Pub Date : 2018-03-01DOI: 10.1109/PDP2018.2018.00081
E. Doynikova, Igor Kotenko
The paper evolves an approach for proactive automatic cyber security incident response. The approach is based on usage of data from open sources, analytical modeling and a hierarchical integrated set of heterogeneous security metrics. The paper outlines the features of the analytical models that are crucial for countermeasure selection. It determines a set of security metrics for countermeasure selection. The algorithms that implement the suggested multi-layer countermeasure selection technique are specified. Introduction of the layers allows getting the result at any time with the maximum accuracy depending on the available data. The experiments that demonstrate the efficiency of the suggested technique are outlined.
{"title":"The Multi-Layer Graph Based Technique for Proactive Automatic Response Against Cyber Attacks","authors":"E. Doynikova, Igor Kotenko","doi":"10.1109/PDP2018.2018.00081","DOIUrl":"https://doi.org/10.1109/PDP2018.2018.00081","url":null,"abstract":"The paper evolves an approach for proactive automatic cyber security incident response. The approach is based on usage of data from open sources, analytical modeling and a hierarchical integrated set of heterogeneous security metrics. The paper outlines the features of the analytical models that are crucial for countermeasure selection. It determines a set of security metrics for countermeasure selection. The algorithms that implement the suggested multi-layer countermeasure selection technique are specified. Introduction of the layers allows getting the result at any time with the maximum accuracy depending on the available data. The experiments that demonstrate the efficiency of the suggested technique are outlined.","PeriodicalId":333367,"journal":{"name":"2018 26th Euromicro International Conference on Parallel, Distributed and Network-based Processing (PDP)","volume":"137 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116208705","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-03-01DOI: 10.1109/PDP2018.2018.00113
T. Dietrich, S. Bernuzzi, B. Brügmann, W. Tichy
The recent detection of gravitational waves and electromagnetic counterparts emitted during and after the collision of two neutron stars marks a breakthrough in the field of multi-messenger astronomy. Numerical relativity simulations are the only tool to describe the binary's merger dynamics in the regime when speeds are largest and gravity is strongest. In this work we report state-of-the-art binary neutron star simulations for irrotational (non-spinning) and spinning configurations. The main use of these simulations is to model the gravitational-wave signal. Key numerical requirements are the understanding of the convergence properties of the numerical data and a detailed error budget. The simulations have been performed on different HPC clusters, they use multiple grid resolutions, and are based on eccentricity reduced quasi-circular initial data. We obtain convergent waveforms with phase errors of 0.5-1.5rad accumulated over ~12 orbits to merger. The waveforms have been used for the construction of a phenomenological waveform model which has been applied for the analysis of the recent binary neutron star detection. Additionally, we show that the data can also be used to test other state-of-the-art semi-analytical waveform models.
{"title":"High-Resolution Numerical Relativity Simulations of Spinning Binary Neutron Star Mergers","authors":"T. Dietrich, S. Bernuzzi, B. Brügmann, W. Tichy","doi":"10.1109/PDP2018.2018.00113","DOIUrl":"https://doi.org/10.1109/PDP2018.2018.00113","url":null,"abstract":"The recent detection of gravitational waves and electromagnetic counterparts emitted during and after the collision of two neutron stars marks a breakthrough in the field of multi-messenger astronomy. Numerical relativity simulations are the only tool to describe the binary's merger dynamics in the regime when speeds are largest and gravity is strongest. In this work we report state-of-the-art binary neutron star simulations for irrotational (non-spinning) and spinning configurations. The main use of these simulations is to model the gravitational-wave signal. Key numerical requirements are the understanding of the convergence properties of the numerical data and a detailed error budget. The simulations have been performed on different HPC clusters, they use multiple grid resolutions, and are based on eccentricity reduced quasi-circular initial data. We obtain convergent waveforms with phase errors of 0.5-1.5rad accumulated over ~12 orbits to merger. The waveforms have been used for the construction of a phenomenological waveform model which has been applied for the analysis of the recent binary neutron star detection. Additionally, we show that the data can also be used to test other state-of-the-art semi-analytical waveform models.","PeriodicalId":333367,"journal":{"name":"2018 26th Euromicro International Conference on Parallel, Distributed and Network-based Processing (PDP)","volume":"4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127904667","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-03-01DOI: 10.1109/PDP2018.2018.00072
G. Speyer, Juan Rodriguez, T. Bencomo, Seungchan Kim
EDDY (Evaluation of Differential DependencY) interrogates transcriptomic data to identify differential genetic dependencies within a biological pathway. Through its probabilistic framework with resampling and permutation, aided by the incorporation of annotated gene sets, EDDY demonstrated superior sensitivity to other methods. However, this statistical rigor incurs considerable computational cost, limiting its application to larger datasets. The ample and independent computation coupled with manageable memory footprint positioned EDDY as a strong candidate for graphical processing unit (GPU) implementation. Custom kernels decompose the independence test loop, network construction, network enumeration, and Bayesian network scoring to accelerate the computation. GPU-accelerated EDDY consistently exhibits two orders of magnitude in performance enhancement, allowing the statistical rigor of the EDDY algorithm to be applied to larger datasets.
