June 2014 - firedrake - imperial.ac.uk

GitHub private repositories for OP2 and firedrake
by Florian Rathgeber 30 Jun '14

30 Jun '14

I have applied for an academic upgrade and we have been granted a free silver plan with 20 private repositories each for the OP2 and firedrake organisations. Florian

1 0

Re: [firedrake] Fwd: [FEniCS] Summer intern to work on python 3 support / FIAT PR
by David Ham 30 Jun '14

30 Jun '14

Does this actually change anything we care about, or is it orthogonal? On 29 June 2014 21:27, McRae, Andrew <a.mcrae12(a)imperial.ac.uk> wrote: > Originally posted to the FEniCS mailing list. This is relevant for us > (or maybe just "me"), too. > > Andrew > > ---------- Forwarded message ---------- > From: Nico Schlömer <nico.schloemer(a)gmail.com> > Date: 29 June 2014 21:18 > Subject: Re: [FEniCS] Summer intern to work on python 3 support > To: Martin Sandve Alnæs <martinal(a)simula.no> > Cc: Jan Blechta <blechta(a)karlin.mff.cuni.cz>, Aslak Bergersen < > aslak.bergersen(a)gmail.com>, "fenics(a)fenicsproject.org" < > fenics(a)fenicsproject.org> > > > FWIW, I created a pull request that improves FIAT and removes almost > all of ScientificPython, cf. > < > https://bitbucket.org/fenics-project/fiat/pull-request/2/remove-almost-all-… > >. > If this gets reviewed (and possibly improved) soon, we might get to > Python3-ready FEniCS quiet quick. > > Cheers, > Nico > -- Dr David Ham Departments of Mathematics and Computing Imperial College London http://www.imperial.ac.uk/people/david.ham

1 0

Fwd: [FEniCS] Summer intern to work on python 3 support / FIAT PR
by Andrew McRae 29 Jun '14

29 Jun '14

Originally posted to the FEniCS mailing list. This is relevant for us (or maybe just "me"), too. Andrew ---------- Forwarded message ---------- From: Nico Schlömer <nico.schloemer(a)gmail.com> Date: 29 June 2014 21:18 Subject: Re: [FEniCS] Summer intern to work on python 3 support To: Martin Sandve Alnæs <martinal(a)simula.no> Cc: Jan Blechta <blechta(a)karlin.mff.cuni.cz>, Aslak Bergersen < aslak.bergersen(a)gmail.com>, "fenics(a)fenicsproject.org" < fenics(a)fenicsproject.org> FWIW, I created a pull request that improves FIAT and removes almost all of ScientificPython, cf. < https://bitbucket.org/fenics-project/fiat/pull-request/2/remove-almost-all-… >. If this gets reviewed (and possibly improved) soon, we might get to Python3-ready FEniCS quiet quick. Cheers, Nico

