MATLAB confusion

[Shadow]

Ok, I have to compare two iterative methods, Jacobi method and Gauss-Seidel and their convergence through the residuals in MATLAB and that's where I'm running into a bit of a wall.

I've gotten the Jacobi method coded to solve the linear equations but the residual convergence is giving me problems.
Checking up some codes available online, they've done the convergence check using norm function, like this.

Code: Select all

while relerr > tol
relerr = norm( A*x-b,inf )/norm( b,inf );

which is

norm(x,inf) The infinity norm, or largest row sum of A, max(sum(abs(A')))

and my lecturer's info gives residual as

Code: Select all

 r=abs(b-A*x)

Anyone know why the residual is done with that norm thing instead of simply this way

Code: Select all

while relerr > tol
relerr = abs(b-A*X(:,niter))
%niter =current iteration count, so the X(:,niter) picks up all the rows but only the last column with the most accurate iteration guesses.

and what's the advantage when done using the norm function is any?

[Sweendoggy]

did you time each calculation method? I'm sure one is more efficient than the other

[Shadow]

Not yet, I was searching for info on the convergence test, stumbled onto the code with that norm function and went "WTF is this??"
I'm still woefully inept when it comes to matlab.

[Shadow]

Yeah, the norm function does seem to be much faster than using the simple abs function. Now I just need to know why.