On the potential of significance-driven execution for energy-aware HPC

Dynamic Voltage and Frequency Scaling (DVFS) exhibits fundamental limitations as a method to reduce energy consumption in computing systems. In the HPC domain, where performance is of highest priority and codes are heavily optimized to minimize idle time, DVFS has limited opportunity to achieve substantial energy savings. This paper explores if operating processors Near the transistor Threshold Volt- age (NTV) is a better alternative to DVFS for break- ing the power wall in HPC. NTV presents challenges, since it compromises both performance and reliability to reduce power consumption. We present a first of its kind study of a significance-driven execution paradigm that selectively uses NTV and algorithmic error tolerance to reduce energy consumption in performance- constrained HPC environments. Using an iterative algorithm as a use case, we present an adaptive execution scheme that switches between near-threshold execution on many cores and above-threshold execution on one core, as the computational significance of iterations in the algorithm evolves over time. Using this scheme on state-of-the-art hardware, we demonstrate energy savings ranging between 35% to 67%, while compromising neither correctness nor performance.

Energy-Aware Rate and Description Allocation Optimized Video Streaming for Mobile D2D Communications

The proliferation problem of video streaming applications and mobile devices has prompted wireless network operators to put more efforts into improving quality of experience (QoE) while saving resources that are needed for high transmission rate and large size of video streaming. To deal with this problem, we propose an energy-aware rate and description allocation optimization method for video streaming in cellular network assisted device-to-device (D2D) communications. In particular, we allocate the optimal bit rate to each layer of video segments and packetize the segments into multiple descriptions with embedded forward error correction (FEC) for realtime streaming without retransmission. Simultaneously, the optimal number of descriptions is allocated to each D2D helper for transmission. The two allocation processes are done according to the access rate of segments, channel state information (CSI) of D2D requester, and remaining energy of helpers, to gain the highest optimization performance. Simulation results demonstrate that our proposed method (named OPT) significantly enhances the performance of video streaming in terms of high QoE and energy saving.

HpMC: An Energy- Aware Management System for Multi-Level Memory Architectures

DRAM technology faces density and power challenges to increase capacity because of limitations of physical cell design. To overcome these limitations, system designers are exploring alternative solutions that combine DRAM and emerging NVRAM technologies. Previous work on heterogeneous memories focuses, mainly, on two system designs: PCache, a hierarchical, inclusive memory system, and HRank, a flat, non-inclusive memory system. We demonstrate that neither of these designs can universally achieve high performance and energy efficiency across a suite of HPC workloads. In this work, we investigate the impact of a number of multilevel memory designs on the performance, power, and energy consumption of applications. To achieve this goal and overcome the limited number of available tools to study heterogeneous memories, we created HMsim, an infrastructure that enables n-level, heterogeneous memory studies by leveraging existing memory simulators. We, then, propose HpMC, a new memory controller design that combines the best aspects of existing management policies to improve performance and energy. Our energy-aware memory management system dynamically switches between PCache and HRank based on the temporal locality of applications. Our results show that HpMC reduces energy consumption from 13% to 45% compared to PCache and HRank, while providing the same bandwidth and higher capacity than a conventional DRAM system.

Cross-layer Energy-Efficiency Optimization of Packet Based Wireless MIMO Communication Systems

Energy in today's short-range wireless communication is mostly spent on the analog- and digital hardware rather than on radiated power. Hence,purely information-theoretic considerations fail to achieve the lowest energy per information bit and the optimization process must carefully consider the overall transceiver. In this paper, we propose to perform cross-layer optimization, based on an energy-aware rate adaptation scheme combined with a physical layer that is able to properly adjust its processing effort to the data rate and the channel conditions to minimize the energy consumption per information bit. This energy proportional behavior is enabled by extending the classical system modes with additional configuration parameters at the various layers. Fine grained models of the power consumption of the hardware are developed to provide awareness of the physical layer capabilities to the medium access control layer. The joint application of the proposed energy-aware rate adaptation and modifications to the physical layer of an IEEE802.11n system, improves energy-efficiency (averaged over many noise and channel realizations) in all considered scenarios by up to 44%.

Using Computational Significance and Resilience in System Software Stacks.:Keynote Talk

This talk explores how the runtime system and operating system can leverage metrics that express the significance and resilience of application components in order to reduce the energy footprint of parallel applications. We will explore in particular how software can tolerate and indeed exploit higher error rates in future processors and memory technologies that may operate outside their safe margins.

VEDA:Variation-aware energy-efficient discrete Wavelet Transform architecture

In this paper, we present a unified approach to an energy-efficient variation-tolerant design of Discrete Wavelet Transform (DWT) in the context of image processing applications. It is to be noted that it is not necessary to produce exactly correct numerical outputs in most image processing applications. We exploit this important feature and propose a design methodology for DWT which shows energy quality tradeoffs at each level of design hierarchy starting from the algorithm level down to the architecture and circuit levels by taking advantage of the limited perceptual ability of the Human Visual System. A unique feature of this design methodology is that it guarantees robustness under process variability and facilitates aggressive voltage over-scaling. Simulation results show significant energy savings (74% - 83%) with minor degradations in output image quality and avert catastrophic failures under process variations compared to a conventional design. © 2010 IEEE.

A Programming Model and Runtime System for Significance-Aware Energy-Efficient Computing.

