Dear all, I would like to follow up on the discussion at the Dynare conference on computation time at higher order. As far as I understand it, we have a performance problem with estimation at higher order. Usually the Dynare preprocessor computes the symbolic derivatives once and then we access them to compute the solution. But when using k_order_pert.dll, the Dynare++ engine will preprocess the model in each call for each parameter set. Is this interpretation correct? If yes, this looks like a bottleneck we should try to avoid. Best, Johannes -- Johannes Pfeifer Friesenwall 104 50672 Köln Tel.: +49-221-29873852 Mobil.: +49-170-6936820 Germany
Hi Johannes, this is not correct. k_order_pert.dll execute <modfile>_dynamic to obtain the 1,2,3rd derivatives of the the model. Best Michel Johannes Pfeifer writes:
Dear all,
I would like to follow up on the discussion at the Dynare conference on computation time at higher order. As far as I understand it, we have a performance problem with estimation at higher order. Usually the Dynare preprocessor computes the symbolic derivatives once and then we access them to compute the solution. But when using k_order_pert.dll, the Dynare++ engine will preprocess the model in each call for each parameter set. Is this interpretation correct?
If yes, this looks like a bottleneck we should try to avoid.
Best,
Johannes
-- Michel Juillard
participants (2)
-
Johannes Pfeifer -
Michel Juillard