Parameter UpperCutoff not working

[ZacharieALES]

The parameter CPXPARAM_MIP_Tolerances_UpperCutoff should enable to cut any node of the branch and bound which is above a given value.

I try to use it on a toy knapsack example but it does not seem to work. I set the cutoff value to 18 but a solution of value 19 is returned:

using JuMP
using CPLEX

m = Model(CPLEX.Optimizer)
B = 12
p = [9, 8, 10, 9, 6]
w = [4, 2, 8, 8, 7]

pbSize = 5
@variable(m, x[1:pbSize], Bin) # x[i] = 1 iff object i is in the knapsack
@objective(m, Max, sum(x[i] * p[i] for i in 1:pbSize)) # the price of the objects in the knapsack is maximised
@constraint(m, sum(x[i] * w[i] for i in 1:pbSize) <= B) # the capacity of the knapsack is satisfied

set_optimizer_attribute(m, "CPXPARAM_MIP_Tolerances_UpperCutoff", 18)

optimize!(m)

@show JuMP.objective_value(m) # Displays 19 (which should not be allowed as 18 is set as a maximum)

Is it normal? Did I miss something?

I just want to add that I am aware that the use of this parameter alone can be replaced by the addition of a constraint (e.g., expression of the objective <= 18). However, I want to use this parameter to stop the optimization once a solution under 18 is found (as presented here). I also know that it could be done through a lazycallback but the use of callbacks slow the resolution and I would like to avoid that.

Cplex version: 12.10
Cplex.jl version: 0.9.3

[ZacharieALES]

My bad, in a maximization it is the parameter LowerCutoff that should be used and it would not prevent a solution of value 19 to be obtained but a solution of value <18...

[odow]

In a maximization it is the parameter LowerCutoff that should be used and it would not prevent a solution of value 19 to be obtained but a solution of value <18

Correct. See the documentation here: https://www.ibm.com/docs/ro/cofz/12.10.0?topic=parameters-lower-cutoff

Closing because this is not a bug in CPLEX.jl.