March 2014 - firedrake - imperial.ac.uk

last day at imperial
by Loriant, Nicolas 24 Mar '14

24 Mar '14

Hi, As you might already be aware of: this Wednesday will be my last day at the office. I will be heading to Eastside at 5:30pm, and I hope you will join me for drinks :) Cheers, Nicolas

1 0

Re: [firedrake] dmplex migration
by Fabio Luporini 24 Mar '14

24 Mar '14

For my branch: - pyop2-ir-perf-eval I choose option (c). It's still under development, but it's there just, as you said, for benchmarking. Thanks -- Fabio 2014-03-24 14:59 GMT+00:00 Mitchell, Lawrence < lawrence.mitchell(a)imperial.ac.uk>: > Dear firedrakers, > > if you've been paying attention recently, you'll have noticed we just > transitioned to a dmplex based mesh infrastructure. Now would appear to be > a good time to do a little bit of branch spring cleaning. For each of the > branches listed please tell me if you would like it: > > (a) kept and pulled forward onto master (still under active development) > (b) pulled forward onto master and proposed for merging > (c) left alone (under development but you'll do it yourself) > (d) removed (the branch is abandoned) > > I've organised the branches by "owner", that is the person who has made > the most commits on it, but please check the entire list. > > David's branches: > > - form_cache is in (a), but already rebased > - public_coordinates is in (b) > - real-function-space > Discussions suggest we need to actually implement these, but we probably > need to sit down and think about how to do it properly. > - wave-demo > This made the video on the front page of the website, should probably be > pulled forward and documented in the same way as the other demos. > > Doru's branches: > > - extrusion-facets-rhs > This appears under active development as in in review I think > - rebased-efr > I think this is the same functionality as extrusion-facets-rhs but > rebased on a different commit, abandon? > - extrusion-facets > Abandoned in favour of extrusion-facets-rhs? > - extrusion-documentation > In review but needing multiple fixes to text before merging > - extrusion-mixed-bcs > I think this is abandoned, it's got a blowing up discretisation of > extruded linear boussinesq we wrote for the October gung-ho meeting. > - ext_bc_demo > no clue > - extrusion-vfs > Has some tests, I don't know if they're in master > - extrusion-vfs-new > has some tests and some IR-related changes. > - extrusion > has some tests > > Florian's branches: > > - split-forms > Needs pulling forward and merging, I think. > - split-kernels > Abandoned in favour of split-forms? > - fenics-1.3 > Needs to come forward I think > - pytest-2.5.2 > Asks for newer pytest versions on travis > - petsc-ppa > Probably abandon ware, since we don't have petsc packages for now. > - moar-lint > Adds even tighter linking to travis > > > Lawrence's branches: > > - more-nullspace > In progress > - matrix-free > Needs updating to master, I don't know what state the stuff Rod has been > working on is in, that may be a better bet. > - blocked-matrices > Abandon ware, will drop. > > Fabio's branches: > > - pyop2-ir-perf-eval > A bunch of stuff implementing a benchmarking test suite I think. > > Colin's branches: > > - hydrostatic: > A test case where the test and trial spaces differ. Colin, do you want > to keep this one? > > Simon's branches: > > - firedrake-adjoint > Various bits and pieces of firedrake-adjoint testing, I think this can > be dropped. > > Cheers, > > Lawrence >

1 0

FEniCS '14 OuterProduct/Extrusion abstract
by Andrew McRae 20 Mar '14

20 Mar '14

I've set up a FEniCS '14 repo at https://bitbucket.org/mapdes/fenics14 Draft version of our submission attached. Order of authors not yet decided.

1 0

first tests
by Colin Cotter 14 Mar '14

14 Mar '14

Dear Firedrakers, Thanks to great work from Andrew and Doru with lots of help from Lawrence and others, we now have enough stuff working in the extruded meshes to do some first testing of my planned discretisation methods for the weather prediction applications. I got quite excited about this, so I made a video: https://www.youtube.com/watch?v=SB0ufc77V84&feature=youtu.be Displayed is the temperature field (actually the perturbation from a stratified mean temperature) in a vertical channel. The temperature field is a P0xP1 tensor product field that turns out to be necessary for large scale weather applications, and has only been possible due to Andrew's enhancements of UFL/FIAT/FFC. It's a fairly shocking resolution, sorry. all the best --Colin

1 0

R space
by Cotter, Colin J 12 Mar '14

12 Mar '14

Dear firedrakers, Can we have a discussion about the "R" space (function space of global constants), which if I understand correctly has not been implemented? Clearly, the R space is not good for performance, since it inserts an entire column or row into the matrix of non-zero entries, and the high performance solution is to solve the problem with a Krylov solver by projecting out the constant from the solver as Lawrence has recently implemented. However, the R space is useful for verifying other parts of Firedrake by e.g. computing eigenvalues of the Laplacian etc. I argue that we want this feature for testing, even though we don't expect to use it for big production code. --cjc

2 2

Re: [firedrake] R space
by David Ham 12 Mar '14

12 Mar '14

The challenge here is having a sane parallel implementation. In essence R causes the halo to expand to cover the entire domain. Directly implementing that would (a) be a lot of work and (b) suck from a performance perspective. You might argue that you don't care about (b) but (a) is still a problem. David On 12 March 2014 06:35, Cotter, Colin J <colin.cotter(a)imperial.ac.uk> wrote: > Dear firedrakers, > Can we have a discussion about the "R" space (function space of global > constants), which if I understand correctly has not been implemented? > > Clearly, the R space is not good for performance, since it inserts an > entire column or row into the matrix of non-zero entries, and the high > performance solution is to solve the problem with a Krylov solver by > projecting out the constant from the solver as Lawrence has recently > implemented. However, the R space is useful for verifying other parts of > Firedrake by e.g. computing eigenvalues of the Laplacian etc. I argue that > we want this feature for testing, even though we don't expect to use it for > big production code. > > --cjc > -- Dr David Ham Departments of Mathematics and Computing Imperial College London http://www.imperial.ac.uk/people/david.ham

2 1

snakes
by Kelly, Paul H J 11 Mar '14

11 Mar '14

Perhaps some of you saw this already... http://www.mcsweeneys.net/articles/faq-the-snake-fight-portion-of-your-thes…

1 0

Agenda for tomorrow's PRISM workshop RSM G41
by David Ham 11 Mar '14

11 Mar '14

Hi All, Here's tomorrows agenda. It's a great lineup of speakers. Attendance is on pain of pain ;) 1400 Martin Alnæs (FEniCS) 1430 Jon Hill (Fluidity) 1450 Dave Moxey (Nektar++) 1510 Lawrence Mitchell (Firedrake) 1530 Coffee 1550 Chris Cantwell (Nektar++) 1610 Michael Lange (Fluidity/Firedrake) 1630 Peter Vincent (PyFR) 1700 Eastside -- Dr David Ham Departments of Mathematics and Computing Imperial College London http://www.imperial.ac.uk/people/david.ham

1 0

PRISM workshop next Wed March 12 2-5:30
by Kelly, Paul H J 06 Mar '14

06 Mar '14

The PRISM project is a "Platform Grant" aiming to integrate and push forward Firedrake, Fluidity, PyFR and Nektar++. It provides a background of funding for what we and our friends do over the next five years. This meeting will have talk s from all four groups + Martin. Plan for beer for afterwards. You really should be there. Paul

1 0

numerics and HPC meeting at UCL 14-15 April
by Lawrence Mitchell 05 Mar '14

05 Mar '14

Hi all, this meeting may be of interest to some. http://www.maths.manchester.ac.uk/news-and-events/events/anahpc13/ Lawrence

1 0