{"title":"Using the Integrated GPU to Improve CPU Sort Performance","authors":"Grigore Lupescu, E. Slusanschi, N. Tapus","doi":"10.1109/ICPPW.2017.19","DOIUrl":"https://doi.org/10.1109/ICPPW.2017.19","url":null,"abstract":"In this paper we discuss the potential of the integrated GPU to accelerate sorting by performing a partial sort prior to a comparison based CPU sort. We experiment along with several CPU comparison based sorting algorithms and outline the performance gain for a random input data set. We then analyze different x86 SoC architectures, and show that by sorting chunks stored inside the onchip GPU memory, we can almost eliminate the impact the memory hierarchy has on performance. Finally, we discuss how our approach is different from previous designs, being specifically tailored for an SoC with an integrated GPU, able to improve on most of its known limitations (memory bandwidth, complexity, performance etc).","PeriodicalId":353142,"journal":{"name":"2017 46th International Conference on Parallel Processing Workshops (ICPPW)","volume":"27 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122449944","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}
{"title":"An Efficient Filter Strategy for Theta-Join Query in Distributed Environment","authors":"Wenjie Liu, Zhanhuai Li, Yuntao Zhou","doi":"10.1109/ICPPW.2017.24","DOIUrl":"https://doi.org/10.1109/ICPPW.2017.24","url":null,"abstract":"Theta-join query is a very popular application in traditional databases, but due to tremendous computation cost and communication cost in distributed environment, it is not efficiently processed for big data. Current researches focus on processing theta-join by using MapReduce framework, which mainly consider the overheads of load balance in the network, when the data sets become larger, massive intermediate results lead to high communication cost. In this work, we propose a filter method for theta-join to reduce the computation and communication cost in distributed environment, which can effectively improve the theta-join query. We consider both the load balance in the cluster and the memory cost in the parallel framework. We have implemented our method in a popular general-purpose data processing framework, Spark. The experimental results demonstrate that our method can significantly improve the performance of theta-joins comparing the state-of-art solutions.","PeriodicalId":353142,"journal":{"name":"2017 46th International Conference on Parallel Processing Workshops (ICPPW)","volume":"75 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123835120","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}
Harald Servat, Jesús Labarta, Hans-Christian Hoppe, Judit Giménez, Antonio J. Peña
{"title":"Integrating Memory Perspective into the BSC Performance Tools","authors":"Harald Servat, Jesús Labarta, Hans-Christian Hoppe, Judit Giménez, Antonio J. Peña","doi":"10.1109/ICPPW.2017.42","DOIUrl":"https://doi.org/10.1109/ICPPW.2017.42","url":null,"abstract":"The growing gap between processor and memory speeds results in complex memory hierarchies as processors evolve to mitigate such differences by taking advantage of locality of reference. In this direction, the BSC performance analysis tools have been recently extended to provide insight relative the application memory accesses depicting their temporal and spatial characteristics, correlating with the source-code and the achieved performance simultaneously. These extensions rely on the Precise Event-Based Sampling (PEBS) mechanism available in recent Intel processors to capture information relative to the application memory accesses. The sampled information is processed with the Folding mechanism to provide a detailed temporal evolution of the memory accesses and in conjunction with the achieved performance and the source-code counterpart. The results obtained from the combination of these tools help application developers to understand better how the application behaves and how the system performs. We demonstrate the value of the complete work-flow by exploring an already optimized state-of-the-art benchmark, providing detailed insight of their memory access behavior.","PeriodicalId":353142,"journal":{"name":"2017 46th International Conference on Parallel Processing Workshops (ICPPW)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131132167","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}
{"title":"Turning GPUs into Floating Devices over the Cluster: The Beauty of GPU Migration","authors":"Javier Prades, F. Silla","doi":"10.1109/ICPPW.2017.30","DOIUrl":"https://doi.org/10.1109/ICPPW.2017.30","url":null,"abstract":"Virtualization techniques have shown to report benefits to data centers and other computing facilities. In this regard, not only virtual machines allow reducing the size of the computing infrastructure while increasing overall resource utilization but also virtualizing individual components of computers may provide significant benefits. This is the case, for example, for the remote GPU virtualization technique, implemented in several frameworks during the recent years. The large degree of flexibility provided by the remote GPU virtualization technique can, however, be further increased by applying the migration mechanism to it, so that the GPU part of applications can be live migrated to another GPU elsewhere in the cluster during execution time in a transparent way. In this paper we present a discussion about how the migration mechanism has been applied to different GPU virtualization frameworks. We also provide a big picture about the possibilities that migrating the GPU part of applications can provide to data centers and other computing facilities. We finally present the first results of an ongoing work consisting on applying the migration mechanism to the rCUDA remote GPU virtualization framework.","PeriodicalId":353142,"journal":{"name":"2017 46th International Conference on Parallel Processing Workshops (ICPPW)","volume":"51 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133572755","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}
{"title":"Exploiting Key-Value Data Stores Scalability for HPC","authors":"C. Cugnasco, Y. Becerra, J. Torres, E. Ayguadé","doi":"10.1109/ICPPW.2017.25","DOIUrl":"https://doi.org/10.1109/ICPPW.2017.25","url":null,"abstract":"BigData revolutionised the IT industry. It first interested the OLTP systems. Distributed Hash Tables replaced Traditional SQL databases as they guaranteed low response time on simple read/write requests. The second wave recast the data warehousing: map-reduce systems spread as they proved to scale linearly long-running computational workloads on commodity servers. The focus now is on real-time analytics. Being able to analyse massive quantities of data in a short time enables multiple HPC applications and interactive analysis and visualization. In this paper, we study the performance of a system that employs the DHT architecture to achieve fast in local analysis on indexed data. We observed that the number of keys, nodes, and the hardware characteristics strongly influence the actual scalability of the system. Therefore, we developed a mathematical model that allows finding the right system configuration to meet desired performance for each kind of query type. We also show how our model can be used to find the right architecture for each distributed application.","PeriodicalId":353142,"journal":{"name":"2017 46th International Conference on Parallel Processing Workshops (ICPPW)","volume":"97 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134576601","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}
