Commit 5c893f50 authored by Michel Juillard's avatar Michel Juillard
Browse files

making extended path ready for parallel computing with parfor

parent 020328a8
......@@ -80,18 +80,22 @@ function [flag,endo_simul,err] = solve_stochastic_perfect_foresight_model(endo_s
% each block of Y with ny rows are unfolded column wise
dimension = ny*(sum(nnodes.^(0:order-1),2)+(periods-order)*world_nbr);
if order == 0
i_upd = ny+(1:ny*periods);
i_upd_r = (1:ny*periods);
i_upd_y = i_upd_r + ny;
else
i_upd = zeros(dimension,1);
i_upd(1:ny) = ny+(1:ny);
i_upd_r = zeros(dimension,1);
i_upd_y = i_upd_r;
i_upd_r(1:ny) = (1:ny);
i_upd_y(1:ny) = ny+(1:ny);
i1 = ny+1;
i2 = 2*ny;
n1 = 2*ny+1;
n2 = 3*ny;
n1 = ny+1;
n2 = 2*ny;
for i=2:periods
k = n1:n2;
for j=1:nnodes^min(i-1,order)
i_upd(i1:i2) = (n1:n2)+(j-1)*ny*(periods+2);
i_upd_r(i1:i2) = (n1:n2)+(j-1)*ny*periods;
i_upd_y(i1:i2) = (n1:n2)+ny+(j-1)*ny*(periods+2);
i1 = i2+1;
i2 = i2+ny;
end
......@@ -99,12 +103,13 @@ function [flag,endo_simul,err] = solve_stochastic_perfect_foresight_model(endo_s
n2 = n2+ny;
end
end
icA = [find(lead_lag_incidence(1,:)) find(lead_lag_incidence(2,:))+world_nbr*ny ...
find(lead_lag_incidence(3,:))+2*world_nbr*ny]';
h1 = clock;
for iter = 1:maxit
h2 = clock;
A = sparse([],[],[],dimension,dimension,(periods+2)*world_nbr*nnz(jacobian));
res = zeros(dimension,1);
A1 = sparse([],[],[],ny*(sum(nnodes.^(0:order-1),2)+1),dimension,(order+1)*world_nbr*nnz(jacobian));
res = zeros(ny,periods,world_nbr);
i_rows = 1:ny;
i_cols = find(lead_lag_incidence');
i_cols_p = i_cols(1:nyp);
......@@ -114,91 +119,85 @@ function [flag,endo_simul,err] = solve_stochastic_perfect_foresight_model(endo_s
i_cols_Ap = i_cols_p;
i_cols_As = i_cols_s;
i_cols_Af = i_cols_f - ny;
for i = 1:periods
if i <= order+1
i_w_p = 1;
for j = 1:nnodes^(i-1)
innovation = exo_simul;
if i > 1
innovation(i+1,:) = nodes(mod(j-1,nnodes)+1,:);
end
if i <= order
for k=1:nnodes
y = [Y(i_cols_p,i_w_p);
Y(i_cols_s,j);
Y(i_cols_f,(j-1)*nnodes+k)];
[d1,jacobian] = dynamic_model(y,innovation,params,steady_state,i+1);
if i == 1
% in first period we don't keep track of
% predetermined variables
i_cols_A = [i_cols_As - ny; i_cols_Af];
A(i_rows,i_cols_A) = A(i_rows,i_cols_A) + weights(k)*jacobian(:,i_cols_1);
else
i_cols_A = [i_cols_Ap; i_cols_As; i_cols_Af];
A(i_rows,i_cols_A) = A(i_rows,i_cols_A) + weights(k)*jacobian(:,i_cols_j);
end
res(i_rows) = res(i_rows)+weights(k)*d1;
i_cols_Af = i_cols_Af + ny;
end
else
for i = 1:order+1
i_w_p = 1;
for j = 1:nnodes^(i-1)
innovation = exo_simul;
if i > 1
innovation(i+1,:) = nodes(mod(j-1,nnodes)+1,:);
end
if i <= order
for k=1:nnodes
y = [Y(i_cols_p,i_w_p);
Y(i_cols_s,j);
Y(i_cols_f,j)];
Y(i_cols_f,(j-1)*nnodes+k)];
[d1,jacobian] = dynamic_model(y,innovation,params,steady_state,i+1);
if i == 1
% in first period we don't keep track of
% predetermined variables
i_cols_A = [i_cols_As - ny; i_cols_Af];
A(i_rows,i_cols_A) = jacobian(:,i_cols_1);
A1(i_rows,i_cols_A) = A1(i_rows,i_cols_A) + weights(k)*jacobian(:,i_cols_1);
else
i_cols_A = [i_cols_Ap; i_cols_As; i_cols_Af];