{"title":"GPU-Accelerated Differential Dependency Network Analysis","authors":"G. Speyer, Juan Rodriguez, T. Bencomo, Seungchan Kim","doi":"10.1109/PDP2018.2018.00072","DOIUrl":"https://doi.org/10.1109/PDP2018.2018.00072","url":null,"abstract":"EDDY (Evaluation of Differential DependencY) interrogates transcriptomic data to identify differential genetic dependencies within a biological pathway. Through its probabilistic framework with resampling and permutation, aided by the incorporation of annotated gene sets, EDDY demonstrated superior sensitivity to other methods. However, this statistical rigor incurs considerable computational cost, limiting its application to larger datasets. The ample and independent computation coupled with manageable memory footprint positioned EDDY as a strong candidate for graphical processing unit (GPU) implementation. Custom kernels decompose the independence test loop, network construction, network enumeration, and Bayesian network scoring to accelerate the computation. GPU-accelerated EDDY consistently exhibits two orders of magnitude in performance enhancement, allowing the statistical rigor of the EDDY algorithm to be applied to larger datasets.","PeriodicalId":333367,"journal":{"name":"2018 26th Euromicro International Conference on Parallel, Distributed and Network-based Processing (PDP)","volume":"10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125483686","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-03-01DOI: 10.1109/PDP2018.2018.00073
G. Romano, E. Genuardi, R. Calogero, S. Ferrero
In the last years, B/T cell clonality assessment and Minimal Residual Disease (MRD) monitoring acquired a strong prediction value in the therapy response evaluation of haematologic B disorders, improving patients outcome prediction. Polymerase Chain Reaction (PCR) based methods are the most standardized and widely used techniques, allowing a risk stratification in a variable proportion of patients, depending on the analyzed disease. Since its recently introduction, Next Generation Sequencing (NGS) technology could increase the number of patients with a traceable disease during the clinical course. This issue is strictly associated with an appropriate computational analysis of the huge volume of complex data obtained by NGS. In this context, recently, we presented an innovative bioinformatics approach, called HashClone, an easy-to-use and reliable bioinformatics tool that simultane- ously provides clonality assessment and MRD detection over time in patients affected by Mantle Cell Lymphoma (MCL). Actually, HashClone original strategy is organized in three steps that provide the simultaneous analysis of a set of samples reads returning to the corresponding clonotypes list, in which each clone is featured by frequency reads and aligned target nomenclature notification with respect to the reference database [1]. HashClone is composed by four C++ applications combined to implement B-cells clonality assessment in patient's samples. Since its successful preliminary application, in this paper, we present ParallelHashClone, an improved version with a parallel implementation of HashClone suite. In detail, the parallelization of this two applications allows to analyze more efficiently the samples from the same patient in parallel. Moreover we integrated ParallelHashClone in a Docker container platform that allows to easily install and run the application since the Docker packages ParallelHashClone with all its dependencies and libraries. We tested ParallelHashClone version for four MCL-NGS data analysis, showing comparable performances with respect to the original HashClone version in B-lymphoprolipherative molecular clonality assessment.
