September 2020 - firedrake

Continuing informal reading group
by Lawrence Mitchell 06 Oct '20

06 Oct '20

Dear firedrakers, after our (I think) reasonable success running an online reading group over the summer, I propose continuing it as the Autumn kicks in. Over the summer we read through Joachim Schoeberl's thesis on parameter robust multigrid methods (see https://github.com/firedrakeproject/firedrake/wiki/schoeberl-thesis) I propose that we switch to going through papers, rather than complete theses. These sessions are reasonably informal, and I (at least) learned some things this time round. Given that we've just closed out one topic, I'm soliciting both timeslots for interested parties, and potential papers (and volunteers to lead on the reading!). If you're interested in participating, please fill in this doodle selecting times that work for you: https://doodle.com/poll/vt8smu76c9m66hbv (ignore the dates, we'll try and make it a weekly thing) I'll then try and find a time that suits most people, and we can decide on a paper to start with. Thanks, Lawrence

2 4

strange nonlinear solver divergence
by arash pourakaberian 29 Sep '20

29 Sep '20

Dear Firedrake developer team, I could not use streamplot, so updated the firedrake yesterday to fix the problem. It still doesn't work and additionally, I encounter a strange nonlinear solver divergence error for the same problem that it was able to solve before the package was updated. I don't know if the problem is due to petsc configuration or something else. Would you please help me resolve the issue? Two figures are attached to show the problem could be solved without error before the update. The text file attached shows the SNES monitor output. It is worth to mention the problem can still be solved when the gradient of the function at the top and bottom boundaries of the domain is not large. Thank you very much in advance. Yours Sincerely, Arash -- *Arash Pourakaberian* *M.Sc. in Petroleum Production Engineering,* *Department of Chemical and Petroleum Engineering,* *Sharif University of Technology,* *Tehran, Iran* *che.sharif.edu/~pourakaberian.arash <http://che.sharif.edu/~pourakaberian.arash>*

2 2

Re: [firedrake] Is it possible to calculate a flux value on the edges of a cell in Firedrake?
by Lawrence Mitchell 17 Sep '20

17 Sep '20

Reinstating the firedrake mailing list cc (please do keep it cc'd since others may have more useful comments, or wish to know the answer). > On 15 Sep 2020, at 10:57, Mohammad Sarraf Joshaghani <m.sarraf.j(a)gmail.com> wrote: > > Thanks a lot for your prompt response and help. > >>> >>> The inner(grad(sol),n)*v in last two terms give me an error: >>> UFL:ERROR Shapes do not match: <Grad id=4986788992> and <Zero id=4998074176>. >>> >>> But the projection was okay when changed them to sol*v. Do you know where is this error coming from? >>> (FYI: my sol Function is defined on DG1 functionspace) >> >> I always mix up the definitions of jump(foo, n). I actually prefer just to write out what you want explicitly in terms of the restricted values. > > I think I was not clear here. My code works when define F as: > F = (fluxes*w)('+')*dS + fluxes*w*ds - avg(sol)*w('+')*dS - sol*w*ds‘ > > But when change it to a desired flux that contains grad(sol) such as below: > F = (fluxes*w)('+')*dS + fluxes*w*ds - avg(inner(grad(sol), n))*w('+')*dS - inner(grad(sol), n)*w*ds > > It returns error even when changing to grad(sol) only on ds term. > This is the error I get: > UFL:ERROR Shapes do not match: <Grad id=4986788992> and <Zero id=4998074176>. Somehow one of the two terms was simplified to zero (perhaps n is zero)? This from firedrake import * mesh = UnitSquareMesh(1, 1) V = FunctionSpace(mesh, "DG", 1) Vt = FunctionSpace(mesh, "HDiv Trace", 0) v = TestFunction(Vt) sol = Function(V) n = FacetNormal(mesh) inner(grad(sol), n)*v*ds assemble(inner(grad(sol), n)*v*ds) works for me, but perhaps I am doing something different. Thanks, Lawrence

2 3

Is it possible to calculate a flux value on the edges of a cell in Firedrake?
by Mohammad Sarraf Joshaghani 15 Sep '20

