[Clp] Performance test much worse than GLPK, any way to optimize it?

[Emerson]

Hi Michael,
Thanks for replying. Good news is : problem solved!
I firstly used coinbuild but nothing changed. Then I changed the solver
type from initialSolve( ) to Babak's suggestion:

ClpSimplex *Model ....

ClpSolve *solvectl;

solvectl= new ClpSolve;

solvectl->setSolveType(ClpSolve::useDual);

solvectl->setPresolveType(ClpSolve::presolveOn);

Model->initialSolve(*solvectl);

delete solvectl;

Although the pointer case doesn't work(don't know why), I changed it to
ClpSimplex Model, and everything works!
Now the performance increases so much, from 58 seconds down to 2 seconds on
my network model, while GLPK takes 21 seconds. I hope everything goes well
after this...

Thanks all,
Emer

On Tue, Jun 30, 2015 at 6:48 PM, Michael Hennebry <
hennebry at web.cs.ndsu.nodak.edu> wrote:

> The first thing is to know what you are measuring.
> CPU time and wall clock time can be rather different.
> Also, what events are you timing?
> Do your times include input and output times?
>
> --
> Michael   hennebry at web.cs.ndsu.NoDak.edu
> "SCSI is NOT magic. There are *fundamental technical
> reasons* why it is necessary to sacrifice a young
> goat to your SCSI chain now and then."   --   John Woods
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/clp/attachments/20150630/d5e90e69/attachment.html>