Parallelism and Auto-Tuner of Irregular Iteration in Reservoir Simulation

In the field of reservoir numerical simulation, irregular iteration problem is one of the most important reasons to affect the efficiency of large-scale computing. Memory access patterns can´t be determined at compiler-time, which brings us difficulties to parallel. This library uses Inspector/Executor model based on distributed clusters, and improves parallel efficiency utilizing alternate tiling technology. Besides, we get the optimal efficiency by auto-tuner, which can find the right parameters. Finally, numerical results on CLUSTER show that the method has a good parallel efficiency as well as scalability, and the auto-tuner possesses practicality to a certain extent.