A Fast Fully Parallel Ant Colony Optimization Algorithm Based on CUDA for Solving TSP

Figure 2

Architecture of a Pascal-based GPU. It contains an interface connected with peripheral component interconnect express (PCI-e), a memory unit with caches and controllers, and several graphics processing clusters (GPCs) with streaming multiprocessors (SMs). GPC is a basic unit for carrying out computational tasks. GPCs work under the control of the thread engine, which dispatches tasks and the SMs are responsible for executing specific operations in parallel.