July 2014 - firedrake - imperial.ac.uk

solver options
by Colin Cotter 05 Jul '14

05 Jul '14

Hi all, We have been trying to use the shiny new option to specify a different matrix for the solver, but within LinearVariationalSolver. With this usage, it fails for us, probably because we are mispecifying the options. The following code tries to solve it first using the solve() command, then with LinearVariationalSolver. https://gist.github.com/colinjcotter/fd72f3a838ce46d0fe7d Please can you tell us what we are doing wrong? cheers --cjc

3 2

Video of flow past a square
by Jacobs, Christian T 04 Jul '14

04 Jul '14

Adding to the collection of pretty Firedrake-related videos, here's a video of flow past a square: https://www.youtube.com/watch?v=JIBKMF6rK8s&feature=youtu.be Cheers, Christian

1 0

Re: [firedrake] compile problem
by Andrew McRae 04 Jul '14

04 Jul '14

On second thoughts, I think you're just missing a comma in front of the t=t? On 3 July 2014 18:57, Shipton, Jemma <j.shipton(a)imperial.ac.uk> wrote: > Hi All, > > Sorry to post a question in the chat and then disappear - I'd lost track > of time and needed to dash home. Anyway, I've copied and pasted my > question... > > 18:26 < jemma> I have pulled some code Colin added this morning > 18:27 < jemma> and it doesn't work - although Colin was quite happily > running > it before he went to watch tennis > 18:27 < jemma> can anyone see why there'd be a problem with this: > 18:27 < jemma> u = project(Expression(("1.0 + > > sin(pi*(x[0]-t))*sin(pi*x[1])*cos(pi*t/T0)","cos(pi*(x[0]-t))*cos(pi*x[1])*cos(pi*t/T0)")t=t,T0=T0),V1) > 18:28 < jemma> where V1 comes from: > 18:28 < jemma> V1_v = OuterProductElement(IUh1, IUv0) > 18:28 < jemma> V1_h = OuterProductElement(IUh0, IUv1) > 18:28 < jemma> V1_elt = HDiv(V1_v) + HDiv(V1_h) > 18:28 < jemma> V1 = FunctionSpace(mesh,V1_elt) > 18:28 < jemma> t and T0 are defined > 18:29 < jemma> maybe I'm missing a ufl/pyop2/something update - I haven't > updated in a while... > > When I run the code won't even compile: > > pyop2.exceptions.CompilationError: Unable to compile code > > I've attached the log and err files. The problem seems to be the variables > t and T0 in the expression - although they are defined in previous lines: > > t = 0.0 > T0 = 1.0 > > The script I'm trying to run is at: > https://bitbucket.org/colinjcotter/vertical-tests/src/c1a2a565912564897fdbe… > > Thanks! > > Jemma >

4 4

A fun video to end the day
by Andrew McRae 03 Jul '14

03 Jul '14

DG advection + extruded circle mesh + coordinate-field editing = https://dl.dropboxusercontent.com/u/62309663/tyre.mp4

1 0

Re: [firedrake] compile problem
by Andrew McRae 03 Jul '14

03 Jul '14

Can you try the FFC branch "const-access"? On 3 July 2014 18:57, Shipton, Jemma <j.shipton(a)imperial.ac.uk> wrote: > Hi All, > > Sorry to post a question in the chat and then disappear - I'd lost track > of time and needed to dash home. Anyway, I've copied and pasted my > question... > > 18:26 < jemma> I have pulled some code Colin added this morning > 18:27 < jemma> and it doesn't work - although Colin was quite happily > running > it before he went to watch tennis > 18:27 < jemma> can anyone see why there'd be a problem with this: > 18:27 < jemma> u = project(Expression(("1.0 + > > sin(pi*(x[0]-t))*sin(pi*x[1])*cos(pi*t/T0)","cos(pi*(x[0]-t))*cos(pi*x[1])*cos(pi*t/T0)")t=t,T0=T0),V1) > 18:28 < jemma> where V1 comes from: > 18:28 < jemma> V1_v = OuterProductElement(IUh1, IUv0) > 18:28 < jemma> V1_h = OuterProductElement(IUh0, IUv1) > 18:28 < jemma> V1_elt = HDiv(V1_v) + HDiv(V1_h) > 18:28 < jemma> V1 = FunctionSpace(mesh,V1_elt) > 18:28 < jemma> t and T0 are defined > 18:29 < jemma> maybe I'm missing a ufl/pyop2/something update - I haven't > updated in a while... > > When I run the code won't even compile: > > pyop2.exceptions.CompilationError: Unable to compile code > > I've attached the log and err files. The problem seems to be the variables > t and T0 in the expression - although they are defined in previous lines: > > t = 0.0 > T0 = 1.0 > > The script I'm trying to run is at: > https://bitbucket.org/colinjcotter/vertical-tests/src/c1a2a565912564897fdbe… > > Thanks! > > Jemma >

