ffc team mailing list archive

["Garth N. Wells" <gnw20@xxxxxxxxx>]

On 20 November 2011 19:09, Anders Logg <logg@xxxxxxxxx> wrote:
> On Sun, Nov 20, 2011 at 07:48:19PM +0100, Marie E. Rognes wrote:
>> On 11/20/2011 07:29 PM, Garth N. Wells wrote:
>> >I'm trying to track down a bug that leads to a silent failure. I'm as
>> >far as a call to the FFC function
>> >
>> >    ffc.tensor.estimate_cost
>> >
>> >which fails to return. This function is called from the function
>> >
>> >     ffc.analysis._auto_select_representation
>> >
>> >What I don't get is why it's even being called. Since the code that
>> >I'm running is:
>> >
>> >    from dolfin import *
>> >    mesh = UnitSquare(32, 32)
>> >    V = FunctionSpace(mesh, "Lagrange", 1)
>> >    print "------------- end"
>> >
>> >It's only a FunctionSpace, so why is FFC going through code that is
>> >for analysing forms?
>> >
>> >
>>
>>
>> Because of the use of a dummy form for generating function
>> spaces, see def jit_element(...) in ffc/jitcompiler.py.
>>
>> This has already resurfaced recently, cf
>>
>>     https://bugs.launchpad.net/dolfin/+bug/884670
>
> Not going via the form jit compiler would require some refactoring of
> the jit code in FFC. This would be good to do in the long run, but I
> don't think it's a good idea to do at the moment since the jit chain
> is complex, fragile and error prone. Changes would likely introduce
> bugs.
>
> Do you have an example of a form that hangs ffc.tensor.estimate_cost?
> It should be easy to debug.
>

I think that it likely a symptom rather than the cause. I think a
build configuration is the problem, and related to shared libs. See

  https://bugs.launchpad.net/dolfin/+bug/892714

Garth

Garth

> --
> Anders
>
> _______________________________________________
> Mailing list: https://launchpad.net/~ffc
> Post to     : ffc@xxxxxxxxxxxxxxxxxxx
> Unsubscribe : https://launchpad.net/~ffc
> More help   : https://help.launchpad.net/ListHelp
>