Parallel Jacobi Algorithm11

PARALLEL JACOBI
ALGORITHM
Rayan Alsemmeri
Amseena Mansoor
LINEAR SYSTEMS
Jacobi method is used to solve linear systems of the
form Ax=b, where A is the square and invertible.
Recall that if A is invertible there is unique solution
METHODS SOLVE LINEAR SYSTEMS
Direct solvers
Gaussian elimination
LU decomposition
Iterative solvers
Stationary iterative solvers
 Jacobi
 Gauss-Seidel
 Successive over-relaxation
Non-Stationary iterative methods

 Generalized minimum residual (GMRES)
 Conjugate gradient
Direct vs Iterative
 Direct Method
 -Dense systems
Gaussian Eliminations
Changes sparsity pattern -introduces non-zero entries which were originally
zero
 Iterative Method
 Sparse systems(usually come in very large size)
Jacobi method:
Main source is numerical approximation of PDE
-
ITERATIVE METHODS
Starts with an initial approximation for the solution
vector (x0)
 At each iteration algorithm updates the x vector by
using the sytem Ax=b
 During the iterations coefficient A, matrix is not
changed so sparsity is preserved
Each iteration involves a matrix-vector product
If A is sparse this product is efficiently done
Jacobi Algorithm
The first iterative technique is called the Jacobi method. This
method makes two assumptions: First, the system given by
has a unique solution

Jacobi Method
The coefficient matrix A has no zeros on its main diagonal. If any of the diagonal entries
are zero, then rows or columns must be interchanged to obtain a coefficient matrix that
has nonzero entries on the main diagonal.
To begin the Jacobi method, solve the first equation for x1, the second equation for x2
and so on, as follows.
How to apply the jacobi method
Continue the iterations until two successive approximations are identical when rounded to
three significant digits.
To begin, write the system in the form
Example of Jacobi
Stopping Criteria
Difference between two consecutive approximations
component wise is less than some tolerance
There exist other ways of computing distance between
two vectors, using norms
Jacobi iteration
 x10 
a11 x1  a12 x2    a1n xn  b1  0
a21x1  a22 x2    a2 n xn  b2  x2 
x 
0
  
 0
an1 x1  an 2 x2    ann xn  bn  xn 
1
x 
1
n (bn  an1 x1  an 2 x2    ann1 xn 1 )
0 0 0
ann
SEQUENTIAL JACOBI ALGORITHM
Ax  b D is diagonal matrix
L is lower triangular matrix
A  D  L U U is upper triangular
matrix
k 1 1
x  D (b  ( L  U ) x ) k
Pseudo Code for Jacobi
X-new //new approximation
X-old//previous approximation
Tol//given(specified by the number)
Counter=0//counts number of iterations
Iter-max//maximum number of iterations(specified by problem)
While(diff>tol &&counter<iter_max)
{
X_new=D-1(b-(L+u)X_old);
Diff=X_new-X_old;
X_old=X_new;
Counter=counter+1;
}
Does Jacobi Always Converge?
As k,under what conditions on A the sequence {xk}
converges to the solution vector
For the same A matrix, one method may converge
while the other may diverge
Example of Divergence
How to guarantee the convergence
The coefficient matrix of A should be strictly
diagonally dominant matrix
If the coefficient matrix of A is not strictly diagonally
dominant matrix we can exchange the rows to keep it
strictly diagonally dominant
+……. 𝑎 11 𝑎 12 𝑎 13
𝑎 21 𝑎 22 𝑎 23
+……. 𝑎 31 𝑎 32 𝑎 33
+…….
Theorem
If A is strictly diagonally dominant, then the system of
linear equations given by, has a unique solution to
which the Jacobi method will converge for any initial
approximation
Parallel Implementations of Jacobi
Algorithm
x k 1  D 1 (b  ( L  U ) x k )
Xk+1 D-1 b L+U XK
             
             
         
             
       -    
             
             
     
      

 
   
    
   
      
  
Parallel Jacobi Algorithm
Row wise Matrix Vector multiplication
Shared-memory parallelization very straightforward
Consider distributed memory machine using MPI
Row wise with shared memory
Xk+1 D-1 b L+U XK
-
Pseudo code of Jacobi distributed memory
systems
1. Distribute D-1,b,L+U row wise at each node
2. Distribute initial guess X0 to all nodes
3. Perform Jacobi iterative at each node to compute
corresponding parts
4. Broadcast all parts of new approximation to the
master process(Let us say p=0)
5. Distribute new approximation to all nodes row wise
6. Repeat from 3
Complexity
The most expensive part is matrix vector

multiplication which is of order O(n2)
But, with p-threads we have the complexity O(n2/p)
Conclusion
Easier implementation in shared memory
Various Distribution schemes for distributed
system(block-cycle)
Modifications of Jacobi Method
-Gauss Seidel & Successive Over Relaxation(SOR)
References
http://www.amazon.com/Parallel-Programming-Mult
icore-Cluster-Systems/dp/364204817X
http://college.cengage.com/mathematics/larson/elem
entary_linear/5e/students/ch08-10/chap_10_2.pdf
www.eee.metu.edu.tr/~skoc/ee443/iterative_
methods.ppt
Thank You!!!!!!

Parallel Jacobi Algorithm11

Uploaded by

Copyright:

Available Formats

Parallel Jacobi Algorithm11

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Parallel Jacobi Algorithm11

Uploaded by

Copyright:

Available Formats

PARALLEL JACOBI

Non-Stationary iterative methods

has a unique solution

The most expensive part is matrix vector

You might also like