On Mon, Jul 30, 2012 at 9:01 PM, Roy Stogner <royst...@ices.utexas.edu>wrote:
>
>
> On Mon, 30 Jul 2012, Paul T. Bauman wrote:
>
> 3. I had to remove the calculate_dphi = true assignment in the
>> get_dphidxi, etc. calls, but I've at least left the asserts there.
>> The problem is that we may need dphidxi without actually needing
>> dphi. The only time this will break is if the user *only* calls
>> get_phi, but also calls get_dphidxi. Can anyone think of a case
>> where this would happen (I couldn't come up with one)?
>>
>
> Sure: any application code which only uses the component arrays
> without using the gradient array will now be failing that assertion in
> debug mode or probably accessing OOB memory in opt mode.
>
I don't follow: if the user calls get_dphidx, dphi and all the components
will be calculated and everything is fine (calculate_dphi gets set to true
in all those methods). This case will only break if get_dphi, get_dphidx,
etc. are *not* called, but the user wants get_dphieta (*reference* shape
function derivatives, changed to eta instead of xi for clarity). So if we
really want to support this case, how do you want to support it? I need
those functions to be public for the current implementation of the
FETransformation object (I pass a const FEGenericBase& to get access to the
shape function derivatives which were precomputed in init_shape_functions).
Do we remove those methods and move the reference shape function derivative
calculations out of init_shape_functions?
> Why can't we set "calculate_dphi = true" there? I'd rather take the
> component arrays out entirely than leave them in there booby-trapped.
>
1. Performance - The reference shape function derivatives are needed for
calculating the divergence and curl, as well as the gradient, but I may not
need the gradients. The curl-curl model problem is an example of this.
Thus, as it was, calling get_dphidxi in the FETransformation object would
force dphi to be calculated when I didn't want it to be.
2. Laziness/time - I don't want to deal with the HCurl shape function
gradient right now and with the assignment, my curl-curl example would fail
with a libmesh_not_implemented() for the dphi calculation.
------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and
threat landscape has changed and how IT managers can respond. Discussions
will include endpoint security, mobile security and the latest in malware
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
Libmesh-devel mailing list
Libmesh-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/libmesh-devel
