<br><font size=2 face="sans-serif">Yiming,</font>
<br>
<br><font size=2 face="sans-serif">This "sample" Clp/Samples/network.cpp
was really for testing. You would just need one call e.g. dual().</font>
<br>
<br><font size=2 face="sans-serif">However as I never announced any network
capability in Clp, I thought I would also post this on Coin-lpsolver</font>
<br><font size=2 face="sans-serif">-----------------------</font>
<br>
<br><font size=2 face="sans-serif">The original purpose of the sample was
twofold - to test the ClpMatrixBase concept and to give some of the advantages
of a specialized network code.</font>
<br>
<br><font size=2 face="sans-serif">In Clp a matrix is an abstract class
so special classes can be used for specialized cases. The default
matrix class is ClpPackedMatrix which is just a CoinPackedMatrix. To
see how easy it was to create new classes I created a CLPPlusMinusOneMatrix
class which saves memory and should be faster on most machines if all elements
are +- 1.0. I also created a ClpNetworkMatrix class (one or two elements
a column, + and - one if two, + or - one if one). In this case Clp
also uses a network factorization so is significantly faster. Code
testing this is also in unitTest.cpp.</font>
<br>
<br><font size=2 face="sans-serif">This will not be as fast as a specialized
network code as it is not tightly integrated but for the same reason it
should be more flexible.</font>
<br>
<br><font size=2 face="sans-serif">John Forrest</font>