We introduce a task-based programming model and runtime system that exploit the observation that not all parts of a program are equally significant for the accuracy of the end-result, in order to trade off the quality of program outputs for increased energy-efficiency. This is done in a structured and flexible way, allowing for easy exploitation of different points in the quality/energy space, without adversely affecting application performance. The runtime system can apply a number of different policies to decide whether it will execute less-significant tasks accurately or approximately.

The experimental evaluation indicates that our system can achieve an energy reduction of up to 83% compared with a fully accurate execution and up to 35% compared with an approximate version employing loop perforation. At the same time, our approach always results in graceful quality degradation.

Circuit-aware design of energy-efficient massive MIMO systems

Densification is a key to greater throughput in cellular networks. The full potential of coordinated multipoint (CoMP) can be realized by massive multiple-input multiple-output (MIMO) systems, where each base station (BS) has very many antennas. However, the improved throughput comes at the price of more infrastructure; hardware cost and circuit power consumption scale linearly/affinely with the number of antennas. In this paper, we show that one can make the circuit power increase with only the square root of the number of antennas by circuit-aware system design. To this end, we derive achievable user rates for a system model with hardware imperfections and show how the level of imperfections can be gradually increased while maintaining high throughput. The connection between this scaling law and the circuit power consumption is established for different circuits at the BS.

Resting energy expenditure in non-ventilated, non-sedated patients recovering from serious traumatic brain injury: Comparison of prediction equations with indirect calorimetry values

Background & aims: Little is known about energy requirements in brain injured (TBI) patients, despite evidence suggesting adequate nutritional support can improve clinical outcomes. The study aim was to compare predicted energy requirements with measured resting energy expenditure (REE) values, in patients recovering from TBI.

Methods: Indirect calorimetry (IC) was used to measure REE in 45 patients with TBI. Predicted energy requirements were determined using FAO/WHO/UNU and Harris–Benedict (HB) equations. Bland– Altman and regression analysis were used for analysis.

Results: One-hundred and sixty-seven successful measurements were recorded in patients with TBI. At an individual level, both equations predicted REE poorly. The mean of the differences of standardised areas of measured REE and FAO/WHO/UNU was near zero (9 kcal) but the variation in both directions was substantial (range 591 to þ573 kcal). Similarly, the differences of areas of measured REE and HB demonstrated a mean of 1.9 kcal and range 568 to þ571 kcal. Glasgow coma score, patient status, weight and body temperature were signi?cant predictors of measured REE (p < 0.001; R²⁼ 0.47).

Conclusions: Clinical equations are poor predictors of measured REE in patients with TBI. The variability in REE is substantial. Clinicians should be aware of the limitations of prediction equations when estimating energy requirements in TBI patients.

Strategies for Energy-Efficient Resource Management of Hybrid Programming Models

Many scientific applications are programmed using hybrid programming models that use both message passing and shared memory, due to the increasing prevalence of large-scale systems with multicore, multisocket nodes. Previous work has shown that energy efficiency can be improved using software-controlled execution schemes that consider both the programming model and the power-aware execution capabilities of the system. However, such approaches have focused on identifying optimal resource utilization for one programming model, either shared memory or message passing, in isolation. The potential solution space, thus the challenge, increases substantially when optimizing hybrid models since the possible resource configurations increase exponentially. Nonetheless, with the accelerating adoption of hybrid programming models, we increasingly need improved energy efficiency in hybrid parallel applications on large-scale systems. In this work, we present new software-controlled execution schemes that consider the effects of dynamic concurrency throttling (DCT) and dynamic voltage and frequency scaling (DVFS) in the context of hybrid programming models. Specifically, we present predictive models and novel algorithms based on statistical analysis that anticipate application power and time requirements under different concurrency and frequency configurations. We apply our models and methods to the NPB MZ benchmarks and selected applications from the ASC Sequoia codes. Overall, we achieve substantial energy savings (8.74 percent on average and up to 13.8 percent) with some performance gain (up to 7.5 percent) or negligible performance loss.

Containing the Nanometer "Pandora-Box": Cross-Layer Design Techniques for Variation Aware Low Power Systems

The demand for richer multimedia services, multifunctional portable devices and high data rates can only been visioned due to the improvement in semiconductor technology. Unfortunately, sub-90 nm process nodes uncover the nanometer Pandora-box exposing the barriers of technology scaling-parameter variations, that threaten the correct operation of circuits, and increased energy consumption, that limits the operational lifetime of today's systems. The contradictory design requirements for low-power and system robustness, is one of the most challenging design problems of today. The design efforts are further complicated due to the heterogeneous types of designs ( logic, memory, mixed-signal) that are included in today's complex systems and are characterized by different design requirements. This paper presents an overview of techniques at various levels of design abstraction that lead to low power and variation aware logic, memory and mixed-signal circuits and can potentially assist in meeting the strict power budgets and yield/quality requirements of future systems.

A Significance-Driven Programming Framework for Energy-Constrained Approximate Computing

Approximate execution is a viable technique for energy-con\-strained environments, provided that applications have the mechanisms to produce outputs of the highest possible quality within the given energy budget.
We introduce a framework for energy-constrained execution with controlled and graceful quality loss. A simple programming model allows users to express the relative importance of computations for the quality of the end result, as well as minimum quality requirements. The significance-aware runtime system uses an application-specific analytical energy model to identify the degree of concurrency and approximation that maximizes quality while meeting user-specified energy constraints. Evaluation on a dual-socket 8-core server shows that the proposed
framework predicts the optimal configuration with high accuracy, enabling energy-constrained executions that result in significantly higher quality compared to loop perforation, a compiler approximation technique.