Reputation: 836
I am working through The NURBS Book by Piegl and Tiller. For the global interpolation algorithm, they require you to provide two utility routines for solving a system of linear equations:
LUDecomposition(A, q, sbw)
to decompose theq x q
coefficient matrix with semibandwidthsbw
into lower and upper triangular components; for simplicity we assumeA
is anq x q
square array, but a utility shoule be used which only stores the nonzero band.
ForwardBackward(A, q, sbw, rhs, sol)
to perform the forward/backward substitution (see [Press88]); rhs[] is the right hand side of the system (the coordinates of the Q_k), and sol[] is the solution vector (coordinates of the P_i).
Checking the reference Press88, I found that it is Numerical Recipes in C. I should be able to rework the algorithm in that book to get the ForwardBackward
function, but as far as the LUDecomposition
goes, where can I find one that works for the special case of a matrix with diagonal bands?
Upvotes: 1
Views: 118
Reputation:
Matlab code for LU decomp of Tridiagonal matrix
function [u1,d1,l1] = decomt(u,d,l)
%length of diagonal
n=length(d);
u1=u;
d1 = d;
l1=l;
%perform LU decomp
d1(1) = d(1);
for i =2:n
l1(i-1) = l(i-1)/d1(i-1); %update lower triangle
d1(i)= d(i) - (l(i-1)/d1(i-1))*u(i-1); % update diagonal
end
end
In order to solve, front sub and back sub.
function [x] = solvet(u,d,l,b)
n=length(d);
x = (1:n);
y =(1:n);
% Solve tridiag LUx=b
% Step 1 Solve Ly=b for y
y(1) = b(1)
for i=2:n
y(i) = b(i) - l(i-1)*y(i-1);
end
%Step 2 : Solve Ux=y for x
x(n) = y(n)/d(n);
for i=(n-1):-1:1
x(i) = (y(i)-u(i)*x(i+1))/d(i);
end
end
Upvotes: 0