Fine-grained Parallel Ant Colony System for Shared-Memory Architectures

Although Ant Colony Systems (ACS) have gained much attention in last two decades but slow execution and convergence speed are still two challenges for these metaheuristic algorithms. Many parallel implementations have been proposed for faster execution. However, most of available implementations use coarse-grained synchronization mechanisms that are not efficient and scalable. In this work, we have taken a fine-grained (ant-level) approach that is more efficient and scalable. We have used traveling salesman problem as a test case and have presented a parallel finegrained implementation for shared-memory multi-core systems. Our experimental results show that our proposed parallel implementation can achieve considerably higher speedup values on modern multicore processors. General Terms Optimization, Parallel Processing.