Hi,

I have constructed a MILP with the size of:

______________________________________________________________

CPU model: Intel(R) Xeon(R) Platinum 8370C CPU @ 2.80GHz, instruction set SSE2|AVX|AVX2|AVX512]
Thread count: 24 physical cores, 48 logical processors, using up to 48 threads
Optimize a model with 1081904 rows, 968127 columns and 12886704 nonzeros

_______________________________________________________________

I am using an older version of AIMMS: 4.92.12. I am using the GMP library to generate and solve my instances.

I solved 10 instances with each 4*5=20 different parameter settings (so 200 solver sessions in total). My procedures are set up as follows (simplified version): Import instances → Generate 4*5 =20 versions of each instance → start asynchronous execution of 20 instances → WaitforSingleCompletion → Save individual instance data. 

I use a virtual machine with 24 cores and 48 threads. I use Gurobi as a solver and have a WLS licence for 200 parallel solver sessions. The thread limit I use is 12 threads, because I believe this sets the maximum number of threads per solver session. So I can run 4 instances in parallel.

During my experiments, I experienced a problem; it seems like Aimms cannot handle this experiment (i.e. task manager shows that Aimms is not responding, and CPU usage is 2). However, the gurobi.log file and the "projectname".lis file indicate that Gurobi is still solving the problem using the branch and bound method. Hence, a low CPU usage is illogical.

Can you help me figure out why my experiments are not finished as expected? I believe it has something to do with the (data) communication between AIMMS and Gurobi and the fact that AIMMS handles its procedures in a sequential order by means of an event stream/logger.

Kind regards,

Ieke Schrader