Re: [Libmesh-devel] Elem::contains_point()

[John P. <jwp...@gm...>]

On Tue, Jun 19, 2012 at 1:14 PM, John Peterson <jwp...@gm...> wrote:
> On Tue, Jun 19, 2012 at 11:27 AM, Roy Stogner <roy...@ic...> wrote:
>>
>> Not sure what the best solution is.  Add a sanity check by calculating
>> xyz for the results of the inverse_map and making sure the distance
>> from the original point is under tol*h_max?  Seems unnecessarily
>> expensive in 90% of case, but I suppose we'd only need to do the check
>> in 1D and 2D cases.
>
> The attached patch implements this fix, it works for my simple test
> case and a more involved one I have...
>
> It could be made a bit faster by adding an FEInterface::map() function
> in the same vein as FEInterface::inverse_map() which forwards to
> FE<Dim,T>::map().  This seems like it would also be a useful feature
> in general...

OK, FEInterface::map() is added and Elem::point_test() uses it.

Considering that Elem::point_test() already contains a call to
FEInterface::inverse_map(), adding one more call to FE::map() makes it
only trivially more expensive...

-- 
John