Originally posted by: yanzhiping

Excuse me,

In the attached log file, I set the mip gap to 35% which can be seen from the parameter CPXPARAM_MIP_Tolerances_MIPGap.

Why does the problem not stop when it reached the gap 31.38% after 4904.86 sec which is less than 35%? and it took 4000s more while stuck at the same gap.

Thanks.

This could be a thread-synchronization issue: at some point each thread is solving one node of the search tree. One thread eventually finds that good solution. The other threads however will first finish the node they are processing before they check whether another thread has found a new solution.

Does the problem persist if you reduce to 1 thread?

Also, it looks the root relaxation solves quickly in 90 seconds but then processing 10 nodes requires almost 3000 seconds. Maybe there are some parameters you could tweak to improve on that. Can you share the model?