{"title":"ParallelHashClone: A Parallel Implementation of HashClone Suite for Clonality Assessment from NGS Data","authors":"G. Romano, E. Genuardi, R. Calogero, S. Ferrero","doi":"10.1109/PDP2018.2018.00073","DOIUrl":"https://doi.org/10.1109/PDP2018.2018.00073","url":null,"abstract":"In the last years, B/T cell clonality assessment and Minimal Residual Disease (MRD) monitoring acquired a strong prediction value in the therapy response evaluation of haematologic B disorders, improving patients outcome prediction. Polymerase Chain Reaction (PCR) based methods are the most standardized and widely used techniques, allowing a risk stratification in a variable proportion of patients, depending on the analyzed disease. Since its recently introduction, Next Generation Sequencing (NGS) technology could increase the number of patients with a traceable disease during the clinical course. This issue is strictly associated with an appropriate computational analysis of the huge volume of complex data obtained by NGS. In this context, recently, we presented an innovative bioinformatics approach, called HashClone, an easy-to-use and reliable bioinformatics tool that simultane- ously provides clonality assessment and MRD detection over time in patients affected by Mantle Cell Lymphoma (MCL). Actually, HashClone original strategy is organized in three steps that provide the simultaneous analysis of a set of samples reads returning to the corresponding clonotypes list, in which each clone is featured by frequency reads and aligned target nomenclature notification with respect to the reference database [1]. HashClone is composed by four C++ applications combined to implement B-cells clonality assessment in patient's samples. Since its successful preliminary application, in this paper, we present ParallelHashClone, an improved version with a parallel implementation of HashClone suite. In detail, the parallelization of this two applications allows to analyze more efficiently the samples from the same patient in parallel. Moreover we integrated ParallelHashClone in a Docker container platform that allows to easily install and run the application since the Docker packages ParallelHashClone with all its dependencies and libraries. We tested ParallelHashClone version for four MCL-NGS data analysis, showing comparable performances with respect to the original HashClone version in B-lymphoprolipherative molecular clonality assessment.","PeriodicalId":333367,"journal":{"name":"2018 26th Euromicro International Conference on Parallel, Distributed and Network-based Processing (PDP)","volume":"159 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116423187","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-03-01DOI: 10.1109/PDP2018.2018.00061
Francisco Muñoz-Martínez, M. Acacio
In this work, we address the challenge of designing an efficient warp scheduler for throughput processors by proposing SAWS (Simple and Adaptive Warp Scheduler). Differently from previous approaches which target a particular type of applications, SAWS considers several simple scheduling algorithms and tries to use the one that best fits each application or phase within an application. Through detailed simulations we demonstrate that a practical implementation of SAWS can obtain IPC values that closely match the best scheduling algorithm in each case.
{"title":"SAWS: Simple and Adaptive Warp Scheduling for Improved Performance in Throughput Processors","authors":"Francisco Muñoz-Martínez, M. Acacio","doi":"10.1109/PDP2018.2018.00061","DOIUrl":"https://doi.org/10.1109/PDP2018.2018.00061","url":null,"abstract":"In this work, we address the challenge of designing an efficient warp scheduler for throughput processors by proposing SAWS (Simple and Adaptive Warp Scheduler). Differently from previous approaches which target a particular type of applications, SAWS considers several simple scheduling algorithms and tries to use the one that best fits each application or phase within an application. Through detailed simulations we demonstrate that a practical implementation of SAWS can obtain IPC values that closely match the best scheduling algorithm in each case.","PeriodicalId":333367,"journal":{"name":"2018 26th Euromicro International Conference on Parallel, Distributed and Network-based Processing (PDP)","volume":"53 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132282101","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-03-01DOI: 10.1109/PDP2018.2018.00091
Anas Toma, Alexander Starinow, J. E. Lenssen, Jian-Jia Chen
In this paper, we present a middleware to save energy in mobile computing devices that offload tasks to a remote server in the cloud. Saving energy in these devices is very important to prolong the battery life and avoid overheating. The middleware uses an available nearby device called auxiliary server either as a surrogate for the remote one, or as a proxy to pass the data between the mobile device and the remote server. The main idea is to reduce the energy consumption of the communication with the remote server by using a high-speed or a low-power local connection with the auxiliary server instead. The paper also analyzes when it is beneficial to use the auxiliary server based on the response time from the remote server and the bandwidth of the remote connection. The proposed middleware is evaluated using different benchmarks, including commonly used applications in mobile devices, and simulations. Furthermore, it is compared to state-of-the art approaches in this area. The experiments show that The middleware is energy-efficient especially when the bandwidth of the remote communication is relatively low or the server is overloaded.
