accelerating Dynare, up to 10x, with SilverBullet

Hello everybody
First of all, thanks a lot to the creators of Dynare: it is a truly amazing tool
Then, my contribution: I think I've found a pretty straightforward way to accelerate the estimation of big DSGE models using Dynare
It reduces the use of global variables, just like dynare 4 is trying to do, but through a much simpler approach (it shouldn't take more than a few hours to add this trick to dynare 3)
For a model with 84 equations and 102 global parameters, it is reducing estimation time by a factor of nearly 10 (from 22.3 hours to 2.3 hours)
See the pdf for an explanation of the process, and the xls for a macro that does the trick automatically
(you need to be logged in to see these)
First of all, thanks a lot to the creators of Dynare: it is a truly amazing tool
Then, my contribution: I think I've found a pretty straightforward way to accelerate the estimation of big DSGE models using Dynare
It reduces the use of global variables, just like dynare 4 is trying to do, but through a much simpler approach (it shouldn't take more than a few hours to add this trick to dynare 3)
For a model with 84 equations and 102 global parameters, it is reducing estimation time by a factor of nearly 10 (from 22.3 hours to 2.3 hours)
See the pdf for an explanation of the process, and the xls for a macro that does the trick automatically
(you need to be logged in to see these)