Extended path & bytecode

Extended Path
1

Hello Dynare folks,

Dynare 4.6.2 has been fixed with the bug for extended path simulations with the bytecode option.
However when I run an extended path simulation with the bytecode option, I get the following error:

Unable to resolve the name M_.params.

Error in extended_path_core (line 44)
    [flag, tmp] = bytecode('dynamic', endo_simul, exo_simul, M_.params,
    endo_simul, periods);

To solve the issue, line 44 of ep/extended_path_core, change ‘M_.params’ with ‘M.params’ when calling bytecode() function.

Kindly,
gauthier

2

Thanks for noticing and reporting this. This will be fixed in the next release.

Best,

3

Thank you Sebastien for your very quick reply.

I just figured out that the code feature two other issues. Running the code in its actual form (simple version of extended path with order=0 + lmmcp=1) provides the following error message:

Unable to perform assignment because dot indexing is not supported for
variables of this type.

Error in extended_path_core (line 82)
    [info_convergence, tmp] =
    extended_path_homotopy(endo_simul, exo_simul, M, options, oo, pfm,
    ep, order, algo, 2, debug);

tmp has been wrongly set line 44: ‘tmp’ is set as a matrix and is next called as a structure line 82.
I am not sure that ‘tmp’ needs to be set line 44 as it will stores the residual of the dynamic equations. So i would simply write line 44:
[info_convergence] = bytecode(…

With this typo solved, another error pops up:

Unrecognized function or variable 'info_convergence'.

Error in extended_path_core (line 80)
if ~info_convergence && ~options.no_homotopy

if the bytecode function returns flag=0 line 44, then ‘info_convergence’ is not set, so the code line 80 cannot work.

Kindly,
gauthier

4

Thanks for noticing this other problem.

The bug will be fixed in the next Dynare release. If you are impatient, you can try the changes contained in this commit: https://git.dynare.org/Dynare/dynare/-/commit/e012db29f5369dfe4c70656a969b16bdcb44e0cc

Best

1 Like
5

Thank you Sebastien it works perfectly!

6

Hello @sebastien

Just a quick question related to bytecode option with EP: is it possible to disable the messages when running an EP simulation? The message looks like:

MODEL SIMULATION: (method=Sparse LU)

  Simulate iteration no 1     
  max. error=2.4008185606e-02       
  sqr. error=7.4504184070e-04       
  abs. error=4.2973750031e-02

I would like the code to run silently to only track the final output, not the intermediary one.

Kindly,
gauthier

7

Hi Gauthier,

I worked on this issue (removing useless outputs from bytecode solvers) years ago and I am not sure I finished… Since I decided not to use bytecode anymore (for extended path simulations). What are the values of options_.ep.debug and options_.ep.verbosity in your case?

Best,
Stéphane.

8

Hello Stéphane,

Thanks for your quick feedback.

Both options_.ep.debug and options_.ep.verbosity are set to zero, but thanks to your message I just figured out that’s actually options_.verbosity that controls the messaging from extended path. It works like a charm now.

Thanks again,
gauthier