fiat team mailing list archive

[Robert Kirby <robert.c.kirby@xxxxxxxxx>]

I know that ffc requires the dmats function, and I don't want to break it.

However, dmats is one of the most opaque and poorly understood aspects of
FIAT for new folks reading
the code.

With automatic differentiation, I think I can make the whole code cleaner
by getting rid of dmats internally and providing a method that will compute
it on demand.

The tabulate method will then just use AD to tabulate the jet of the basis
functions and do a mat-vec of
coefficients.

Any comments?  I ask since I'm going to try to put constrained spaces back
in (feature request for BDFM elements by David Ham)
and might like to clean things up a bit.

Rob


On Mon, Mar 21, 2011 at 2:31 PM, Robert Kirby <robert.c.kirby@xxxxxxxxx>wrote:

> OK.  I'll try to work on this.
>
>
>
> On Mon, Mar 21, 2011 at 1:55 PM, Anders Logg <logg@xxxxxxxxx> wrote:
>
>> On Mon, Mar 21, 2011 at 01:41:39PM -0500, Robert Kirby wrote:
>> > Anders, am I correct in thinking that ffc is currently based on the main
>> fiat
>> > branch lp:fiat?
>>
>> Yes, and it is identical to the 0.9.9 release I made a few weeks ago,
>> except for a very small fix (adding a missing date to the ChangeLog).
>>
>> > David Ham is asking for BDFM, and that will involve putting constrained
>> element
>> > spaces back in.
>> > I want to make sure I do it in the right spot.
>>
>> It will be good to get BDFM back into FIAT.
>>
>> --
>> Anders
>>
>
>