An efficient voltage scaling algorithm for complex SoCs with few number of voltage modes

Abstract

Increasing demand for larger high-performance applications requires developing more complex systems with hundreds of processing cores on a single chip. To allow dynamic voltage scaling in each on-chip cores individually, many on-chip voltage regulators must be used. However, the limitations in implementation of onchip inductors can reduce the efficiency, accuracy and the number of voltage modes generated by regulators. Therefore the future voltage scheduling algorithms must be efficient, even in the presence of few voltage modes; and fast, in order to handle complex applications. Techniques proposed to date, need many fine-grained voltage modes to produce energy efficient results and their quality degrades significantly as the number of modes decreases. This paper presents a new technique called Adaptive Stochastic Gradient Voltage and Task Scheduling (ASG-VTS) that quickly generates very energy efficient results irrespective of the number of available voltage modes. The results of comparing our algorithm to the most efficient approaches (RVS and EE-GLSA) show that in the presence of only four valid modes, the ASG-VTS saves up to 26% and 33% more energy. On the other hand, other approaches require at least ten modes to reach the same level of energy saving that ASG-VTS achieves with only four modes. Therefore our algorithm can also be used to explore and minimize the number of required voltage levels in the system. Categories and Subject Descriptors

A Novel Voltage Scaling Algorithm through Ant Colony Optimization for Embedded Distributed Systems

“…Here, we divide range between V dd and V t of each PE evenly into 4 or 30 modes. The results would be improved little when the number of modes is greater than 30 [5]. Note that, under the fine-grained 30 modes, the three algorithms of EE-GLSA, ASG-VTS and ACO-VTS produce nearly the same results, and ACO-VTS's results are more optimal than the other two under most test benches, with approximately equal time complexities.…”

“…Implementing the parallelizing ACO-VTS and its approximation approach by C++ with MPI, we compare the results of the two algorithms operated on the Blade Server with EE-GLSA and ASG-VTS based on benchmarks tgff1-tgff25 used in [1,2,4,5]. These benchmarks are generated using TGFF [8], a tool producing pseudo-random task graphs, so it is quite practical in the real world.…”

“…So far, all the algorithms above are sensitive to the count of voltage modes provided by the processing element. But, in [5], Bita Gorjiara proposes an algorithm ASG-VTS, which could produce results not affected by the count of the modes, that selects a voltage mode for each tasks stochastically and sorts the tasks based on the state of the voltage modes. ASG-VTS is also an efficient algorithm, but, although it generates quite energy-efficient results, some of the results seem not so optimal as the algorithms before.…”

Chemical Reaction Optimization for Task Mapping in Heterogeneous Embedded Multiprocessor Systems

“…There are many papers [2,4] that focus solely on the TSVS subproblem and there are also some papers [3,9] that assume that the task ordering is known and focus only on voltage scaling.…”

A Multi-core Testbed on Desktop Computer for Research on Power/Thermal Aware Resource Management

“…Gojiara et al [23] developed a technique which quickly generates very energy efficient results irrespective of the number of available voltage models. Dabiri et al [16] consider the effects of voltage on single event upsets (SEUs) to develop a method for energy optimization under SEU constraints.…”