A(i_rows,i_cols_A) = jacobian(:,i_cols_j);
A1(i_rows,i_cols_A) = A1(i_rows,i_cols_A) + weights(k)*jacobian(:,i_cols_j);
end
res(i_rows) = d1;
res(:,i,j) = res(:,j,i)+weights(k)*d1;
i_cols_Af = i_cols_Af + ny;
end
i_rows = i_rows + ny;
if mod(j,nnodes) == 0
i_w_p = i_w_p + 1;
end
if i > 1
if mod(j,nnodes) == 0
i_cols_Ap = i_cols_Ap + ny;
end
i_cols_As = i_cols_As + ny;
else
y = [Y(i_cols_p,i_w_p);
Y(i_cols_s,j);
Y(i_cols_f,j)];
[d1,jacobian] = dynamic_model(y,innovation,params,steady_state,i+1);
if i == 1
% in first period we don't keep track of
% predetermined variables
i_cols_A = [i_cols_As - ny; i_cols_Af];
A1(i_rows,i_cols_A) = jacobian(:,i_cols_1);
else
i_cols_A = [i_cols_Ap; i_cols_As; i_cols_Af];
A1(i_rows,i_cols_A) = jacobian(:,i_cols_j);
end
res(:,i,j) = d1;
i_cols_Af = i_cols_Af + ny;
end
i_cols_p = i_cols_p + ny;
i_cols_s = i_cols_s + ny;
i_cols_f = i_cols_f + ny;
elseif i == periods
if i == order+2
i_cols_A = [i_cols_Ap; i_cols_As; i_cols_Af];
i_rows = i_rows + ny;
if mod(j,nnodes) == 0
i_w_p = i_w_p + 1;
end
for j=1:world_nbr
[d1,jacobian] = dynamic_model(Y(i_cols,j),exo_simul, ...
params,steady_state,i+1);
A(i_rows,i_cols_A(i_cols_T)) = jacobian(:,i_cols_T);
res(i_rows) = d1;
i_rows = i_rows + ny;
i_cols_A = i_cols_A + ny;
end
else
if i == order+2
i_cols_A = [i_cols_Ap; i_cols_As; i_cols_Af];
if i > 1
if mod(j,nnodes) == 0
i_cols_Ap = i_cols_Ap + ny;
end
i_cols_As = i_cols_As + ny;
end
for j=1:world_nbr
[d1,jacobian] = dynamic_model(Y(i_cols,j), ...
exo_simul,params,steady_state,i+1);
A(i_rows,i_cols_A) = jacobian(:,i_cols_j);
res(i_rows) = d1;
i_rows = i_rows + ny;
i_cols_A = i_cols_A + ny;
end
i_cols_p = i_cols_p + ny;
i_cols_s = i_cols_s + ny;
i_cols_f = i_cols_f + ny;
end
nzA = cell(periods,world_nbr);
parfor j=1:world_nbr
i_rows_y = find(lead_lag_incidence')+(order+1)*ny;
offset_c = ny*(sum(nnodes.^(0:order-1),2)+j-1);
offset_r = (j-1)*ny;
for i=order+2:periods
[d1,jacobian] = dynamic_model(Y(i_rows_y,j), ...
exo_simul,params, ...
steady_state,i+1);
if i == periods
[ir,ic,v] = find(jacobian(:,i_cols_T));
else
[ir,ic,v] = find(jacobian(:,i_cols_j));
end
nzA{i,j} = [offset_r+ir,offset_c+icA(ic), v]';
res(:,i,j) = d1;
i_rows_y = i_rows_y + ny;
offset_c = offset_c + world_nbr*ny;
offset_r = offset_r + world_nbr*ny;
end
i_cols = i_cols + ny;
end
err = max(abs(res));
err = max(abs(res(i_upd_r)));
if err < tolerance
stop = 1;
if verbose
......@@ -214,8 +213,10 @@ function [flag,endo_simul,err] = solve_stochastic_perfect_foresight_model(endo_s
% pause
break
end
dy = -A\res;
Y(i_upd) = Y(i_upd) + dy;
A2 = [nzA{:}]';
A = [A1; sparse(A2(:,1),A2(:,2),A2(:,3),ny*(periods-order-1)*world_nbr,dimension)];
dy = -A\res(i_upd_r);
Y(i_upd_y) = Y(i_upd_y) + dy;
end
if ~stop
......
......@@ -79,10 +79,10 @@ copyfile('rbcii_steady_state.m','rbcii_steadystate2.m');
ts = extended_path([],100);
options_.ep.stochastic.order = 1;
profile on
// profile on
ts1_4 = extended_path([],100);
profile off
profile viewer
// profile off
// profile viewer
@#else
shocks;
......
Supports Markdown
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment