237 lines
8.4 KiB
Matlab
Executable File
237 lines
8.4 KiB
Matlab
Executable File
%%***************************************************************
|
|
%% linsysolve: solve linear system to get dy, and direction
|
|
%% corresponding to unrestricted variables.
|
|
%%
|
|
%% [xx,coeff,L,resnrm] = linsysolve(schur,UU,Afree,EE,rhs);
|
|
%%
|
|
%% child functions: symqmr.m, mybicgstable.m, linsysolvefun.m
|
|
%%
|
|
%% SDPT3: version 3.1
|
|
%% Copyright (c) 1997 by
|
|
%% K.C. Toh, M.J. Todd, R.H. Tutuncu
|
|
%% Last Modified: 16 Sep 2004
|
|
%%***************************************************************
|
|
|
|
function [xx,coeff,L,resnrm] = linsysolve(par,schur,UU,Afree,EE,rhs);
|
|
|
|
global solve_ok existspcholsymb exist_analytic_term
|
|
global nnzmat nnzmatold matfct_options matfct_options_old use_LU
|
|
global Lsymb msg diagR diagRold perturb_opt
|
|
|
|
spdensity = par.spdensity;
|
|
printlevel = par.printlevel;
|
|
iter = par.iter;
|
|
|
|
m = length(schur);
|
|
if (iter==1);
|
|
use_LU = 0; matfct_options_old = '';
|
|
diagR = ones(m,1); perturb_opt = 1;
|
|
end
|
|
if isempty(nnzmatold); nnzmatold = 0; end
|
|
diagRold = diagR;
|
|
%%
|
|
%% schur = schur + rho*diagschur + lam*AAt
|
|
%%
|
|
diagschur = full(abs(diag(schur)));
|
|
if (min(diagR) < 1e-6) & (max(diagR) > 1e-2);
|
|
perturb_opt = 2;
|
|
end
|
|
if (perturb_opt == 2) & (iter > 10)
|
|
pertdiag = min(1e-2,1e-15*max(1,diagschur)./max(1e-12,diagR));
|
|
idx = find(diagschur < 1e-6);
|
|
if (norm(rhs(idx)) < 1e-6) & (max(diagschur) > 1e-2)
|
|
pertdiag(idx) = 1e6*ones(length(idx),1);
|
|
end
|
|
mexschurfun(schur,pertdiag);
|
|
if (printlevel); fprintf('*'); end
|
|
else
|
|
diagschur = max(0,full(diag(schur)));
|
|
minrho(1) = max(1e-15, 1e-6/3.0^iter);
|
|
minlam(1) = max(1e-10, 1e-4/2.0^iter);
|
|
minrho(2) = max(1e-04, 1/1.5^iter);
|
|
rho = min(minrho(1),minrho(2)*(1+norm(rhs))/(1+norm(diagschur.*par.y)));
|
|
lam = min(minlam(1),0.1*rho*norm(diagschur)/par.normAAt);
|
|
if (exist_analytic_term); rho = 0; end
|
|
mexschurfun(schur,rho*diagschur);
|
|
mexschurfun(schur,lam*par.AAt);
|
|
%%if (printlevel); fprintf(' %2.1e %2.1e ',rho,lam); end
|
|
end
|
|
%%
|
|
%% assemble coefficient matrix
|
|
%%
|
|
len = size(Afree,2);
|
|
if ~isempty(EE)
|
|
EE(:,[1 2]) = len + EE(:,[1 2]); %% adjust for ublk
|
|
end
|
|
EE = [(1:len)' (1:len)' zeros(len,1); EE];
|
|
if isempty(EE)
|
|
coeff.mat22 = [];
|
|
else
|
|
coeff.mat22 = spconvert(EE);
|
|
end
|
|
coeff.mat12 = [Afree, UU];
|
|
coeff.mat11 = schur; %% important to use perturbed schur matrix
|
|
ncolU = size(coeff.mat12,2);
|
|
%%
|
|
%% pad rhs with zero vector
|
|
%% decide which solution methods to use
|
|
%%
|
|
rhs = [rhs; zeros(m+ncolU-length(rhs),1)];
|
|
if (ncolU > 300); use_LU = 1; end
|
|
%%
|
|
%% Cholesky factorization
|
|
%%
|
|
