[metapost] memory problem

[Boguslaw Jackowski]

Denis:
> But this doesn't answer the slow solving of a large number
> of independent equations. I think we should work on a non-metaobj
> example and focus on it.

Taco:
> Agreed. This is core functionality, and as such, its running time
> should be x^3. 
>
> make that:
>   should _not_ be x^3

Knuth applies ``the standard technique of Gaussian elimination''
to convert a system of equations to the diagonal form. The complexity
of this method is x^3. Is it possible ( practically, i.e., not
using tricks of the kind of Strassen's fast matrix multiplication,
http://mathworld.wolfram.com/StrassenFormulas.html ) to gain
a significant speed-up? Sorry for a silly question, but I'm not
``au courant'' with respect to recent advances in numerical
methods...

Cheers.

-- 

Jacko