September 2015 - firedrake

supply Jp
by Colin Cotter 22 Sep '15

22 Sep '15

Dear all, I would like to supply a preconditioning matrix for Newton that depends on current state. Does it do the Right Thing if I just use the field that is being solved into for this? cheers --cjc -- http://www.imperial.ac.uk/people/colin.cotter www.cambridge.org/9781107663916

2 1

Re: [firedrake] Slow compilation with fd_bendy
by Fabio Luporini 18 Sep '15

18 Sep '15

Mmm, that is weird. Something odd is going on, I'm curious. At least O1 should be way faster than O2. Are you up-to-date? Will look into it, and will get back to you soon. Thanks! -- Fabio 2015-09-17 18:05 GMT+01:00 Tuomas Karna <tuomas.karna(a)gmail.com>: > Thanks Fabio, > > "parameters["coffee"] = {}" makes a huge difference, compiles in 1 min. > Using O1 is almost as slow as O2 (~13 min). > > - Tuomas > > > On 09/17/2015 04:09 AM, Fabio Luporini wrote: > > Hello Tuomas! > > For disabling coffee adding "parameters["coffee"] = {}" at the top of your > file should do the trick. > > You could retain some caffeine while decreasing code generation time by > instead setting "parameters["coffee"] = {'O1'=True}" (note: default is > O2=True, which is more aggressive - as expected - than O1) > > Cheers > > -- Fabio > > 2015-09-16 22:42 GMT+01:00 Tuomas Karna <tuomas.karna(a)gmail.com>: > >> Hello Firedrakers, >> >> I'm using "fd_bendy" branches of FFC and UFL, and it takes ~15 min to >> compile the attached example (with master branches it takes only ~1 >> min). Is there any way to speed it up? Are there some optimization flags >> I can turn off? I'm running small test cases so runtime efficiency is >> not a concern. >> >> Cheers, >> >> Tuomas >> >> > > > _______________________________________________ > firedrake mailing listfiredrake@imperial.ac.ukhttps://mailman.ic.ac.uk/mailman/listinfo/firedrake > > >

2 1

Re: [firedrake] Slow compilation with fd_bendy
by Andrew McRae 17 Sep '15

17 Sep '15

Interesting, thanks. I didn't expect the FFC:COFFEE ratio to be so form-dependent. On 17 September 2015 at 16:45, Lawrence Mitchell < lawrence.mitchell(a)imperial.ac.uk> wrote: > > > On 17 Sep 2015, at 12:14, Andrew McRae <A.T.T.McRae(a)bath.ac.uk> wrote: > > > > COFFEE shouldn't be particularly slow unless you forgot to update in the > last 3 months... > > My timings suggested almost all the time is spent in coffee (O2) > > Lawrence > _______________________________________________ > firedrake mailing list > firedrake(a)imperial.ac.uk > https://mailman.ic.ac.uk/mailman/listinfo/firedrake >

2 1

Re: [firedrake] Slow compilation with fd_bendy
by Fabio Luporini 17 Sep '15

17 Sep '15

Hello Tuomas! For disabling coffee adding "parameters["coffee"] = {}" at the top of your file should do the trick. You could retain some caffeine while decreasing code generation time by instead setting "parameters["coffee"] = {'O1'=True}" (note: default is O2=True, which is more aggressive - as expected - than O1) Cheers -- Fabio 2015-09-16 22:42 GMT+01:00 Tuomas Karna <tuomas.karna(a)gmail.com>: > Hello Firedrakers, > > I'm using "fd_bendy" branches of FFC and UFL, and it takes ~15 min to > compile the attached example (with master branches it takes only ~1 > min). Is there any way to speed it up? Are there some optimization flags > I can turn off? I'm running small test cases so runtime efficiency is > not a concern. > > Cheers, > > Tuomas > >

