Review Article

Milestones in the Development of Iterative Solution Methods

Algorithm 2

Preconditioned conjugate gradient algorithm.
Given        𝐱 ( 0 ) , 𝜀          initial guess and
                  stopping tolerance
Set         𝐱 ( 0 ) , 𝐠 = 𝐴 𝐱 𝐛 ,
            𝐡 = [ 𝐵 ] 1 𝐠
     𝛿 0 = 𝐠 𝑇 𝐡
     𝐝 = 𝐡        initial search direction
Repeat    until convergence
     𝐡 = 𝐴 𝐝
     𝜏 = 𝛿 0 / ( 𝐝 𝑇 𝐡 )
     𝐱 = 𝐱 + 𝜏 𝐝       new approximation
     𝐠 = 𝐠 + 𝜏 𝐡       new (iterative) residual
     𝛿 1 = 𝐠 𝑇 𝐠
     𝐡 = [ 𝐵 ] 1 𝐠       new pseudoresidual
     𝛿 1 = 𝐠 𝑇 𝐡
    if 𝛿 1 𝜀 then stop
     𝛽 = 𝛿 1 / 𝛿 0 , 𝛿 0 = 𝛿 1
     𝐝 = 𝐡 + 𝛽 𝐝        new search direction