L = []; resnrm = []; xx = inf*ones(m,1);
|
|
if (~use_LU)
|
|
solve_ok = 1; solvesys = 1;
|
|
nnzmat = mexnnz(coeff.mat11);
|
|
nnzmatdiff = (nnzmat ~= nnzmatold);
|
|
if (nnzmat > spdensity*m^2) | (m < 500)
|
|
matfct_options = 'chol';
|
|
else
|
|
if (par.matlabversion < 7.3)
|
|
matfct_options = 'spchol';
|
|
else
|
|
matfct_options = 'spcholmatlab';
|
|
end
|
|
end
|
|
if (printlevel>2); fprintf(' %s',matfct_options); end
|
|
if strcmp(matfct_options,'chol')
|
|
if issparse(schur); schur = full(schur); end;
|
|
if (iter<=5); %% to fix strange anonmaly in Matlab
|
|
mexschurfun(schur,1e-20,2);
|
|
end
|
|
L.matfct_options = 'chol';
|
|
L.perm = [1:m];
|
|
[L.R,indef] = chol(schur);
|
|
diagR = diag(L.R).^2;
|
|
%%semilogy(diagschur,'*'); hold on;
|
|
%%semilogy(diagR,'or'); hold off; pause(0.1);
|
|
if (indef)
|
|
diagR = diagRold;
|
|
solve_ok = -2; solvesys = 0;
|
|
msg = 'linsysolve: Schur complement matrix not pos. def.';
|
|
if (printlevel); fprintf('\n %s',msg); end
|
|
end
|
|
elseif strcmp(matfct_options,'spcholmatlab')
|
|
if ~issparse(schur); schur = sparse(schur); end;
|
|
L.matfct_options = 'spcholmatlab';
|
|
[L.R,indef,L.perm] = chol(schur,'vector');
|
|
L.Rt = L.R';
|
|
%%diagR(L.perm(1:length(L.R)),1) = diag(L.R).^2;
|
|
if (indef)
|
|
diagR = diagRold;
|
|
solve_ok = -2; solvesys = 0;
|
|
msg = 'linsysolve: Schur complement matrix not pos. def.';
|
|
if (printlevel); fprintf('\n %s',msg); end
|
|
end
|
|
elseif strcmp(matfct_options,'spchol')
|
|
if ~issparse(schur), schur = sparse(schur); end;
|
|
if (nnzmatdiff | ~strcmp(matfct_options,matfct_options_old))
|
|
[Lsymb,flag] = symbcholfun(schur,par.cachesize);
|
|
if (flag)
|
|
solve_ok = -2; solvesys = 0;
|
|
existspcholsymb = 0;
|
|
msg = 'linsysolve: symbolic factorization fails';
|
|
if (printlevel); fprintf('\n %s',msg); end
|
|
use_LU = 1;
|
|
else
|
|
existspcholsymb = 1;
|
|
end
|
|
end
|
|
if (existspcholsymb)
|
|
L = sparcholfun(Lsymb,schur);
|
|
L.matfct_options = 'spchol';
|
|
L.d(find(L.skip)) = 1e20;
|
|
if any(L.skip) & (ncolU)
|
|
solve_ok = -3; solvesys = 0;
|
|
existspcholsymb = 0;
|
|
use_LU = 1;
|
|
if (printlevel)
|
|
fprintf('\n L.skip exists but ncolU > 0.');
|
|
fprintf('\n switch to LU factor.');
|
|
end
|
|
end
|
|
end
|
|
end
|
|
if (solvesys)
|
|
if (ncolU)
|
|
tmp = coeff.mat12'*linsysolvefun(L,coeff.mat12)-coeff.mat22;
|
|
if issparse(tmp); tmp = full(tmp); end
|
|
tmp = 0.5*(tmp + tmp');
|
|
[L.Ml,L.Mu,L.Mp] = lu(tmp);
|
|
tol = 1e-16;
|
|
idx = find(abs(diag(L.Mu)) < tol);
|
|
if ~isempty(idx) & (printlevel); fprintf('**'); end
|
|
end
|
|
[xx,resnrm,solve_ok] = symqmr(coeff,rhs,L);
|
|
if (solve_ok<=0) & (printlevel)
|
|
fprintf('\n warning: symqmr fails: %3.1f.',solve_ok);
|
|
end
|
|
end
|
|
if (solve_ok < 0)
|
|
if (m < 5000 & strcmp(matfct_options,'chol')) | ...
