[Coin-discuss] "zero distance in CbcRounding - debug" message using Cbc/Clp

[Kish Shen]

John,

Thanks for your reply.

> As it is a heuristic it obviously doesn't matter much.

I wonder if the message is showing some unexpected behaviour for the
heuristic -- this MIP problem (which seems to be some modified version of
MIPLIB's noswot) is one of the few that shows this message (when integer
presolving is turned off), and it is also the problem that shows the
greatest difference in solving times with other solvers -- it is between
150 and 600 times slower than CPLEX (depending on the options such as
presolve on or off etc.), and is even 4 to 10 times slower than
glpk -- the next largest ratio between CPLEX and Cbc/Clp is under 50 times
(and that is for a small problem), and most are under 10 times; and almost no
other problem is slower than glpk.

Cheers,

Kish Shen
IC-Parc
Imperial College London
U.K.