Hi ​@afattahi. How to warm start a model depends on the model type (LP or MIP). For LP, if the solver is using the primal or dual simplex algorithm then you should set the Solvers General option 'accept basis’ to 1 (AIMMS will store the basis automatically by default). The barrier algorithm cannot efficiently use a warm start. You could use barrier for the first solve and then use the primal or dual simplex algorithm for the successive solves (assuming barrier does a crossover step, which is normally done by default).

For MIP you can provide a MIP start using the procedure GMP::Solution::SetMIPStartFlag. Note that using a MIP start will not help much if the "proving optimality” part takes the most time.

How to efficiently do model updates using GMP functionality depends on your AIMMS project. Please send an email to our support to discuss in detail.

I am not familiar with the GUSS functionality of AIMMS but, after taking a glance at it, it seems that AIMMS does not have something similar.

Hi ​@Marcel Hunting,
Thanks for your explanation.

I am dealing with a large LP and usually only aim for the barrier solution. However, if it can reduce the runtime of successive GMPs through warm start, it would make sense to have the crossover enabled for the first run. 

I (or my colleague) will send an email to support for further discussing this.