Logo des Repositoriums

Auflistung nach Schlagwort "power management"

1 - 2 von 2
  • Textdokument
    Remote AVX Overhead: Detection and Mitigation
    (Tagungsband des FG-BS Frühjahrstreffens 2021, 2021) Gottschlag, Mathias
    Due to power constraints, recent Intel CPUs reduce their frequency when executing AVX2 and AVX-512 instructions. Often, this frequency reduction affects other applications as well, which reduces overall performance and prevents contemporary operating systems from fairly distributing system resources. In our work, we show that these problems are fundamental problems of power-limited computing. We analyze the problems and show a method to quantify the underlying AVX overhead. Based on our analysis, we then describe a set of operating system techniques to improve performance and scheduler fairness. Our results show the importance of active management of hardware-controlled frequency scaling by the OS. Based on this observation, we sketch improved hardware-software interfaces which could further reduce AVX overhead and improve the efficacy of our approach.
  • Textdokument
    Toward Dynamic Orchestration of Data/Power/Process Management for Hybrid Memory Based Systems
    (Tagungsband des FG-BS Herbsttreffens 2021, 2021) Arima, Eishi; Trinitis, Carsten; Schulz, Martin
    The exponential growth of the transistor count on VLSI circuits, known as Moore’s law, is slowing down, and the end of the technology scaling is predicted to be inevitable. As a consequence, computing system architectures are gradually shifting toward extremely heterogeneous designs consisting of multiple different hardware devices or accelerators in each component. As one example, over the past few years the industry has begun to support hybrid memory systems in their products based on emerging memory device technologies, most prominently HBM (High-Bandwidth Memory) and NVRAM (Non-Volatile RAM). This hardware trend has opened up new research opportunities in the system software and operating system area. In this position paper, we focus on data, power and process management in hybrid memory based systems, with a particular focus on a coordinated and dynamic approach. This is based on our key insight, which is brought by our prior studies, that the on such systems memory access/utilization behavior as well as the memory management policy plays an important role for various optimizations, including power management and process (or job) scheduling. In this position paper, we clarify the problem, provide a high-level software architecture, and finally discuss the major challenges to realize it.