1 |
Fortunately very irregular problems like power systems tend not to be huge.
-
100's or 10,000's not billions of points
-
Thus can use very difficult optimized "full matrix exploiting zero structure" algorithms
|
2 |
In PDE case, one can show for some elliptic problems that red black can be better than original ordering
|
3 |
For some hyperbolic equations used in computational fluid dynamics, this is not so clear and SOR methods are used
|
4 |
If one has more complex differencing schemes e.g. fourth order differencing, then red-black does not work but a scheme with more colors (and more update phases) can be devised
-
Difficulty is irregular graphs not complex stencils
|