January 2016 - firedrake - imperial.ac.uk

Re: [firedrake] How do I cite Firedrake?
by Andrew McRae 11 Jan '16

11 Jan '16

On 7 January 2016 at 09:43, Lawrence Mitchell < lawrence.mitchell(a)imperial.ac.uk> wrote: > > > On 07/01/16 08:51, David Ham wrote: > > Hi Justin, > > > > That's exactly the right paper to cite for Firedrake in general. If > > your work particularly depends on features which have their own > > publications, then you might also cite those. At this stage, anything > > using quadrilateral elements or extrusion should cite McRae et al. > > (2016) > > That's this one: > > @article{McRae2015, > title = {Automated generation and symbolic manipulation of > tensor product finite elements}, > author = {McRae, Andrew TT and Bercea, Gheorge-Teodor and > Mitchell, Lawrence and Ham, David A and Cotter, > Colin J}, > Note that should be Gheorghe, not Gheorge. I take full responsibility for misspelling his name on the original arXiv submission! > archivePrefix ="arXiv", > journal = {Submitted to SIAM Journal on Scientific Computing}, > eprint = {1411.2940}, > year = {2015}, > url = {http://arxiv.org/abs/1411.2940} > } > > > > and anything which has a particular reliance on kernel > > performance should cite Luporini et al. (2015). > > @article{Luporini2015, > author = {Luporini, Fabio and Varbanescu, Ana Lucia and Rathgeber, > Florian and Bercea, Gheorghe-Teodor and Ramanujam, J. and Ham, David > A. and Kelly, Paul H. J.}, > title = {Cross-Loop Optimization of Arithmetic Intensity for Finite > Element Local Assembly}, > journal = {ACM Trans. Archit. Code Optim.}, > issue_date = {January 2015}, > volume = {11}, > number = {4}, > month = jan, > year = {2015}, > issn = {1544-3566}, > pages = {57:1--57:25}, > articleno = {57}, > numpages = {25}, > url = {http://doi.acm.org/10.1145/2687415}, > doi = {10.1145/2687415}, > acmid = {2687415}, > publisher = {ACM}, > address = {New York, NY, USA}, > keywords = {Finite element integration, SIMD vectorization, > compilers, local assembly, optimizations}, > } > > > > If you say anything > > about UFL, you should cite Martin Alnæs' paper on the subject, and > > it's likely that most papers using Firedrake would also want to be > > citing PETSc and/or one of the packages PETSc provides an interface to > > (depending on what you are using). > > > > One of the things on the vapourware todo list is to provide a > > PETSc-style citation option to Firedrake to autogenerate the list of > > citations used in a particular run. > > Running with -citations will do, at least, the PETSc part for now. > > Lawrence > >

3 2

Forced update of firedrake petsc branch
by Lawrence Mitchell 11 Jan '16

11 Jan '16

Dear all, the next time you update firedrake, you will have to manually update the git repository for PETSc. For a while we've been carrying a commit fixing a bug that is now fixed upstream, as a result, the firedrake branch of PETSc has been reset to match upstream PETSc. To update PETSc, you'll need to do the following: 1. Change to the petsc src directory of your firedrake installation: cd path/to/firedrake/src/petsc 2. Update the PETSc source tree. git fetch origin # Only do this next step if you have NO CHANGES TO PETSc # that you want to keep. If you do have local changes, # please email and we can fix up the instructions. git reset --hard origin/firedrake 3. Update firedrake as normal with the firedrake-update script, only remembering to additionally passing "--rebuild" firedrake-update --normal-update-options --rebuild 4. There is no step 4, you should be good to go. Please let us know if any of this doesn't work and/or makes no sense. Cheers, Lawrence

1 0

Re: [firedrake] calling firedrake in idle
by David Ham 11 Jan '16

11 Jan '16

Dear Angwenyi, This is actually an idle/virtualenv issue rather than a firedrake one. Basically idle is running the system python rather than the one in the virtualenv. You can get around this by explicitly using the virtualenv python when running idle. python /usr/local/bin/idle On Mon, 11 Jan 2016 at 13:13 Angwenyi David <kipkoej(a)gmail.com> wrote: > Dear all, > > I activated firedrake in the mac terminal and then opened idle from the > terminal but I get the following error while trying to import firedrake in > idle: > > Traceback (most recent call last): > File "<pyshell#1>", line 1, in <module> > from firedrake import * > ImportError: No module named firedrake > > How can I get around this problem? > > Thanks in advance. > > Angwenyi > _______________________________________________ > firedrake mailing list > firedrake(a)imperial.ac.uk > https://mailman.ic.ac.uk/mailman/listinfo/firedrake >