1 0

petsc
by Colin Cotter 25 Jun '14

25 Jun '14

Hi Folks, I've tried to update my petsc4py. I started by re-pip-installing petsc, and then ran "sudo python setup.py --install" in an updated petsc4py directory, and got the output which can be read at the bottom of this email, which was unexpected. cheers --cjc cjc1@ae-gauss:~/firedrake/petsc4py$ sudo python setup.py install running install running build running build_src running build_py running build_ext PETSC_DIR: /usr/local/lib/python2.7/dist-packages/petsc PETSC_ARCH: scalar-type: real precision: double language: CONLY compiler: /usr/bin/mpicc.openmpi linker: /usr/bin/mpicc.openmpi building 'PETSc' extension /usr/bin/mpicc.openmpi -pthread -fPIC -Wall -Wwrite-strings -Wno-strict-aliasing -Wno-unknown-pragmas -O -fPIC -fno-strict-aliasing -DNDEBUG -g -fwrapv -O2 -Wall -g -fstack-protector --param=ssp-buffer-size=4 -Wformat -Wformat-security -Werror=format-security -DPETSC_DIR=/usr/local/lib/python2.7/dist-packages/petsc -I/usr/lib/openmpi/include/openmpi -I/usr/lib/openmpi/include -I/usr/local/lib/python2.7/dist-packages/petsc/include -Isrc/include -I/usr/local/lib/python2.7/dist-packages/numpy/core/include -I/usr/include/python2.7 -c src/PETSc.c -o build/temp.linux-x86_64-2.7/src/PETSc.o In file included from /usr/local/lib/python2.7/dist-packages/numpy/core/include/numpy/ndarraytypes.h:1761:0, from /usr/local/lib/python2.7/dist-packages/numpy/core/include/numpy/ndarrayobject.h:17, from /usr/local/lib/python2.7/dist-packages/numpy/core/include/numpy/arrayobject.h:4, from src/include/petsc4py/numpy.h:11, from src/petsc4py.PETSc.c:347, from src/PETSc.c:3: /usr/local/lib/python2.7/dist-packages/numpy/core/include/numpy/npy_1_7_deprecated_api.h:15:2: warning: #warning "Using deprecated NumPy API, disable it by " "#defining NPY_NO_DEPRECATED_API NPY_1_7_API_VERSION" [-Wcpp] In file included from src/PETSc.c:3:0: src/petsc4py.PETSc.c: In function ‘__pyx_pf_8petsc4py_5PETSc_5LGMap_8create’: src/petsc4py.PETSc.c:71088:3: warning: passing argument 3 of ‘ISLocalToGlobalMappingCreate’ makes pointer from integer without a cast [enabled by default] /usr/local/lib/python2.7/dist-packages/petsc/include/petscis.h:143:29: note: expected ‘const PetscInt *’ but argument is of type ‘PetscInt’ src/petsc4py.PETSc.c:71088:3: error: incompatible type for argument 4 of ‘ISLocalToGlobalMappingCreate’ /usr/local/lib/python2.7/dist-packages/petsc/include/petscis.h:143:29: note: expected ‘PetscCopyMode’ but argument is of type ‘PetscInt *’ src/petsc4py.PETSc.c:71088:3: error: incompatible type for argument 5 of ‘ISLocalToGlobalMappingCreate’ /usr/local/lib/python2.7/dist-packages/petsc/include/petscis.h:143:29: note: expected ‘struct _p_ISLocalToGlobalMapping **’ but argument is of type ‘PetscCopyMode’ src/petsc4py.PETSc.c:71088:3: error: too many arguments to function ‘ISLocalToGlobalMappingCreate’ /usr/local/lib/python2.7/dist-packages/petsc/include/petscis.h:143:29: note: declared here src/petsc4py.PETSc.c: In function ‘__pyx_pf_8petsc4py_5PETSc_5LGMap_14getBlockSize’: src/petsc4py.PETSc.c:71405:3: warning: implicit declaration of function ‘ISLocalToGlobalMappingGetBlockSize’ [-Wimplicit-function-declaration] src/petsc4py.PETSc.c: In function ‘__pyx_pf_8petsc4py_5PETSc_5LGMap_18getBlockIndices’: src/petsc4py.PETSc.c:71752:3: warning: implicit declaration of function ‘ISLocalToGlobalMappingGetBlockIndices’ [-Wimplicit-function-declaration] src/petsc4py.PETSc.c:71803:7: warning: implicit declaration of function ‘ISLocalToGlobalMappingRestoreBlockIndices’ [-Wimplicit-function-declaration] src/petsc4py.PETSc.c: In function ‘__pyx_pf_8petsc4py_5PETSc_5LGMap_22getBlockInfo’: src/petsc4py.PETSc.c:72204:3: warning: implicit declaration of function ‘ISLocalToGlobalMappingGetBlockInfo’ [-Wimplicit-function-declaration] src/petsc4py.PETSc.c:72259:7: warning: implicit declaration of function ‘ISLocalToGlobalMappingRestoreBlockInfo’ [-Wimplicit-function-declaration] src/petsc4py.PETSc.c: In function ‘__pyx_pf_8petsc4py_5PETSc_5LGMap_26applyBlock’: src/petsc4py.PETSc.c:72724:3: warning: implicit declaration of function ‘ISLocalToGlobalMappingApplyBlock’ [-Wimplicit-function-declaration] src/petsc4py.PETSc.c: In function ‘__pyx_pf_8petsc4py_5PETSc_5LGMap_32applyBlockInverse’: src/petsc4py.PETSc.c:73315:3: warning: implicit declaration of function ‘ISGlobalToLocalMappingApplyBlock’ [-Wimplicit-function-declaration] src/petsc4py.PETSc.c: In function ‘__pyx_pf_8petsc4py_5PETSc_4DMDA_62getScatter’: src/petsc4py.PETSc.c:186920:3: error: too few arguments to function ‘DMDAGetScatter’ /usr/local/lib/python2.7/dist-packages/petsc/include/petscdmda.h:71:29: note: declared here error: command '/usr/bin/mpicc.openmpi' failed with exit status 1

