Skip to content

FORTRAN program that implements hybrid parallel computing (OpenMP and MPI) to solve the general equilibrium solution of a heterogeneous agent model that simulates the dynamics of U.S. wealth distribution.

Notifications You must be signed in to change notification settings

libinruan/dsge_openmp_mpi_parallel_computing

Repository files navigation

!! Hybrid MPI with OpenMP version !! The implementation of the parallel version Nelder-Mead algorithm in global optimization.

Fortran files:

(1) main.f90 focuses on the implementation of the Nelder-Mead simplex method on 10 dimensional parameter space. (MPI)
(2) equilibrium.f90 solves for the steady state of the economy.
(3) model.f90 solves the Bellman equations for saving and consumption decisions across generations. (OpenMP)
(4) variable.f90 initializes most variables to be used in the program.
(5) toolbox.f90 contains all the subroutines that executes evoked numerical methods.

Input files:

*Mode == 0-2
(1) _survival_probability.txt
(2) _efficiency_units.txt
(3) _1parameter.txt

*Mode == 2 
(4) _ShortListStartPoint.csv: generated by "sorting_on_funcval.py" serving as the short list of starting points of the Nelder-Mead algorithm. "sorting_on_funcval.py" merges and sorts the outcome obtained in the MPI_Exercise_Mode==1 case.
(5) _startingpoint_1.csv: specified in _1parameter.txt to tell the program which starting point from the short list to use. 
(6) !!parameter "nrow" has to be the same as the size of a single amoeba. For example, in the 10 dimensional case, we need at least 11 vertices. To form 3 subgroup with a single amoeba, we need 3*4 (12) vertices. With the addition of the system root, the total nodes required should be 12+1 (13).

Supplemental files:

(1) zv1.pbs: the PBS script with the specification of HPC resources to be submitted to MSI's work queue.
(2) makefile: the script with the commands for compiling the Fortran files 

About

FORTRAN program that implements hybrid parallel computing (OpenMP and MPI) to solve the general equilibrium solution of a heterogeneous agent model that simulates the dynamics of U.S. wealth distribution.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages