stepan
/
dynare
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

1) eliminated us of struct2local; 

2) changes to header/comments.
time-shift
Marco Ratto 2010-05-31 11:55:25 +02:00
parent 55764e667f
commit 839b77d679
2 changed files with 103 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

35
matlab/random_walk_metropolis_hastings.m
View File

@ -19,6 +19,22 @@ function random_walk_metropolis_hastings(TargetFun,ProposalFun,xparam1,vv,mh_bou
% SPECIAL REQUIREMENTS
% None.
% PARALLEL CONTEXT
% The most computationally intensive part of this function may be executed
% in parallel. The code sutable to be executed in
% parallel on multi core or cluster machine (in general a 'for' cycle),
% is removed from this function and placed in random_walk_metropolis_hastings_core.m funtion.
% Then the DYNARE parallel package contain a set of pairs matlab functions that can be executed in
% parallel and called name_function.m and name_function_core.m.
% In addition in parallel package we have second set of functions used
% to manage the parallel computation.
%
% This function was the first function to be parallelized, later other
% functions have been parallelized using the same methodology.
% Then the comments write here can be used for all the other pairs of
% parallel functions and also for management funtions.
% Copyright (C) 2006-2008,2010 Dynare Team
%
% This file is part of Dynare.
@ -47,6 +63,10 @@ InitSizeArray = min([repmat(MAX_nruns,nblck,1) fline+nruns-1],[],2);
load([MhDirectoryName '/' ModelName '_mh_history.mat'],'record');
% Snapshot of the current state of computing. It necessary for the parallel
% execution (i.e. to execute in a corretct way portion of code remotely or
% on many core). The mandatory variables for local/remote parallel
% computing are stored in localVars struct.
localVars = struct('TargetFun', TargetFun, ...
'ProposalFun', ProposalFun, ...
@ -66,15 +86,18 @@ localVars.InitSizeArray=InitSizeArray;
localVars.record=record;
localVars.varargin=varargin;
% tic,
% The user don't want to use parallel computing, or want to compute a
% single chain. In this cases Random walk Metropolis-Hastings algorithm is
% computed sequentially.
if isnumeric(options_.parallel) || (nblck-fblck)==0,
fout = random_walk_metropolis_hastings_core(localVars, fblck, nblck, 0);
record = fout.record;
else
% global variables for parallel routines
% Parallel in Local or remote machine.
else
% Global variables for parallel routines.
globalVars = struct('M_',M_, ...
'options_', options_, ...
'bayestopt_', bayestopt_, ...
@ -109,8 +132,6 @@ irun = fout(1).irun;
NewFile = fout(1).NewFile;
% ComptationalTime=toc,
% record.Seeds.Normal = randn('state');
% record.Seeds.Unifor = rand('state');
save([MhDirectoryName '/' ModelName '_mh_history.mat'],'record');
@ -119,4 +140,4 @@ disp(['MH: Total number of generated files : ' int2str(NewFile(1)*nblck) '.
disp(['MH: Total number of iterations : ' int2str((NewFile(1)-1)*MAX_nruns+irun-1) '.'])
disp('MH: average acceptation rate per chain : ')
disp(record.AcceptationRates);
disp(' ')
disp(' ')

77
matlab/random_walk_metropolis_hastings_core.m
View File

@ -1,6 +1,53 @@
function myoutput = random_walk_metropolis_hastings_core(myinputs,fblck,nblck,whoiam, ThisMatlab)
% PARALLEL CONTEXT
% This function contain the most computationally intensive portion of code in
% random_walk_metropolis_hastings (the 'for xxx = fblck:nblck' loop). The branches in 'for'
% cycle and are completely independent than suitable to be executed in parallel way.
%
% INPUTS
% o myimput [struc] The mandatory variables for local/remote
% parallel computing obtained from random_walk_metropolis_hastings.m
% function.
% o fblck and nblck [integer] The Metropolis-Hastings chains.
% o whoiam [integer] In concurrent programming a modality to refer to the differents thread running in parallel is needed.
% The integer whoaim is the integer that
% allows us to distinguish between them. Then it is the index number of this CPU among all CPUs in the
% cluster.
% o ThisMatlab [integer] Allows us to distinguish between the
% 'main' matlab, the slave matlab worker, local matlab, remote matlab,
% ... Then it is the index number of this slave machine in the cluster.
% OUTPUTS
% o myoutput [struc]
% If executed without parallel is the original output of 'for b =
% fblck:nblck' otherwise a portion of it computed on a specific core or
% remote machine. In this case:
% record;
% irun;
% NewFile;
% OutputFileName
%
% ALGORITHM
% Portion of Metropolis-Hastings.
%
% SPECIAL REQUIREMENTS.
% None.
% Copyright (C) 2006-2008 Dynare Team
% PARALLEL CONTEXT
% The most computationally intensive part of this function may be executed
% in parallel. The code sutable to be executed in parallel on multi core or cluster machine,
% is removed from this function and placed in random_walk_metropolis_hastings_core.m funtion.
% Then the DYNARE parallel package contain a set of pairs matlab functios that can be executed in
% parallel and called name_function.m and name_function_core.m.
% In addition in the parallel package we have second set of functions used
% to manage the parallel computation.
%
% This function was the first function to be parallelized, later other
% functions have been parallelized using the same methodology.
% Then the comments write here can be used for all the other pairs of
% parallel functions and also for management funtions.
% Copyright (C) 2006-2008,2010 Dynare Team
%
% This file is part of Dynare.
%
@ -24,7 +71,33 @@ end
global bayestopt_ estim_params_ options_ M_ oo_
struct2local(myinputs);
% reshape 'myinputs' for local computation.
% In order to avoid confusion in the name space, the instruction struct2local(myinputs) is replaced by:
TargetFun=myinputs.TargetFun;
ProposalFun=myinputs.ProposalFun;
xparam1=myinputs.xparam1;
vv=myinputs.vv;
mh_bounds=myinputs.mh_bounds;
ix2=myinputs.ix2;
ilogpo2=myinputs.ilogpo2;
ModelName=myinputs.ModelName;
fline=myinputs.fline;
npar=myinputs.npar;
nruns=myinputs.nruns;
NewFile=myinputs.NewFile;
MAX_nruns=myinputs.MAX_nruns;
d=myinputs.d;
InitSizeArray=myinputs.InitSizeArray;
record=myinputs.record;
varargin=myinputs.varargin;
% Necessary only for remote computing!
if whoiam
Parallel=myinputs.Parallel;
MasterName=myinputs.MasterName;
DyMo=myinputs.DyMo;
end
% (re)Set the penalty
bayestopt_.penalty = Inf;