2 1

Re: [firedrake] Slow compilation with fd_bendy
by Andrew McRae 17 Sep '15

17 Sep '15

COFFEE shouldn't be particularly slow unless you forgot to update in the last 3 months... On 17 September 2015 at 12:09, Fabio Luporini <f.luporini12(a)imperial.ac.uk> wrote: > Hello Tuomas! > > For disabling coffee adding "parameters["coffee"] = {}" at the top of your > file should do the trick. > > You could retain some caffeine while decreasing code generation time by > instead setting "parameters["coffee"] = {'O1'=True}" (note: default is > O2=True, which is more aggressive - as expected - than O1) > > Cheers > > -- Fabio > > 2015-09-16 22:42 GMT+01:00 Tuomas Karna <tuomas.karna(a)gmail.com>: > >> Hello Firedrakers, >> >> I'm using "fd_bendy" branches of FFC and UFL, and it takes ~15 min to >> compile the attached example (with master branches it takes only ~1 >> min). Is there any way to speed it up? Are there some optimization flags >> I can turn off? I'm running small test cases so runtime efficiency is >> not a concern. >> >> Cheers, >> >> Tuomas >> >> >

2 1

Re: [firedrake] undefined symbol: kspfgmresmodifypcksp_
by David Ham 17 Sep '15

17 Sep '15

Just as a side idea, apt-get is *not* required in order to use the install script, so you could try running that script and see if it Just Works. If you are concerned to make sure it never tries to run apt, you can pass the --no_package_manager option to the install script. On Thu, 17 Sep 2015 at 12:03 Mitchell, Lawrence < lawrence.mitchell(a)imperial.ac.uk> wrote: > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > On 17/09/15 11:57, Buesing, Henrik wrote: > > > > > >> -----Ursprüngliche Nachricht----- Von: > >> firedrake-bounces(a)imperial.ac.uk [mailto:firedrake- > >> bounces(a)imperial.ac.uk] Im Auftrag von Buesing, Henrik Gesendet: > >> 17 September 2015 12:29 An: firedrake(a)imperial.ac.uk Betreff: Re: > >> [firedrake] undefined symbol: kspfgmresmodifypcksp_ > >> > >> > >> > >>> -----Ursprüngliche Nachricht----- Von: > >>> firedrake-bounces(a)imperial.ac.uk [mailto:firedrake- > >>> bounces(a)imperial.ac.uk] Im Auftrag von Buesing, Henrik > >>> Gesendet: 17 September 2015 11:43 An: firedrake(a)imperial.ac.uk > >>> Betreff: Re: [firedrake] undefined symbol: > >>> kspfgmresmodifypcksp_ > >>> > >>>> -----Ursprüngliche Nachricht----- Von: > >>>> firedrake-bounces(a)imperial.ac.uk [mailto:firedrake- > >>>> bounces(a)imperial.ac.uk] Im Auftrag von Buesing, Henrik > >>>> Gesendet: 17 September 2015 10:53 An: > >>>> firedrake(a)imperial.ac.uk Betreff: Re: [firedrake] undefined > >>>> symbol: kspfgmresmodifypcksp_ > >>>> > >>>> @Lawrence: I will go through your tests. At the moment I try > >>>> to use my own petsc version (and not the pip installed one) > >>>> for petsc4py. This crashed because I did not have > >>>> "--download-ctetgen --download-triangle - -download-chaco". > >>>> I'm doing a petsc ./configure atm. I will see how that goes. > >>> > >>> [Buesing, Henrik] This gave at least different errors (see the > >>> attached pyop2 test). I get "Import errors" No module named > >>> sparsity cannot import name plan No module named computeind > >> > >> [Buesing, Henrik] There are shared object files > >> > >> computeind.so plan.so sparsity.so > >> > >> in the pyop2 directory > >> > >> build/lib.linux-x86_64-2.7/pyop2 > >> > >> Maybe he does not find these files somehow? > >> > > > > [Buesing, Henrik] I have added the above directory to the > > PYTHONPATH. He now does not complain about sparsity anymore. I get > > a > > > > Intel MKL FATAL ERROR: Cannot load libmkl_mc3.so or libmkl_def.so. > > > > Well, I guess that has to do with Petsc and its blas-lapack-dir. I > > will try to reconfigure with --download-fblaslapack and not set the > > directory. > > I would start by trying to get petsc4py working on top of the PETSc > you built manually (rather than through pip), which I think seemed to > be working. Then we need to build PyOP2 with the correct > PETSc/petsc4py combo. The MKL error may be similar to this error > mentioned in the PETSc FAQ: > > http://www.mcs.anl.gov/petsc/documentation/faq.html#libimf > > Lawrence > -----BEGIN PGP SIGNATURE----- > Version: GnuPG v1 > > iQEcBAEBAgAGBQJV+p3UAAoJECOc1kQ8PEYvNuwH/iuTymdAT0MofwrnwRV1iRwe > kNj9tC3mN3E3pXVnCml8Wc4IvdGYGM0wgdlk8glUtQSyJNu+YoNuK7KRRTUhsNUr > GLwxX61SV9kuwXfk5XjXjmvLGoJlxn4275jslNlUz8vXRAgajY46pPqXlb7v4D/H > Et5jkHirDVC29lM42P9KmtCmX4zxjZGUpNrUAGQxRm6nbAeA+cplifHJ6zlb1Pf0 > ewU4adl2ywxagtvX8Nxy5uIQt3pJmF6QX+nJBsei1OjVsoFMGMz869nEZuG6jqN9 > yxkbno5PBS8LsLTAyVYsYScRo1pxWE8uaq7T4TMWrQSE+/j/85l5ZapOKuoUvmo= > =WUld > -----END PGP SIGNATURE----- > > _______________________________________________ > firedrake mailing list > firedrake(a)imperial.ac.uk > https://mailman.ic.ac.uk/mailman/listinfo/firedrake >