|
|
(m < 1e5 & strcmp(matfct_options,'spchol')) | ...
|
|
(m < 1e5 & strcmp(matfct_options,'spcholmatlab'))
|
|
use_LU = 1;
|
|
if (printlevel); fprintf('\n switch to LU factor.'); end
|
|
end
|
|
end
|
|
end
|
|
%%
|
|
%% LU factorization
|
|
%%
|
|
if (use_LU)
|
|
nnzmat = mexnnz(coeff.mat11)+mexnnz(coeff.mat12);
|
|
nnzmatdiff = (nnzmat ~= nnzmatold);
|
|
solve_ok = 1; solvesys = 1;
|
|
if ~isempty(coeff.mat22)
|
|
raugmat = [coeff.mat11, coeff.mat12; coeff.mat12', coeff.mat22];
|
|
else
|
|
raugmat = coeff.mat11;
|
|
end
|
|
if (nnzmat > spdensity*m^2) | (m+ncolU < 500)
|
|
matfct_options = 'lu';
|
|
else
|
|
matfct_options = 'splu';
|
|
end
|
|
if (printlevel > 2); fprintf(' %s ',matfct_options); end
|
|
if strcmp(matfct_options,'lu')
|
|
if issparse(raugmat); raugmat = full(raugmat); end
|
|
[L.l,L.u,L.p] = lu(raugmat);
|
|
L.matfct_options = 'lu';
|
|
L.p = sparse(L.p);
|
|
idx = find(abs(diag(L.u)) < 1e-20);
|
|
if ~isempty(idx)
|
|
msg = 'linsysolve: matrix is singular';
|
|
if (printlevel); fprintf('\n %s',msg); end
|
|
solvesys = 0;
|
|
end
|
|
[ii,jj] = find(L.p); [dummy,idx] = sort(ii); L.perm = jj(idx);
|
|
end
|
|
if strcmp(matfct_options,'splu')
|
|
if ~issparse(raugmat); raugmat = sparse(raugmat); end
|
|
if (nnzmatdiff | ~strcmp(matfct_options,matfct_options_old))
|
|
Lsymb.perm = symamd(raugmat);
|
|
end
|
|
L.perm = Lsymb.perm;
|
|
L.matfct_options = 'splu';
|
|
if (par.matlabversion >= 6.5)
|
|
[L.l,L.u,L.p,L.q] = lu(raugmat(L.perm,L.perm));
|
|
else
|
|
[L.l,L.u,L.p] = lu(raugmat(L.perm,L.perm));
|
|
L.q = speye(length(raugmat));
|
|
end
|
|
end
|
|
if (solvesys)
|
|
[xx,resnrm,solve_ok] = mybicgstab(coeff,rhs,L,[],[],printlevel);
|
|
if (solve_ok<=0) & (printlevel)
|
|
fprintf('\n warning: bicgstab fails: %3.1f,',solve_ok);
|
|
end
|
|
end
|
|
end
|
|
if (printlevel>=3); fprintf(' %2.0d',length(resnrm)-1); end
|
|
%%
|
|
nnzmatold = nnzmat; matfct_options_old = matfct_options;
|
|
%%***************************************************************
|