ffc team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Fri, Sep 02, 2005 at 09:43:16AM +0200, jhoffman@xxxxxxxxxxx wrote:
> Anders,
> 
> Some operations for Functions in FFC that would be very useful are:
> multiplication, division, and powers, including taking the square root,
> with the operations defined as the corresponding operations for the nodal
> values.

What do you mean by multiplication? FFC already does
multiplication. Do you mean interpolating a product w1*w2 to remove
one of the factors in the output to reduce the complexity of evaluation?

> Together with access to a local (cell) h in some form, I think we will be
> able to implement a wide range of solvers in Dolfin.
> 
> /Johan

I agree. All this will be added.

There are a couple of things I need to implement in FFC. Which is the
most important? Which should be implemented first?

1. Improve the expressiveness of the form language: division, powers,
square root, predefined h, index ranges (for Garth), operators such as
grad(), rot(), div(), unit tensor etc.

2. Boundary integrals using *ds.

3. Option -f blas for generating BLAS output (to speed up compilation of
complicated forms).

I should be able to start implementing one of these early next week.

/Anders