics=0 --with-numpy=1 --with-fftw=1 --with-scalapack=1 '
	configopts += '--with-blaslapack-lib="[/home/sassy/apps/software/FlexiBLAS/3.0.4-GCC-10.3.0/lib/'
	configopts += 'libflexiblas.a,libgfortran.a]" '
	configopts += '--LIBS="$LIBS -lrt" --with-mpi4py=0 '
	configopts += '--CFLAGS="$CFLAGS" '
	configopts += '--CXXFLAGS="$CXXFLAGS '
	configopts += '-DOMPI_SKIP_MPICXX -DMPICH_SKIP_MPICXX" --with-mpi=1 --with-build-step-np=4 '
	configopts += '--FFLAGS="$FFLAGS" '
	configopts += '--with-shared-libraries=1 --with-debugging=0 --with-pic=1 --with-x=0 '
	configopts += '--with-windows-graphics=0 --with-numpy=1 --with-fftw=1 --with-scalapack=1 '
	configopts += '--with-blaslapack-lib="$BLAS_LAPACK_LIB_DIR"
	'

[sassy-crick]

Thanks for the suggestion. much appreciated. The "$BLAS_LAPACK_LIB_DIR" is actually not needed, what is needed are the two libs libflexiblas.a,libgfortran.a which I replaced with -lflexiblas -lgfortran to stick to dynamic libs.
Please have a look at the log file to make sure I am not accidentally build it with the wrong flags. From what I can see, it is all working as intended but that does not mean it is correct.

[jfgrimm]

Eigen should only need to be a build dependency (since it's header-only). I imagine googletest isn't a runtime requirement either

[sassy-crick]

True. I forgot about Eigen being only header. I was quite happy that it all worked after 2 days of playing with it to be honest. :-)

[boegel]

@sassy-crick Is there a strict need for a newer version of PETSc, or could you also use the existing PETSc-3.15.1-foss-2021a.eb?

[sassy-crick]

@sassy-crick Is there a strict need for a newer version of PETSc, or could you also use the existing PETSc-3.15.1-foss-2021a.eb?

Well, the reason I put myself into that was the installation of FreeFEM where, when I looked at the 3rd party files, I found that:

download('PETSc','https://www.mcs.anl.gov/petsc/mirror/release-snapshots/petsc-3.17.0.tar.gz',

So I thought I better stick to what they are using. I agree with you that is not an ideal situation to have two different versions of the same programs in one toolchain.
To make things more interesting: there is actually PETSc version 3.17.1 out, which in turn requires SLEPC-3.17.1
Of course, all of that also requires SuperLU next to the other dependencies where I have created PRs recently.

I guess one way out of that is to push it up in the toolchain and see if the 3.17.1 versions are ok for that FreeFEM build. How far are we with foss-2022a ?

[boegel]

I guess one way out of that is to push it up in the toolchain and see if the 3.17.1 versions are ok for that FreeFEM build. How far are we with foss-2022a ?

Since 3.17.1 is a bugfix release, that should be fine for FreeFEM. foss/2022a isn't fully settled yet, but how about going with foss/2021b instead?

[sassy-crick]

Since 3.17.1 is a bugfix release, that should be fine for FreeFEM. foss/2022a isn't fully settled yet, but how about going with foss/2021b instead?

There are two things which we need to consider:

• that PR is only part of a larger sets of PRs which are needed for PR 15537
• also, I have asked the user to test FreeFEM to see if that is all working. I am waiting for the results to see if that is all working numerically as well.

That said, I am happy to push that up to version 3.17.1 do the same for PR 15550 (SLEPc) as they are mutually depending on each other in terms of version, and push the rest of the requirements up to foss2021b as well. I did not want to go for a too new compiler initially, just to make sure I am not opening yet another can of worms in that quite complex installation.
Happy with that?

[boegel]

@boegelbot please test @ generoso

[boegelbot]

@boegel: Request for testing this PR well received on login1

PR test command 'EB_PR=15519 EB_ARGS= /opt/software/slurm/bin/sbatch --job-name test_PR_15519 --ntasks=4 ~/boegelbot/eb_from_pr_upload_generoso.sh' executed!

• exit code: 0
• output:

Submitted batch job 9383

Test results coming soon (I hope)...

Details

- notification for comment with ID 1291756479 processed

Message to humans: this is just bookkeeping information for me,
it is of no use to you (unless you think I have a bug, which I don't).

[boegelbot]

Test report by @boegelbot
FAILED
Build succeeded for 1 out of 2 (1 easyconfigs in total)
cns8 - Linux Rocky Linux 8.5, x86_64, Intel(R) Xeon(R) CPU E5-2667 v3 @ 3.20GHz (haswell), Python 3.6.8
See https://gist.github.com/b9fb7290fb0387271f7a46f26ac2782e for a full test report.

[sassy-crick]

Closed this PR as it is using a rather old toolchain.