masterParallel.m and fInputVar

[MaxB]

Dear all,

I have been trying to get Parallel Dynare to work on my 4-core personal computer for a few days, and I am encountering a difficulty that is far behind my skills. Once Dynare (4.3.1) has estimated the posterior mean of my parameters, and so before running the MH algorithms, I get the following error message:

Code: Select all

error: save: error while writing `' to MAT file
error: save: error while writing `fInputVar' to MAT file
error: called from:
error:   C:\dynare\4.3.1\matlab\/parallel\masterParallel.m at line 130, column 9
error:   C:\dynare\4.3.1\matlab\random_walk_metropolis_hastings.m at line 148, column 35
error:   C:\dynare\4.3.1\matlab\dynare_estimation_1.m at line 884, column 13
error:   C:\dynare\4.3.1\matlab\dynare_estimation.m at line 70, column 5
error:   C:\dynare\4.3.1\memoire2\miuf_estim.m at line 156, column 1
error:   C:\dynare\4.3.1\matlab\dynare.m at line 120, column 1


It sounds to me like there is somekind of a bad relationship between my CPUs, but I am puzzled. I have tried to understand going through the ParallelMaster.m file, but it goes far beyond my MatLab understanding, and I can not figure out if it really is a network issue, or if it may be more complicated than that.

Please note that:

• I am using Dynare 4.3.1 on Octave 3.6.1, and I use Windows 7. I am the only user, so I have admin rights, and should not be stopped by writing rights, should I?
• I have installed PsTools, and psexec (using the console) works just fine. psinfo has some trouble about the "uptime line", which says that is an error reading uptime. Could my problem come from here? I read on the internet that this uptime issue happens quite a lot, but I can't say if this has to do with my issue here.
• Also, the very same dynare command with parallel_test at the end works fine.

I join to this message my config file, although that it is very standard.

Thank you very much for your time,

[rattoma]

Hi,

at the point where the error happens the parallel threads did not start yet, so there should be nothing wrong there. It seems a bug for octave. Can you pass me the dynare project you are trying to run?
thanks a lot
Marco

[MaxB]

Hello!

Sure! Here it is, joined to this email. Do not take into account the notes in the comments of the .mod file, it is just a small model I am using to get dynare estimations running. Is works without using the parallel option. I was concerned that it was an Octave-related bug, I don't know how much the developpers are interested in them. (There are a few strange things that happen when running Dynare using Octave, but nothing that would prevent me to work, before this one.)

Thank you very much for your time!

Yours,

Max

EDIT : it sounds like it's actually not working, even without going through the standard procedure. I am going to have a look at the .mod file, sorry for that!

[rattoma]

Indeed I noted the error of the dynare project irrespective to the parallel run.
Please let me know if the parallel problem persists or it is fixed.