Neural Network-based Performance Prediction for Task Migration on S-NUCA Many-Cores

The performance of a task running on a many-core with distributed shared last-level cache (LLC) strongly depends on two parameters: the thermally-safe power budget and the LLC latency. The thread-to-core mapping determines both the parameters and needs to make a trade-off. Arrival and departure of tasks on a many-core change its state significantly in terms of available cores and power budgets. Task migrations can thereupon be used to keep the many-core operating at peak performance. Furthermore, a task's characteristics may change with its execution phases mandating its migration on-the-fly. We propose the first run-time algorithm PCMig that increases the performance of a many-core with distributed shared LLC by migrating tasks based on their phases and the many-core's state. PCMig is based on a model that predicts the performance impact of migrations. We propose a performance prediction model based on a lightweight neural network (NN). To serve as a reference, we also propose an analytical model of the many-core that operates on CPI stacks. The NN-based model achieves a higher prediction accuracy at a lower overhead than an analytical model. PCMig results in an up to 7.3% increase in average performance (up to 20% for individual applications) compared to the state-of-the-art at an overhead of less than 0.5%.