1 1

reproducible parallel summation
by Kelly, Paul H J 23 Jun '14

23 Jun '14

Further to various discussions of reproducibility, I thought you might be interested in this paper: http://hal.archives-ouvertes.fr/docs/00/94/93/55/PDF/superaccumulator.pdf Full-Speed Deterministic Bit-Accurate Parallel Floating-Point Summation on Multi- and Many-Core Architectures Abstract. On modern multi-core, many-core, and heterogeneous architectures, ﬂoating-point computations, especially reductions, may become non-deterministic and thus non-reproducible mainly due to non-associativity of ﬂoating-point operations. We introduce a solution to compute deterministic sums of ﬂoating-point numbers efﬁciently and with the best possible accuracy. Our multi-level algorithm consists of two main stages: a ﬁltering stage that uses fast vectorized ﬂoating-point expansions; an accumulation stage based on superaccumulators in a high-radix carry-save representation. We present implementations on recent Intel desktop and server processors, on Intel Xeon Phi accelerator, and on AMD and NVIDIA GPUs. We show that the numerical reproducibility and bit-perfect accuracy can be achieved at no additional cost for large sums that have dynamic ranges of up to 90 orders of magnitude by leveraging arithmetic units that are left underused by standard reduction algorithms. Paul

6 7

Fwd: Math Libraries
by Kelly, Paul H J 21 Jun '14

21 Jun '14

Meanwhile here is another - thank you to Freddie Witherden for both of these finds: "The Metalibm project provides a tool for the automatic implementation of mathematical (libm) functions. A function f is automatically transformed into Gappa-certified C code implementing an approximation polynomial in a given domain with given accuracy." Code: http://lipforge.ens-lyon.fr/www/metalibm/ Slides: http://tamadiwiki.ens-lyon.fr/tamadiwiki/images/d/d8/DeDinechin2013.pdf The context of this was the "STOKE" PLDI paper I mentioned last week, http://web.stanford.edu/~sharmar/pubs/fp.pdf The point being that MetaLibc produces approximated functions with proven properties, while STOKE stochastically searches for implementations that are "close enough" under testing. Paul Begin forwarded message: From: "Witherden, Freddie" <freddie.witherden08(a)imperial.ac.uk<mailto:freddie.witherden08@imperial.ac.uk>> Date: 21 June 2014 12:43:12 GMT+01:00 To: "Kelly, Paul H J" <p.kelly(a)imperial.ac.uk<mailto:p.kelly@imperial.ac.uk>> Subject: Math Libraries Hi Paul, With regards to our discussion last week on math libraries (and how you can apparently get away with flipping an instruction here and there) I am wondering if you have heard about metalibm: http://lipforge.ens-lyon.fr/www/metalibm/ http://tamadiwiki.ens-lyon.fr/tamadiwiki/images/d/d8/DeDinechin2013.pdf It is from the same group that developed crlibm. Regards, Freddie.