15 Sep '20

Hello, I appreciate if you help me with this question. I am trying to implement a postprocessing flux-limiting scheme for a DG advection problem to eliminate the under/overshoots that appears in the vicinity of the front. In order to achieve this, I need to iterate over each mesh element and then on each facet of the element calculate the value of numerical flux. Based on the sign of this Flux value, a constant edge-specific alpha is calculated and the new corrected solutions are then reconstructed for the cell. In firedrake, Is there any way to iterate over facets of each cell, and calculate an integral of an expression (e.g, avg(inner(grad(sol,n))) ) on that facet? This is how I achieved it in fenics: ----------------------------------------------------- facet_domains = MeshFunction('size_t',mesh,mesh.topology().dim()-1) dS = Measure('dS', subdomain_data = facet_domains) elnum = 0 for cell in cells(mesh): flux=np.zeros(3) for fct in facets(cell): flux[fct.index] = assemble( (avg(inner(grad(sol) , n) ) )*dS(fct.index()) print(flux) . . do some operations based on flux[fct.index] values . . elnum += 1 -------------------------------------------------------- Thanks, Sarraf

2 3

Re: [firedrake] Is it possible to calculate a flux value on the edges of a cell in Firedrake?
by Cotter, Colin J 14 Sep '20

14 Sep '20

Dear Mohammed, Probably your reconstruction involves a (probably local) solve in which case you can do anything that you can express in UFL as a surface integral. If you really need to compute the flux as a separate step then we have trace finite element spaces that you can use. all the best --cjc ________________________________ From: firedrake-bounces(a)imperial.ac.uk <firedrake-bounces(a)imperial.ac.uk> on behalf of Mohammad Sarraf Joshaghani <m.sarraf.j(a)gmail.com> Sent: 14 September 2020 08:01 To: firedrake <firedrake(a)imperial.ac.uk> Subject: [firedrake] Is it possible to calculate a flux value on the edges of a cell in Firedrake? Hello, I appreciate if you help me with this question. I am trying to implement a postprocessing flux-limiting scheme for a DG advection problem to eliminate the under/overshoots that appears in the vicinity of the front. In order to achieve this, I need to iterate over each mesh element and then on each facet of the element calculate the value of numerical flux. Based on the sign of this Flux value, a constant edge-specific alpha is calculated and the new corrected solutions are then reconstructed for the cell. In firedrake, Is there any way to iterate over facets of each cell, and calculate an integral of an expression (e.g, avg(inner(grad(sol,n))) ) on that facet? This is how I achieved it in fenics: ----------------------------------------------------- facet_domains = MeshFunction('size_t',mesh,mesh.topology().dim()-1) dS = Measure('dS', subdomain_data = facet_domains) elnum = 0 for cell in cells(mesh): flux=np.zeros(3) for fct in facets(cell): flux[fct.index] = assemble( (avg(inner(grad(sol) , n) ) )*dS(fct.index()) print(flux) . . do some operations based on flux[fct.index] values . . elnum += 1 -------------------------------------------------------- Thanks, Sarraf

1 0

SIAM CSE21 Code Generation Minisymposterium
by Vorderwuelbecke, Sophia 09 Sep '20

09 Sep '20

Dear Firedrake users, This is a friendly reminder that we are proposing a minisymposterium at SIAM CSE21<https://www.siam.org/conferences/cm/conference/cse21> due to be held March 1-5 2021. CSE21 will be held at least partly online, so this is a great opportunity for everyone in the code generation community to share their work, even if they won’t be in Fort Worth in March. Submitting a poster doesn’t count against the one presentation rule, so please consider submitting even if you are also presenting elsewhere at the conference. The proposed title and abstract are Automatic simulation code generation tools and their applications. Automatic code generation provides a mechanism for creating high performance simulations using sophisticated numerical methods at minimal development cost. Tools such as Firedrake, FEniCS, FreeFEM++, Devito and many others enable developers to work in a highly productive way at a high level of mathematical abstraction, while still creating optimised low-level code which is frequently faster than would have been written by hand. Code generation approaches also facilitate a separation of concerns in which new numerical methods or performance improvements become directly available to application developers without the need for repeated recoding. This minisymposterium provides the opportunity for developers and users of code generation frameworks to present recent developments in both the capabilities and applications of these frameworks. The poster session will also provide the opportunity for informal exchanges of ideas in this community, and to support dialog between developers, and existing and potential users of these frameworks. To take part in our minisymposterium you will need to fill out the survey https://imperial.eu.qualtrics.com/jfe/form/SV_efEu8ddYj1nicdv. The survey helps us to collect poster titles and authors for inclusion into our proposal. The proceeding for minisymposteria is the following: 'Participants must submit their posters using the Conference Management System<https://meetings.siam.org/start.cfm?CONFCODE=cse21>. The poster title should be submitted as “Minisymposterium: [insert poster title]”. After the posters are submitted to the Conference Management System, organizers must submit the minisymposterium proposal form<https://forms.gle/nS7h2mkBmNiZgopm6>.' Therefore you will need to fill out our survey by 20th Semptember, one day before SIAM's deadline for poster submissions. Kind regards, Sophia Vorderwuelbecke

