[firedrake] Issues with LMA of vertical derivative on extruded meshes

[Eike Mueller]

Hi Lawrence,

thanks, I adapted my code to make sure it only disables the COFFEE optimisation in the LMA assembly (I also had to change it in one or two other places). At the moment the banded matrix construction is part of the setup phase, so I don’t really care if it is a bit slower.

I just subscribed to the github issue, so will look out for updates.

Cheers,

Eike

--

Dr Eike Hermann Mueller
Research Associate (PostDoc)

Department of Mathematical Sciences
University of Bath
Bath BA2 7AY, United Kingdom

+44 1225 38 5803
e.mueller at bath.ac.uk
http://people.bath.ac.uk/em459/

> On 6 Jan 2015, at 10:47, Lawrence Mitchell <lawrence.mitchell at imperial.ac.uk> wrote:
> 
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> On 06/01/15 08:52, Eike Mueller wrote:
>> Hi Lawrence,
>> 
>> thanks for the explanation, so my understanding is that in general
>> PyOP2 is not aware of the COFFEE optimisation, and the wrapper code
>> that is generated in my PyOP2 parloop can’t handle the padding.
> 
> Yes, that's right.  You can follow this issue
> https://github.com/coneoproject/COFFEE/issues/14 to see if anything
> will change.  I hope so, since it's not a very nice model.
> 
>> Since I call a PyOP2 parloop directly, I don’t have any control
>> over the wrapper code and at the moment it sounds like the only
>> thing I can do is switch off the COFFEE optimisation (which, I
>> suppose, has implications everwhere in the code, so I really
>> shouldn’t do this).
> 
> You can do the following in build_lma:
> 
> def build_lma(...):
>    parameters["coffee"]["O2"] = False
>    existing_code
>    parameters["coffee"]["O2"] = True
> 
> 
> That way the optimisations are only switched off for that one section.
> 
>> Isn’t then there a problem with all PyOP2 parloops, and are you
>> going to do something about this?
> 
> So the problem is that you're doing something that isn't explicitly
> supported by the PyOP2 data model: using a Kernel designed for
> assembling into a Mat to assemble into a Dat.  It happens to work in
> some cases.
> 
> I think it is probably useful to properly support a "build_lma_matrix"
> style abstraction, but I haven't thought properly about how to do it
> right.
> 
> Lawrence
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.10 (GNU/Linux)
> Comment: Using GnuPG with undefined - http://www.enigmail.net/
> 
> iQEcBAEBAgAGBQJUq68hAAoJECOc1kQ8PEYvxYsH/RyIE4fr3WKapkstsmWckEZX
> MzGN+C2Zl+oYp4VYDDqb8WtjNoVDl0CzqyAeywLDIUOCktFYD/HFLIcE8q3eSVIA
> sAAo/Lt6g0rqBaR6cFyqLWSC4DIcAJAYdEKFfHsOxRxJY1sSzOWMhdl6i35FjMpg
> nTZtUMWZd0gXeLXr44FW2aaHbGlJzgjFrlrLx21kXpt1B+ZM5jl2kDJ9N7gU+l8b
> GUQsDh/jEuOYKP9lKmLgGktIHTEBh7BTPK2jv36iSqZ2jJ7OExAszAympGPkyTd1
> bYcaFXwKQ8IWjbvYfNdQLvOYgbxtGoQuFg5V34q4WOkJ7yb+QRdlyl0Jj/Som4U=
> =CFSf
> -----END PGP SIGNATURE-----
> 
> _______________________________________________
> firedrake mailing list
> firedrake at imperial.ac.uk
> https://mailman.ic.ac.uk/mailman/listinfo/firedrake

-------------- next part --------------
HTML attachment scrubbed and removed