1 0

Re: [firedrake] bcs
by Andrew McRae 17 Jun '14

17 Jun '14

Can you elaborate? 'top' and 'bottom' are there by default, otherwise you rely on markers on the base mesh (1 to 4 on a square) On 17 Jun 2014 12:30, "Cotter, Colin J" <colin.cotter(a)imperial.ac.uk> wrote: > So, what if I want to do the same for an extruded mesh? > > --cjc > ________________________________________ > From: Cotter, Colin J > Sent: 17 June 2014 09:14 > To: firedrake > Subject: RE: [firedrake] bcs > > doh! thanks. > ________________________________________ > From: firedrake-bounces(a)imperial.ac.uk [firedrake-bounces(a)imperial.ac.uk] > on behalf of Lawrence Mitchell [lawrence.mitchell(a)imperial.ac.uk] > Sent: 17 June 2014 09:00 > To: firedrake > Subject: Re: [firedrake] bcs > > > On 17 Jun 2014, at 09:45, "Cotter, Colin J" <colin.cotter(a)imperial.ac.uk> > wrote: > > > > from firedrake import * > > mesh = UnitSquareMesh(3,3,reorder=False) > > V = FunctionSpace(mesh,"CG",1) > > bc = DirichletBC(V,0.,"1") > > 1 not "1" > > > > print bc.nodes > > _______________________________________________ > firedrake mailing list > firedrake(a)imperial.ac.uk > https://mailman.ic.ac.uk/mailman/listinfo/firedrake >

2 1

bcs
by Cotter, Colin J 17 Jun '14

17 Jun '14

Dear firedrakers, The following code: from firedrake import * mesh = UnitSquareMesh(3,3,reorder=False) V = FunctionSpace(mesh,"CG",1) bc = DirichletBC(V,0.,"1") print bc.nodes produces: Traceback (most recent call last): File "test.py", line 5, in <module> print bc.nodes File "/home/cjc1/firedrake/firedrake/firedrake/utils.py", line 43, in __get__ obj.__dict__[self.__name__] = result = self.fget(obj) File "/home/cjc1/firedrake/firedrake/firedrake/bcs.py", line 155, in nodes fs._mesh.exterior_facets.subset(self.sub_domain).indices, File "/home/cjc1/firedrake/firedrake/firedrake/mesh.py", line 185, in subset markers = as_tuple(markers, int) File "/home/cjc1/firedrake/PyOP2/pyop2/utils.py", line 62, in as_tuple raise TypeError("Items need to be of type %s" % type) TypeError: Items need to be of type <type 'int'> Is this a known bug? --cjc

2 3

Demo illustrating more complicated forms
by Marie E. Rognes 16 Jun '14

16 Jun '14

Dear Firedrakers, Please consider the below if you ever need a more complicated variational form for stress testing and benchmarking. https://gist.github.com/meg-simula/3ab3fd63264c8cf1912b Best regards, Marie -- Marie E. Rognes, PhD Head of Department Biomedical Computing Department Simula Research Laboratory P.0.Box 134, 1325 Lysaker, Norway http://home.simula.no/~meg/ Skype: m.e.rognes Cell: +47 45 66 23 96

1 0

Rob Kirby seminar 1200 TODAY!
by David Ham 13 Jun '14

13 Jun '14

Hi All, We have a really great seminar at Midday today in Huxley. I hope to see you all there! David "Mixed finite Element methods for wave equations" 13th June 2014, 12-13, 130 Huxley. Dr Robert Kirby, Department of Mathematics, Baylor University, USA Abstract: Wave equations model many important applications, such as acoustics and oceanic tides. Mixed finite element methods discretize the conservative first-order form of such equations and preserve important mathematical and physical structure. I will begin by surveying some of the basic features of these discretizations. Then, I will present new semi discrete energy estimates for linear tide models. These estimates give long-term stability and, in the absence of forcing terms, give exponential decay of solution profiles to a steady, solenoidal field. However, these results still leave important practical details related to algorithms and software open, such as time-stepping and enabling the right kinds of operations efficiently. I will conclude by surveying some of the challenges that thus presented to software tools such as FEniCS and Firedrake. http://www3.imperial.ac.uk/newsandeventspggrp/imperialcollege/naturalscienc… -- Dr David Ham Departments of Mathematics and Computing Imperial College London http://www.imperial.ac.uk/people/david.ham

1 0