1 0

Fw: Next HPC-Europa3 closing dates: 17th September, 12th November
by Cotter, Colin J 09 Sep '20

09 Sep '20

Dear Firedrakers, I don't expect many people are planning much travel right now, but just to say that if anyone is interested to visit Imperial (from an EU country or Associated State) as part of the scheme below I'd be very happy to discuss. Please contact me off-list for this. all the best --cjc ________________________________ From: INGLIS Catherine <c.inglis(a)epcc.ed.ac.uk> Sent: 09 September 2020 09:33 To: hpce3-hosts(a)mlist.is.ed.ac.uk <hpce3-hosts(a)mlist.is.ed.ac.uk> Subject: Next HPC-Europa3 closing dates: 17th September, 12th November Dear HPC-Europa hosts, I hope you are all still keeping well through these very unusual times. I would like to ask for your help, once again, in publicising the upcoming HPC-Europa3 closing dates. The next one is on 17th September (i.e. 9 days from now), but if this is a bit short notice for people, there will be another one soon after, on 12th November. The September closing date is usually the one for which we receive most applications, but many of us in the HPC-Europa3 team have been working on other things during the pandemic, as the visitor programme has effectively been paused. This means that we have not had much time to dedicate to publicising the programme, and this combined with the uncertainty around when it will be possible to travel normally again, means that I think we are likely to receive far fewer applications this September. However, we would like to emphasise that: * Applications are open as normal, and we are keen for people to continue applying to the programme. * There is more flexibility with visit dates currently – however, visits should be completed by 31st October 2021, unless there is an extension to the programme. * As people may not be able to start their visits very soon, we are able to allocate up to 20% of the resources in advance of the visit, so that people can get started on their project. Please publicise among your contacts. Your help really does make a difference! I’ve attached the Call as a pdf, but the text is also below, if that is easier. Thanks, Catherine HPC-Europa3 Transnational Access programme Collaborative research visits using High Performance Computing Call for applications: next closing dates 17th September and 12th November 2020 HPC-Europa3 funds research visits for computational scientists in any discipline which can use High Performance Computing (HPC). Visits can be made to research institutes in Finland, Germany, Greece, Ireland, Italy, the Netherlands, Spain or the UK. What does HPC-Europa3 provide? • Funding for travel, living and accommodation expenses for visits of up to 13 weeks. • Access to world-class High Performance Computing (HPC) facilities. • Technical support to help you make best use of the HPC systems. • Collaborative environment with an expert in your field of research. Who is the programme open to? • Researchers of all levels, from postgraduate to full professors. • Researchers from academia or industry. • Researchers currently working in a European Union country or Associated State (see http://bit.ly/AssociatedStates for full list of Associated States). * Researchers may not visit a group in the country where they currently work. * A small number of places are available for researchers working outside these countries – please contact staff(a)hpc-europa.org<mailto:staff@hpc-europa.org> for more information. How do I apply? For more information, and to apply online, visit: http://www.hpc-europa.eu The next closing dates are 17th September and 12th November 2020. Closing dates are held 4 times per year. Applications can be submitted at any time. You should receive a decision approximately 6 weeks after the closing date. Follow us on Twitter for project news: https://twitter.com/HPCEuropa3 Catherine Inglis EPCC, The University of Edinburgh Bayes Centre 47 Potterrow Edinburgh EH8 9BT Tel: +44 131 651 3578 / Skype: catherine.inglis / Twitter: @cafi_epcc --- HPC-Europa3: work with Europe's leading researchers - work on Europe's biggest computers. Apply for funding at www.hpc-europa.eu<http://www.hpc-europa.eu>. Next closing date 17th September 2020. Follow https://twitter.com/HPCEuropa3 to be kept informed of future closing dates. -- The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.

