Hello,
I am trying to use cplex.jl to solve linear problems without using JuMP or MOI. I want to use the barrier method without presolve and crossover/vertex, so I used these commands:
CPLEX.set_param!(env, "CPX_PARAM_MIPCBREDLP", 0) CPLEX.set_param!(env, "CPXPARAM_Barrier_Algorithm", 0)
Are these commands correct, and is there a way to be sure they are doing what I want, using some display messages?
Thanks
CPLEX.jl wraps CPLEX’s C API. You can find the full documentation at:
For example, the documentation for CPX_PARAM_MIPCBREDLP can be found:
I don’t think this is what you want because it is about callbacks.
Instead, you’re probably looking for:
This is the setting I use for solving problems with barrier, no presolve, no crossover.
Hope it can be useful.
CPXENV = CPLEX.Env() # create CPLEX environment
# Set parameters
CPLEX.set_param!(CPXENV, "CPXPARAM_ScreenOutput", 1) # Enable output
CPLEX.set_param!(CPXENV, "CPXPARAM_TimeLimit", 3600) # Time limit
CPLEX.set_param!(CPXENV, "CPXPARAM_Threads", 1) # Single thread
CPLEX.set_param!(CPXENV, "CPXPARAM_SolutionType", 2) # No crossover
CPLEX.set_param!(CPXENV, "CPXPARAM_LPMethod", 4) # Use barrier
CPLEX.set_param!(CPXENV, "CPXPARAM_Preprocessing_Presolve", 0) # No presolve
...
CPLEX should display non-default parameters values at the beginning of the optimization.
For the above case, you should see something like this in the log:
...
CPXPARAM_Preprocessing_Presolve 0
CPXPARAM_LPMethod 4
CPXPARAM_Threads 1
CPXPARAM_SolutionType 2
CPXPARAM_TimeLimit 3600
...
FWIW, I’ve observed that CPLEX still performs a “reduced presolve”, even with the CPXPARAM_Preprocessing_Presolve set to 0.
As odow points out, the answer to all these questions will be in CPLEX’s documentation, there’s nothing additional that’s done in the Julia wrapper.
Thank you!
Thank you this is what I was looking for!