1 0

compile problem
by Shipton, Jemma 03 Jul '14

03 Jul '14

Hi All, Sorry to post a question in the chat and then disappear - I'd lost track of time and needed to dash home. Anyway, I've copied and pasted my question... 18:26 < jemma> I have pulled some code Colin added this morning 18:27 < jemma> and it doesn't work - although Colin was quite happily running it before he went to watch tennis 18:27 < jemma> can anyone see why there'd be a problem with this: 18:27 < jemma> u = project(Expression(("1.0 + sin(pi*(x[0]-t))*sin(pi*x[1])*cos(pi*t/T0)","cos(pi*(x[0]-t))*cos(pi*x[1])*cos(pi*t/T0)")t=t,T0=T0),V1) 18:28 < jemma> where V1 comes from: 18:28 < jemma> V1_v = OuterProductElement(IUh1, IUv0) 18:28 < jemma> V1_h = OuterProductElement(IUh0, IUv1) 18:28 < jemma> V1_elt = HDiv(V1_v) + HDiv(V1_h) 18:28 < jemma> V1 = FunctionSpace(mesh,V1_elt) 18:28 < jemma> t and T0 are defined 18:29 < jemma> maybe I'm missing a ufl/pyop2/something update - I haven't updated in a while... When I run the code won't even compile: pyop2.exceptions.CompilationError: Unable to compile code I've attached the log and err files. The problem seems to be the variables t and T0 in the expression - although they are defined in previous lines: t = 0.0 T0 = 1.0 The script I'm trying to run is at: https://bitbucket.org/colinjcotter/vertical-tests/src/c1a2a565912564897fdbe… Thanks! Jemma

1 0

Meeting tomorrow
by Lawrence Mitchell 03 Jul '14

03 Jul '14

Dear firedrakers, for the avoidance of doubt, those of us currently local to Imperial will be holding our weekly meeting tomorrow at 1pm in WPL 301. Lawrence

1 0

Fwd: [PRISM] PRISM Technical Seminar: 10th July 10am
by Lawrence Mitchell 02 Jul '14

02 Jul '14

Dear firedrakers, next week, I'll be continuing the PRISM technical seminar series with an overview of the design and implementation of Firedrake and PyOP2. If you're interested, do come along. 10am Thursday 10th July in Room 266 Roderic Hill (Aeronautic). See abstract below. Cheers, Lawrence -------- Original Message -------- Subject: [PRISM] PRISM Technical Seminar: 10th July 10am Date: Wed, 11 Jun 2014 15:34:11 +0100 From: Chris Cantwell <c.cantwell(a)imperial.ac.uk> Organization: Imperial College London To: <prism(a)imperial.ac.uk> Dear all, The next PRISM technical seminar will take place as follows: Speaker: Lawrence Mitchell Title: The design and implementation of Firedrake and PyOP2 Time: Thursday 10th July, 10am Place: TBC Abstract: The guiding principle in the development of PyOP2 and Firedrake is "get the abstraction right and everything else is easy". In this talk, I will present the abstractions in PyOP2 and and Firedrake (and their implementation). Starting with the data structures and parallel iteration construct in PyOP2 for execution on unstructured meshes, I will move on to explain how these handcuffs influence the design of Firedrake, and why we think they're a good thing. Although Firedrake itself is designed as a finite element package, the separation of mesh iteration into a PyOP2 level leads to the ability to "escape" out of the variational jail in a clean way, and I will describe why we might want to do so. Finally, I'll motivate why we don't want to implement any solvers ourselves, and describe how we interact with PETSc to provide us with both linear and non-linear solvers. Cheers, Chris -- Chris Cantwell Department of Aeronautics Roderic Hill Building Imperial College London South Kensington Campus London SW7 2AZ Email: c.cantwell(a)imperial.ac.uk www.imperial.ac.uk/people/c.cantwell _______________________________________________ prism mailing list prism(a)imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/prism

2 3

Looks like a relevant conference
by Colin Cotter 01 Jul '14

01 Jul '14

http://conan.iwr.uni-heidelberg.de/sppexa-workshop/

1 0