{"title":"Saving Energy for Cloud Applications in Mobile Devices Using Nearby Resources","authors":"Anas Toma, Alexander Starinow, J. E. Lenssen, Jian-Jia Chen","doi":"10.1109/PDP2018.2018.00091","DOIUrl":"https://doi.org/10.1109/PDP2018.2018.00091","url":null,"abstract":"In this paper, we present a middleware to save energy in mobile computing devices that offload tasks to a remote server in the cloud. Saving energy in these devices is very important to prolong the battery life and avoid overheating. The middleware uses an available nearby device called auxiliary server either as a surrogate for the remote one, or as a proxy to pass the data between the mobile device and the remote server. The main idea is to reduce the energy consumption of the communication with the remote server by using a high-speed or a low-power local connection with the auxiliary server instead. The paper also analyzes when it is beneficial to use the auxiliary server based on the response time from the remote server and the bandwidth of the remote connection. The proposed middleware is evaluated using different benchmarks, including commonly used applications in mobile devices, and simulations. Furthermore, it is compared to state-of-the art approaches in this area. The experiments show that The middleware is energy-efficient especially when the bandwidth of the remote communication is relatively low or the server is overloaded.","PeriodicalId":333367,"journal":{"name":"2018 26th Euromicro International Conference on Parallel, Distributed and Network-based Processing (PDP)","volume":"72 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114842985","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-03-01DOI: 10.1109/PDP2018.2018.00095
Yizhuo Wang, Zhiwei Gao, Weixing Ji, Han Zhang, Duzheng Qing
Today large-scale simulation applications are becoming common in research and industry. A significant fraction of them run on multi-core clusters. Current parallel simulation kernels use multi-process and multi-thread to exploit inter-node parallelism and intra-node parallelism on multi-core clusters. We exploit task-base parallelism in parallel discrete event simulation (PDES) kernels, which is more fine-grained than thread-level and process-level parallelism. In our system, every simulation event is wrapped to a task. Work-stealing task scheduling scheme is applied to achieve dynamic load balancing among the multi-cores, and a graph partitioning approach is applied in partitioning simulation entities among the cluster nodes. Experimental results show that our PDES kernel outperforms existing PDES kernels by fully exploiting task parallelism.
{"title":"Exploiting Task-Based Parallelism for Parallel Discrete Event Simulation","authors":"Yizhuo Wang, Zhiwei Gao, Weixing Ji, Han Zhang, Duzheng Qing","doi":"10.1109/PDP2018.2018.00095","DOIUrl":"https://doi.org/10.1109/PDP2018.2018.00095","url":null,"abstract":"Today large-scale simulation applications are becoming common in research and industry. A significant fraction of them run on multi-core clusters. Current parallel simulation kernels use multi-process and multi-thread to exploit inter-node parallelism and intra-node parallelism on multi-core clusters. We exploit task-base parallelism in parallel discrete event simulation (PDES) kernels, which is more fine-grained than thread-level and process-level parallelism. In our system, every simulation event is wrapped to a task. Work-stealing task scheduling scheme is applied to achieve dynamic load balancing among the multi-cores, and a graph partitioning approach is applied in partitioning simulation entities among the cluster nodes. Experimental results show that our PDES kernel outperforms existing PDES kernels by fully exploiting task parallelism.","PeriodicalId":333367,"journal":{"name":"2018 26th Euromicro International Conference on Parallel, Distributed and Network-based Processing (PDP)","volume":"30 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114745787","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-03-01DOI: 10.1109/PDP2018.2018.00084
Hergys Rexha, S. Lafond
In this paper we investigate the relation between energy efficiency model and workload type executed in modern embedded architectures. From the energy efficiency model obtained in our previous work we select a few configuration points to verify that the prediction in terms of relative energy efficiency is maintained through different workload scenarios. A configuration point is defined as a set of platform tunable metrics, such as DVFS point, DPM level and utilization rate. As workloads, we use a combination of synthetic generators and real world applications from the embedded domain. In our experiments we use two different architectures for testing the model generality, which provide examples of real systems. First we have a comparison of the efficiency obtained by the two architecturally different chips (ARM and INTEL) in different configuration points and different workload scenarios. Second we try to explain the different results through the thermal management done by the two different chips. At the end we show that only in the case of workloads highly composed by integer instructions the results from the two architectures converge and show the need for a specific model trained with integer operations.