1 0

Help me
by Emmanuel Yarleque Medina 08 Sep '20

08 Sep '20

Hi users, I am interested in solving problems of high order EDP's using adaptivity strategies, such as h-adaptive and p-adaptive. I would like to know if there is a simple example of how to make such a strategy with Firedrake. Best regards, Emmanuel Y. Medina Ph.D. Student PPGMC - UFJF, Brazil

1 0

Fw: Second PRISM workshop on applications: Beyond CFD that will be held on Wed, 30 September 2020 at 14:00 – 16:00
by Cotter, Colin J 08 Sep '20

08 Sep '20

Dear Firedrakers, You may be interested in the PRISM workshop below. It has an odd title but it is really about composability, coupling and the outer loop. all the best --cjc ________________________________ From: Radomska, Anna <a.radomska(a)imperial.ac.uk> Sent: 08 September 2020 16:00 To: patrick.farrell <patrick.farrell(a)maths.ox.ac.uk>; Cotter, Colin J <colin.cotter(a)imperial.ac.uk>; Ham, David A <david.ham(a)imperial.ac.uk>; Kelly, Paul H J <p.kelly(a)imperial.ac.uk>; Cohen, Jeremy <jeremy.cohen(a)imperial.ac.uk>; Moxey, David <D.Moxey(a)exeter.ac.uk>; Gorman, Gerard J <g.gorman(a)imperial.ac.uk>; Piggott, Matthew D <m.d.piggott(a)imperial.ac.uk> Cc: Sherwin, Spencer J <s.sherwin(a)imperial.ac.uk> Subject: Second PRISM workshop on applications: Beyond CFD that will be held on Wed, 30 September 2020 at 14:00 – 16:00 Dear All, Please join us for the second PRISM workshop on applications: Beyond CFD that will be held on Wed, 30 September 2020 at 14:00 – 16:00. In this workshop we wish to share experiences on application of PDE solvers which go beyond CFD tools, for example composable solvers, stability analysis, inverse problems, data assimilation and uncertainty quantification. The format of the on-line workshop will involve a series of 3 short 10-minutes talks followed by small group discussions and a summary session. The event's programme includes: 2.00pm-2.10pm Matt Knepley (University at Buffalo) 2.10pm-2.20pm Patrick Farrell (University of Oxford) 2.20pm-2.30pm Koki Sagiyama (Imperial College London) on "Firedrake: Solving equations on subspaces" 2.30pm-3.30pm Small group discussions 3.30pm-3.45pm Summary session If you could kindly share this announcement through your network it would be greatly appreciated. To register please sign-up here<https://www.eventbrite.co.uk/e/beyond-cfd-tickets-117983109657> so you can receive a Microsoft Teams calendar invite which will be distributed closer to the date of the event. With best regards, Anna

1 0

Is there a way to calculate a flux on the edges of each cell in Firedrake
by Mohammad Sarraf Joshaghani 07 Sep '20

07 Sep '20

Hello, I am trying to implement a postprocessing flux-limiting scheme for a DG advection problem to eliminate the under/overshoots that appears in the vicinity of the front. In order to achieve this, I need to iterate over each mesh element and then on each facet of the element calculate the value of numerical flux. Based on the sign of this Flux value, a constant edge-specific alpha is calculated and the new corrected solutions are then reconstructed for the cell. In firedrake, Is there any way to iterate over facets of each cell, and calculate an integral of an expression (e.g, avg(inner(grad(sol,n))) ) on that facet? This is how I achieved it in fenics: ----------------------------------------------------- facet_domains = MeshFunction('size_t',mesh,mesh.topology().dim()-1) dS = Measure('dS', subdomain_data = facet_domains) elnum = 0 for cell in cells(mesh): flux=np.zeros(3) for fct in facets(cell): flux[fct.index] = assemble( (avg(inner(grad(sol) , n) ) )*dS(fct.index()) print(flux) . . do some operations based on flux[fct.index] values . . elnum += 1 -------------------------------------------------------- Thank you, Sarraf

1 0