To maintain reliable operation, task allocation for manycore processors must consider the heat interaction of processor cores and network-on-chip routers in performing task assignment. Our approach employs reinforcement learning, a machine learning algorithm that performs task allocation based on current core and router temperatures and a prediction of which assignment will minimize maximum temperature in the future. The algorithm updates prediction models after each allocation based on feedback regarding the accuracy of previous predictions. Our new algorithm is verified via detailed many-core simulation which includes on-chip routing. Our results show that the proposed technique is fast (scheduling performed in < 1 ms) and can efficiently reduce peak temperature by up to 8 o C in a 49-core processor (4.3 o C on average) versus a competing task allocation approach for a series of SPLASH-2 benchmarks.

No abstract

Contemporary multi-core architectures deployed in embedded systems are expected to function near the operational limits of temperature, voltage, and device wear-out. To date, most on-chip sensing systems have been designed to collect and use sensor information for these parameters locally. In this paper, a new sensing system to enhance multi-core dependability which supports both the local and global distribution of sensing data in embedded processors is considered. The benefit of the new sensing architecture is verified using the broadcast of microarchitectural parameter signatures which can be used to identify impending voltage droops. Low-latency broadcasts are supported for a range of sensor data transfer rates. Up to a 9% performance improvement for a 16-core system is determined via the use of the distributed voltage droop sensor information (5.4% on average). The entire sensing system, including broadcasting resources, requires about 2.6% of multi-core area.

No abstract

In order to understand the different performances of robots under different control systems, the author has carried out dynamic optimization research on the control system of robots combined with differential algebraic equations. In this study, the general form of the discrete differential-algebraic equation (DAE) optimization problem using the Orthogonal Configuration of Finite Element (OCFE) method is deeply analyzed, and the equivalent conditions of the direct discrete scheme and the indirect discrete scheme are obtained through rigorous proof. On this basis, a variety of common configuration methods are simulated and analyzed, and it is found that indirect Lobatto configuration can achieve better results in many aspects. The results show that the discrete algorithm using differential algebraic equations can effectively achieve dynamic optimization of the control system, thus achieving the author’s research purpose.