{"title":"Exploring Energy Efficiency Model Generalization on Multicore Embedded Platforms","authors":"Hergys Rexha, S. Lafond","doi":"10.1109/PDP2018.2018.00084","DOIUrl":"https://doi.org/10.1109/PDP2018.2018.00084","url":null,"abstract":"In this paper we investigate the relation between energy efficiency model and workload type executed in modern embedded architectures. From the energy efficiency model obtained in our previous work we select a few configuration points to verify that the prediction in terms of relative energy efficiency is maintained through different workload scenarios. A configuration point is defined as a set of platform tunable metrics, such as DVFS point, DPM level and utilization rate. As workloads, we use a combination of synthetic generators and real world applications from the embedded domain. In our experiments we use two different architectures for testing the model generality, which provide examples of real systems. First we have a comparison of the efficiency obtained by the two architecturally different chips (ARM and INTEL) in different configuration points and different workload scenarios. Second we try to explain the different results through the thermal management done by the two different chips. At the end we show that only in the case of workloads highly composed by integer instructions the results from the two architectures converge and show the need for a specific model trained with integer operations.","PeriodicalId":333367,"journal":{"name":"2018 26th Euromicro International Conference on Parallel, Distributed and Network-based Processing (PDP)","volume":"15 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127649407","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-03-01DOI: 10.1109/PDP2018.2018.00075
M. Hamidi, M. Alishahi, F. Martinelli
This paper presents a general framework for constructing any agglomerative hierarchical clustering algorithm over partitioned data. It is assumed that data is distributed between two (or more) parties horizontally, such that for mutual benefits the participated parties are willing to identify the clusters' structure on their data as a whole, but for privacy restrictions, they avoid to share the original datasets. To this end, in this study, we propose general algorithms based on secure scalar product and secure hamming distance computation to securely compute the desired criteria for shaping the clusters' scheme. The proposed approach covers all possible secure agglomerative hierarchical clustering construction when data is distributed between two (or more) parties, including both numerical and categorical data.
{"title":"A Secure Distributed Framework for Agglomerative Hierarchical Clustering Construction","authors":"M. Hamidi, M. Alishahi, F. Martinelli","doi":"10.1109/PDP2018.2018.00075","DOIUrl":"https://doi.org/10.1109/PDP2018.2018.00075","url":null,"abstract":"This paper presents a general framework for constructing any agglomerative hierarchical clustering algorithm over partitioned data. It is assumed that data is distributed between two (or more) parties horizontally, such that for mutual benefits the participated parties are willing to identify the clusters' structure on their data as a whole, but for privacy restrictions, they avoid to share the original datasets. To this end, in this study, we propose general algorithms based on secure scalar product and secure hamming distance computation to securely compute the desired criteria for shaping the clusters' scheme. The proposed approach covers all possible secure agglomerative hierarchical clustering construction when data is distributed between two (or more) parties, including both numerical and categorical data.","PeriodicalId":333367,"journal":{"name":"2018 26th Euromicro International Conference on Parallel, Distributed and Network-based Processing (PDP)","volume":"20 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132324048","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-03-01DOI: 10.1109/PDP2018.2018.00058
M. Serpa, Arthur M. Krause, E. Cruz, P. Navaux, Marcelo Pasin, P. Felber
Driven by the development of new technologies such as personal assistants or autonomous cars, machine learning has rapidly become one of the most active fields in computer science. The algorithms at the core of machine learning are notoriously demanding in terms of resources. It is therefore of paramount importance to optimize their operation on modern processors. Several approaches have been proposed to accelerate machine learning on GPUs and massively parallel computers, as well as dedicated ASICs. In this paper, we focus on Intel's multi-core Xeon and many-core accelerator Xeon Phi Knights Landing, which can host several hundreds of threads on the same CPU. In such architectures, thread and data mapping are keys for performance. We study the impact of mapping strategies, revealing that, with smart mapping policies, one can indeed significantly speed up machine learning applications on many-core architectures. Execution time was reduced by up to 25.2% and 18.5% on Intel Xeon and Xeon Phi KNL, respectively.
{"title":"Optimizing Machine Learning Algorithms on Multi-Core and Many-Core Architectures Using Thread and Data Mapping","authors":"M. Serpa, Arthur M. Krause, E. Cruz, P. Navaux, Marcelo Pasin, P. Felber","doi":"10.1109/PDP2018.2018.00058","DOIUrl":"https://doi.org/10.1109/PDP2018.2018.00058","url":null,"abstract":"Driven by the development of new technologies such as personal assistants or autonomous cars, machine learning has rapidly become one of the most active fields in computer science. The algorithms at the core of machine learning are notoriously demanding in terms of resources. It is therefore of paramount importance to optimize their operation on modern processors. Several approaches have been proposed to accelerate machine learning on GPUs and massively parallel computers, as well as dedicated ASICs. In this paper, we focus on Intel's multi-core Xeon and many-core accelerator Xeon Phi Knights Landing, which can host several hundreds of threads on the same CPU. In such architectures, thread and data mapping are keys for performance. We study the impact of mapping strategies, revealing that, with smart mapping policies, one can indeed significantly speed up machine learning applications on many-core architectures. Execution time was reduced by up to 25.2% and 18.5% on Intel Xeon and Xeon Phi KNL, respectively.","PeriodicalId":333367,"journal":{"name":"2018 26th Euromicro International Conference on Parallel, Distributed and Network-based Processing (PDP)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130042157","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：