With the increase of demand for computing resources and the struggle to provide the necessary energy, power-aware resource management is becoming a major issue for the High-performance computing (HPC) community. Inclu...
ISBN:
(纸本)9783031695766;9783031695773
With the increase of demand for computing resources and the struggle to provide the necessary energy, power-aware resource management is becoming a major issue for the High-performance computing (HPC) community. Including reliable energy management to a super-computer's resource and job management system (RJMS) is not an easy task. The energy consumption of jobs is rarely known in advance and the workload of every machine is unique and different from the others. We argue that the first step toward properly managing energy is to deeply understand the energy consumption of the workload, which involves predicting the workload's power consumption and exploiting it by using smart power-aware scheduling algorithms. Crucial questions are (i) how sophisticated a prediction method needs to be to provide accurate workload power predictions, and (ii) to what point an accurate workload's power prediction translates into efficient energy management. In this work, we propose a method to predict and exploit HPC workloads' power consumption, with the objective of reducing the supercomputer's power consumption while maintaining the management (scheduling) performance of the RJMS. Our method exploits workload submission logs with power monitoring data, and relies on a mix of light-weight power prediction methods and a classical EASY Backfillling inspired heuristic. We base this study on logs of Marconi 100, a 980 servers supercomputer. We show using simulation that a light-weight history-based prediction method can provide accurate enough power prediction to improve the energy management of a large scale supercomputer compared to energy-unaware scheduling algorithms. These improvements have no significant negative impact on performance.
Query-aware database generator (QAGen) expects to generate an application scenario based on the anonymized query plans as well as the cardinality constraints of all operators. It prefers to have the similar performanc...
ISBN:
(纸本)9789819755516;9789819755523
Query-aware database generator (QAGen) expects to generate an application scenario based on the anonymized query plans as well as the cardinality constraints of all operators. It prefers to have the similar performance to the in-production real performance if applying the generated workload on the generated database. Touchstone is the first work achieving simulating the application with the first 16 queries in TPC-H. However, it is designed based on heuristic rules, and has a weak ability to guarantee the cardinality constraints from match operators, i.e., IN and LIKE, which are important operators for performance optimization. So in this paper, we propose Touchstone(+) to solve the problem QAGen involving match operators by modeling constraints from IN and LIKE into a Constraint Programming (CP) problem. After solving the CP problem, it provides an initial data distribution satisfying cardinality constraints from all match operators for the iterative parameter search algorithm of Touchstone. Experiments have verified the effectiveness of our design and we also open code sources [12] for reproducing all results.
In order to identify the most representative subset of features in high-dimensional data, a feature selection algorithm (AP-MSU) based on feature clustering and information theory is proposed. The algorithm introduces...
ISBN:
(纸本)9789819756179;9789819756186
In order to identify the most representative subset of features in high-dimensional data, a feature selection algorithm (AP-MSU) based on feature clustering and information theory is proposed. The algorithm introduces the AP clustering algorithm and multivariate symmetric uncertainty (MSU) based on the filtering feature selection algorithm's preliminary screening of relevant features, better demonstrating the interactions between multiple feature variables and their interactions with target variables. The features are evaluated sequentially by an MSU-based feature quality metric, which considers both redundancy and interaction among the candidate features in the selected feature set, and removes the redundant features by assessing the ability of the features to provide effective categorization information with a small amount of computation. The experimental results show that the AP-MSU feature selection algorithm can effectively select a good feature set on binary and multi-classified gene expression datasets, and has good classification effect on different classifiers. In addition, the classification accuracy can be improved by the algorithm obtained a lower dimensional subset of features.
Flexible job shop scheduling problem with energy and makespan minimization objectives, and uncertain processing times that are modeled with intervals is addressed in this work. The problem is solved by a genetic algor...
ISBN:
(纸本)9783031611360;9783031611377
Flexible job shop scheduling problem with energy and makespan minimization objectives, and uncertain processing times that are modeled with intervals is addressed in this work. The problem is solved by a genetic algorithm using a lexicographic goal programming approach and the results evaluated with respect to the lower and upper bounds that come from various sources and methods.
Information Retrieval (IR) remains an active, fast-paced area of research. However, most advances in IR have predominantly benefited the so-called "classical" users, e.g., English-speaking adults. We envisio...
ISBN:
(纸本)9783031560682;9783031560699
Information Retrieval (IR) remains an active, fast-paced area of research. However, most advances in IR have predominantly benefited the so-called "classical" users, e.g., English-speaking adults. We envision IR4U2as a forum to spotlight efforts that, while sparse, consider diverse, and often understudied, user groups when designing, developing, assessing, and deploying the IR technologies that directly impact them. The key objectives for IR4U2 are: (1) raise awareness about ongoing efforts focused on IR technologies designed for and used by often understudied user groups, (2) identify challenges and open issues impacting this area of research, (3) ignite discussions to identify common frameworks for future research, and (4) enable cross-fertilization and community-building by sharing lessons learned from research catering to different audiences by researchers and (industry) practitioners across various disciplines.
The rapid expansion of the Internet of Things (IoT) industry highlights the significance of workload characterization when evaluating microprocessors tailored for IoT applications. The streamlined yet comprehensive sy...