2 1

undefined symbol: kspfgmresmodifypcksp_
by Buesing, Henrik 17 Sep '15

17 Sep '15

Dear Firedrake Developers, I tried to start the helmholtz.py example, but I get the following error: File "helmholtz.py", line 47, in <module> from firedrake import * File "/home/user/.local/lib/python2.7/site-packages/firedrake/__init__.py", line 8, in <module> import firedrake.petsc as petsc File "/home/user/.local/lib/python2.7/site-packages/firedrake/petsc.py", line 5, in <module> petsc4py.init(sys.argv) File "/home/user/.local/lib/python2.7/site-packages/petsc4py/__init__.py", line 42, in init PETSc = petsc4py.lib.ImportPETSc(arch) File "/home/user/.local/lib/python2.7/site-packages/petsc4py/lib/__init__.py", line 29, in ImportPETSc return Import('petsc4py', 'PETSc', path, arch) File "/home/user/.local/lib/python2.7/site-packages/petsc4py/lib/__init__.py", line 64, in Import module = imp.load_module(fullname, fo, fn, stuff) ImportError: /home/user/.local/lib/python2.7/site-packages/petsc/lib/libpetsc.so.3.06: undefined symbol: kspfgmresmodifypcksp_ Is there anything I can do about it? Thank you! Henrik -- Dipl.-Math. Henrik Büsing Applied Geophysics and Geothermal Energy E.ON Energy Research Center RWTH Aachen University ----------------------------------------------- Mathieustr. 10 | Tel +49 (0)241 80 49907 52074 Aachen, Germany | Fax +49 (0)241 80 49889 ----------------------------------------------- http://www.eonerc.rwth-aachen.de/GGE hbuesing(a)eonerc.rwth-aachen.de -----------------------------------------------

2 9

Slow compilation with fd_bendy
by Tuomas Karna 17 Sep '15

17 Sep '15

Hello Firedrakers, I'm using "fd_bendy" branches of FFC and UFL, and it takes ~15 min to compile the attached example (with master branches it takes only ~1 min). Is there any way to speed it up? Are there some optimization flags I can turn off? I'm running small test cases so runtime efficiency is not a concern. Cheers, Tuomas