2 1

calling firedrake in idle
by Angwenyi David 11 Jan '16

11 Jan '16

Dear all, I activated firedrake in the mac terminal and then opened idle from the terminal but I get the following error while trying to import firedrake in idle: Traceback (most recent call last): File "<pyshell#1>", line 1, in <module> from firedrake import * ImportError: No module named firedrake How can I get around this problem? Thanks in advance. Angwenyi

2 1

Solving an ODE in firedrake
by Justin Chang 08 Jan '16

08 Jan '16

Hi all, Say I have some ODE of the following: dc/dt = f where f is some analytical expression. So if I used backward euler temporal discretization: (c - c0)/dt = f where c, c0, and f belong to the same function space (P1 lagrange elements). My question now is, is it a good idea to solve the above using the Galerkin weak form? So if I had something like: a = w*c*dx L = w*(dt*f + c0)*dx Would this give an accurate solution? Or is there a better way of solving this ODE? Thanks, Justin

1 0

Surface integrals on custom mesh
by Anna Kalogirou 08 Jan '16

08 Jan '16

Dear all, I have the mesh defined in the file attached. I can successfully import it and use it with Firedrake, but how can I define a surface integral? For example, I want to somehow define a "ds" which will denote the left boundary of my domain, corresponding to y=0. The nodes that will define this "ds" are conveniently the first 20 nodes in the mesh file attached, with y=0 (the second column). Thanks, Anna.

3 3

New job, same place.
by David Ham 08 Jan '16

08 Jan '16

Dear all, As many of you know, there has been some uncertainty about where I would be continuing my career in the future. I am now very happy to let you know that I have just accepted an appointment as senior lecturer (that's associate professor for the Yanks out there) in the maths department at Imperial College London. I look forward to your continued bug reports and to taking Firedrake forward to ever more and better things ;). Regards, David

3 2

Re: [firedrake] How do I cite Firedrake?
by David Ham 07 Jan '16

07 Jan '16

Hi Justin, That's exactly the right paper to cite for Firedrake in general. If your work particularly depends on features which have their own publications, then you might also cite those. At this stage, anything using quadrilateral elements or extrusion should cite McRae et al. (2016) and anything which has a particular reliance on kernel performance should cite Luporini et al. (2015). If you say anything about UFL, you should cite Martin Alnæs' paper on the subject, and it's likely that most papers using Firedrake would also want to be citing PETSc and/or one of the packages PETSc provides an interface to (depending on what you are using). One of the things on the vapourware todo list is to provide a PETSc-style citation option to Firedrake to autogenerate the list of citations used in a particular run. Regards, David On Thu, 7 Jan 2016 at 06:17 Justin Chang <jychang48(a)gmail.com> wrote: > I have used the Firedrake project to produce results in a paper I am > hoping to submit very soon, but how exactly do I cite you guys? > > Would it be > > Firedrake: automating the finite element method by composing abstractions" Rathgeber, > F.; Ham, D. A; Mitchell, L.; Lange, M.; Luporini, F.; McRae, A. T.; Bercea, > G.; Markall, G. R; and Kelly, P. H. *Submitted to ACM TOMS*, . 2015. > > or something else? > > Thanks, > Justin >

2 1

How do I cite Firedrake?
by Justin Chang 07 Jan '16

07 Jan '16

I have used the Firedrake project to produce results in a paper I am hoping to submit very soon, but how exactly do I cite you guys? Would it be Firedrake: automating the finite element method by composing abstractions" Rathgeber, F.; Ham, D. A; Mitchell, L.; Lange, M.; Luporini, F.; McRae, A. T.; Bercea, G.; Markall, G. R; and Kelly, P. H. *Submitted to ACM TOMS*, . 2015. or something else? Thanks, Justin

1 0

PETSc via petsc4py/PyOP2 really slow
by Justin Chang 04 Jan '16

04 Jan '16

Hi all, If I had a purely firedrake code, everything runs fast on my Macbook. But when I need to make direct calls to the TaoSolver via petsc4py or PETSc calls via PyOP2, the program is extremely slow. This was not the case as of several commits ago (like probably two months). I suspect it has something to do with not adding any optimization flags to the PETSc configure options? The PETSc developers suggest adding more than simply --with-debugging=0. For instance, if using GNU compilers one could use COPTFLAGS='-O3 -march=native' It seems to me that whatever compiler optimization flags were used for Firedrake/PyOP2 function calls are not used for direct usage of PETSc function calls? Thanks, Justin

2 2