ISBN:
(数字)9789819703166
ISBN:
(纸本)9789819703159;9789819703166
The rapid expansion of the Internet of Things (IoT) industry highlights the significance of workload characterization when evaluating microprocessors tailored for IoT applications. The streamlined yet comprehensive system stack of an IoT system is highly suitable for synergistic software and hardware co-design. This stack comprises various layers, including programming languages, frameworks, runtime environments, instruction set architectures (ISA), operating systems (OS), and microarchitecture. These layers can be bucketed into three primary categories: the intermediate representation (IR) layer, the ISA layer, and the microarchitecture layer. Consequently, conducting cross-layer workload characterization constitutes the initial stride in IoT design, especially in co-design. In this paper, we use a cross-layer profiling methodology to conduct an exhaustive analysis of IoTBench-an IoT workload benchmark. Each layer's key metrics, including instruction, data, and branch locality, were meticulously examined. Experimental evaluations were performed on both ARM and X86 architectures. Our findings revealed general patterns in how IoTBench's metrics fluctuate with different input data. Additionally, we noted that the same metrics could demonstrate varied characteristics across different layers, suggesting that isolated layer analysis might yield incomplete conclusions. Besides, our cross-layer profiling disclosed that the convolution task, characterized by deeply nested loops, significantly amplified branch locality at the microarchitecture layer on the ARM platform. Interestingly, optimization with the GNU C++ compiler (G++), intended to boost performance, had a counterproductive effect, exacerbating the branch locality issue and resulting in performance degradation.
GIFT is a family of lightweight block ciphers based on SPN structure and composed of two versions named GIFT-64 and GIFT-128. In this paper, we reevaluate the security of GIFT-64 against the rectangle attack under the...
ISBN:
(纸本)9783031533679;9783031533686
GIFT is a family of lightweight block ciphers based on SPN structure and composed of two versions named GIFT-64 and GIFT-128. In this paper, we reevaluate the security of GIFT-64 against the rectangle attack under the related-key setting. Investigating the previous rectangle key recovery attack on GIFT-64, we obtain the core idea of improving the attack-trading off the time complexity of each attack phase. We flexibly guess part of the involved subkey bits to balance the time cost of each phase so that the overall time complexity of the attack is reduced. Moreover, the reused subkey bits are identified according to the linear key schedule of GIFT-64 and bring additional advantages for our attacks. Furthermore, we incorporate the above ideas and propose a dedicated MILP model for finding the best rectangle key recovery attack on GIFT-64. As a result, we get the improved rectangle attacks on 26-round GIFT-64, which are the best attacks on it in terms of time complexity so far.
Table retrieval involves providing a ranked list of relevant tables in response to a search query. A critical aspect of this process is computing the similarity between tables. Recent Transformer-based language models...
ISBN:
(纸本)9783031560590;9783031560606
Table retrieval involves providing a ranked list of relevant tables in response to a search query. A critical aspect of this process is computing the similarity between tables. Recent Transformer-based language models have been effectively employed to generate word embedding representations of tables for assessing their semantic similarity. However, generating such representations for large tables comprising thousands or even millions of rows can be computationally intensive. This study presents the hypothesis that a significant portion of a table's content (i.e., rows) can be removed without substantially impacting its word embedding representation, thereby reducing computational costs while maintaining system performance. To test this hypothesis, two distinct evaluations were conducted. Firstly, an intrinsic evaluation was carried out using two different datasets and five state-of-the-art contextual and not-contextual language models. The findings indicate that, for large tables, retaining just 5% of the content results in a word embedding representation that is 90% similar to the original one. Secondly, an extrinsic evaluation was performed to assess how three different reduction techniques proposed affects the overall performance of the table-based query retrieval system, as measured by MAP, precision, and nDCG. The results demonstrate that these techniques can not only decrease data volume but also improve the performance of the table retrieval system.
Entertainment-oriented singing voice synthesis (SVS) requires a vocoder to generate high-fidelity (e.g. 48 kHz) audio. However, most text-to-speech (TTS) vocoders cannot reconstruct the waveform well in this scenario....
ISBN:
(纸本)9789819743988;9789819743995
Entertainment-oriented singing voice synthesis (SVS) requires a vocoder to generate high-fidelity (e.g. 48 kHz) audio. However, most text-to-speech (TTS) vocoders cannot reconstruct the waveform well in this scenario. In this paper, we propose HiFi-WaveGAN to synthesize the 48 kHz high-quality singing voices in real-time. Specifically, it consists of an Extended WaveNet that served as a generator, a multi-period discriminator proposed in HiFiGAN, and a multi-resolution spectrogram discriminator borrowed from UnivNet. To better reconstruct the high-frequency part from the full-band mel-spectrogram, we incorporate a pulse extractor to generate the constraint for the synthesized waveform. Additionally, an auxiliary spectrogram-phase loss is utilized to approximate the real distribution further. The experimental results (Demo page: https://***/hifi-wavegan/) show that our proposed HiFi-WaveGAN obtains 4.23 in the mean opinion score (MOS) metric for the 48 kHz SVS task, significantly outperforming other neural vocoders.
Tabulation is a well-known technique for improving the efficiency of recursive functions with redundant function calls. A key point in the application of this technique is to identify a suitable representation for the...
ISBN:
(纸本)9789819722990;9789819723003
Tabulation is a well-known technique for improving the efficiency of recursive functions with redundant function calls. A key point in the application of this technique is to identify a suitable representation for the table. In this paper, we propose the use of zippers as tables in the tabulation process. Our approach relies on a generic function zipWithZipper, that makes strong use of lazy evaluation to traverse two zippers in a circular manner. The technique turns out to be particularly efficient when the arguments to recursive calls are closely situated within the function domain. For example, in the case of natural numbers this means function calls on fairly contiguous values. Likewise, when dealing with tree structures, it means functions calls on immediate sub-trees and parent nodes. This results in a concise and efficient zipper-based embedding of attribute grammars.
暂无评论