2 1

Re: [firedrake] undefined symbol: kspfgmresmodifypcksp_
by David Ham 17 Sep '15

17 Sep '15

Hi Henrik, Welcome to Firedrake. That's quite an odd error you have there. The symbol kspfgmresmodifypcksp_ should be defined in libpetsc itself, and indeed that does work in the firedrake install on my machine updated this morning. Can we just get some basic information here. Which OS is this? If Linux, which distribution and release? How did you install Firedrake? Is this using the install script (preferred, but it doesn't look like it from your log) or are you using the step by step instructions? Specifically which version of petsc is this (ie which command did you type to install petsc). Cheers, David On Thu, 17 Sep 2015 at 09:07 Buesing, Henrik <HBuesing(a)eonerc.rwth-aachen.de> wrote: > I tried to rebuild everything from scratch, but the error persists. It > already occurs in the tests for pyop2 (see attachement). > > > Henrik > > > > > > *Von:* firedrake-bounces(a)imperial.ac.uk [mailto: > firedrake-bounces(a)imperial.ac.uk] *Im Auftrag von *Buesing, Henrik > *Gesendet:* 16 September 2015 23:19 > *An:* firedrake(a)imperial.ac.uk > *Betreff:* [firedrake] undefined symbol: kspfgmresmodifypcksp_ > > > > Dear Firedrake Developers, > > I tried to start the *helmholtz.py* example, but I get the following > error: > > > > > > > > > > > > > > > > * File "helmholtz.py", line 47, in <module> from firedrake import * > File "/home/user/.local/lib/python2.7/site-packages/firedrake/__init__.py", > line 8, in <module> import firedrake.petsc as petsc File > "/home/user/.local/lib/python2.7/site-packages/firedrake/petsc.py", line 5, > in <module> petsc4py.init(sys.argv) File > "/home/user/.local/lib/python2.7/site-packages/petsc4py/__init__.py", line > 42, in init PETSc = petsc4py.lib.ImportPETSc(arch) File > "/home/user/.local/lib/python2.7/site-packages/petsc4py/lib/__init__.py", > line 29, in ImportPETSc return Import('petsc4py', 'PETSc', path, arch) > File > "/home/user/.local/lib/python2.7/site-packages/petsc4py/lib/__init__.py", > line 64, in Import module = imp.load_module(fullname, fo, fn, stuff) > ImportError: > /home/user/.local/lib/python2.7/site-packages/petsc/lib/libpetsc.so.3.06: > undefined symbol: kspfgmresmodifypcksp_ *Is there anything I can do about > it? Thank you! > > Henrik > > > -- > Dipl.-Math. Henrik Büsing > Applied Geophysics and Geothermal Energy > E.ON Energy Research Center > RWTH Aachen University > ----------------------------------------------- > Mathieustr. 10 | Tel +49 (0)241 80 49907 > 52074 Aachen, Germany | Fax +49 (0)241 80 49889 > ----------------------------------------------- > http://www.eonerc.rwth-aachen.de/GGE > hbuesing(a)eonerc.rwth-aachen.de > ----------------------------------------------- >

2 1

Re: [firedrake] NonlinearVariationalSolver problem
by Colin Cotter 15 Sep '15

15 Sep '15

Oh no! It's the canonical split() confusion. cheers --cjc On 15 September 2015 at 14:20, Mitchell, Lawrence < lawrence.mitchell(a)imperial.ac.uk> wrote: > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > On 15/09/15 14:18, Shipton, Jemma wrote: > > v,rho = w1.split() > > v, rho = split(w1) > > > w,phi = TestFunctions(WV) > > > Lawrence > -----BEGIN PGP SIGNATURE----- > Version: GnuPG v1 > > iQEcBAEBAgAGBQJV+BskAAoJECOc1kQ8PEYvgrgIAIEFwAQL8PPqWM5rtZwG3ugB > KpgBOiBFuf048QobPXlQBOranGOYplRP11HP0bPdB8pxVPrE2Bl3r/98gnHeJsAL > LnqRNxAxksuecaCfWR5kJSq/IiSWh5R4LcKwOv3Qi7qdk7l022Fa9eTHGBnOuIw/ > nKkkoj0nqV4pW3JnZBPqeJm6DdDWrrJ9OO6M8g6DNmqGannJl/Qjxcdwo9XIiSXd > vi42HfNICH7ffRT4LuIeq8V9kZRrCGyeBUWogRQ9EqljwjBOg4Vky9xtKnMNsP9G > wmfykMnL6EEhy0Kln7H3xu0jenoiF9xc3qESQ5y8C+QLelbmQtZr4bB8C3sJqmw= > =aH7E > -----END PGP SIGNATURE----- > > _______________________________________________ > firedrake mailing list > firedrake(a)imperial.ac.uk > https://mailman.ic.ac.uk/mailman/listinfo/firedrake > -- http://www.imperial.ac.uk/people/colin.cotter www.cambridge.org/9781107663916

1 0