{"title":"On a Storage System Software Stack for Extreme Scale Data Centric Computing","authors":"Sai B. Narasimhamurthy","doi":"10.1109/ICPPW.2017.41","DOIUrl":"https://doi.org/10.1109/ICPPW.2017.41","url":null,"abstract":"This paper presents an overview of a storage system software stack designed specifically for extreme scale data centric computing. The storage system software stack is designed to work for next generation storage system architectures that will need to incorporate multiple data storage device technologies. We also envision such data centric storage systems to have in-storage compute capability to make it suitable for dealing with data analytics, and, pre/post processing steps in data centric workflows. The storage system software stack indeed needs to accommodate such capability.","PeriodicalId":353142,"journal":{"name":"2017 46th International Conference on Parallel Processing Workshops (ICPPW)","volume":"275 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116069608","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}
{"title":"SWAS: Stealing Work Using Approximate System-Load Information","authors":"S. Tzilis, M. Pericàs, P. Trancoso, I. Sourdis","doi":"10.1109/ICPPW.2017.51","DOIUrl":"https://doi.org/10.1109/ICPPW.2017.51","url":null,"abstract":"This paper explores the potential of utilizing approximate system load information to enhance work stealing for dynamic load balancing in hierarchical multicore systems. Maintaining information about the load of a system has not been extensively researched since it is assumed to introduce performance overheads. We propose SWAS, a lightweight approximate scheme for retrieving and using such information, based on compact bit vector structures and lightweight update operations. This approximate information is used to enhance the effectiveness of work stealing decisions. Evaluating SWAS for a number of representative scenarios on a multi-socket multi-core platform showed that work stealing guided by approximate system load information achieves considerable performance improvements: up to 18.5% for dynamic, severely imbalanced workloads; and up to 34.4% for workloads with complex task dependencies, when compared with random work stealing.","PeriodicalId":353142,"journal":{"name":"2017 46th International Conference on Parallel Processing Workshops (ICPPW)","volume":"13 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127581780","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}
{"title":"Hierarchical Read/Write Analysis for Pointer-Based OpenCL Programs on RRAM","authors":"Lin-Ya Yu, Shao-Chung Wang, Jenq-Kuen Lee","doi":"10.1109/ICPPW.2017.20","DOIUrl":"https://doi.org/10.1109/ICPPW.2017.20","url":null,"abstract":"Heterogeneous computing platforms containing a wide range of computing resources from CPUs to specialized hardware accelerators is the trend today resulting from the physical limitations on processors speed and the increasing demand for computing performance. Hence many optimization strategies are studied to get better throughput and lower energy consumption in heterogeneous systems. Various memory technologies such as DRAM, STT-RAM, and RRAM are also developed to help reach the goal. Meanwhile OpenCL is an open programming language standard for programmers to write programs on these hardware accelerators in a heterogeneous system. In this paper, a new static analysis technique, based on OpenCL programming languages can determine read/write characteristics of instances within the program is presented. We consider the computing configuration with both DRAM and RRAM. We try to answer the question of which variables stored on RRAM will benefit energy efficiency with compiler analysis. Our compiler scheme based on Memory SSA enables the analysis to cover pointer-based programs. Our evaluation demonstrates that the read/write information obtained from the proposed design has great potential to achieve high energy-savings. The average error distance of our proposed scheme is 0.3548. In addition, compared to the baseline system, the read/write information help gain average 37.06% energy-saving per kernel.","PeriodicalId":353142,"journal":{"name":"2017 46th International Conference on Parallel Processing Workshops (ICPPW)","volume":"26 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121103924","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}
{"title":"OpenMP Memkind: An Extension for Heterogeneous Physical Memories","authors":"Xi Wang, John D. Leidel, Yong Chen","doi":"10.1109/ICPPW.2017.40","DOIUrl":"https://doi.org/10.1109/ICPPW.2017.40","url":null,"abstract":"Recently, CPU and graphics processors have been increasing the degree of on-chip parallelism in order to combat the decrease in traditional Moore's Law scaling. As a result, these new processors are increasing their appetite for faster memory devices with higher bandwidth. Component manufacturers have resorted to disparate or hierarchical fast memory device architectures such as shared local memory (SLM), scratch pad memory (SPM), and high bandwidth memory (HBM) to provide sufficient bandwidth. Following this trend, the physical memory locality gradually becomes a performance feature that users would like to explicitly manage. Inspired by this idea, this research is conducted to create a heterogeneous memory interface based on a new declarative data storage directive, or \"memkind\", for the OpenMP parallel programming specification to explicitly manage physical memory locality. Our approach is implemented as an OpenMP directive in order to avoid allocating data inside parallel regions, thus avoiding performance degradation due to sequential operating system routines. We demonstrate our approach as an extension to the LLVM OpenMP implementation, that enables the portability of our approach to be rapidly ported to any LLVM-supported architecture target. Our contributions in this work are a detailed design analysis of the memkind directive as well as a detailed implementation in the LLVM compiler infrastructure. We demonstrate the efficacy of our approach using a synthetic benchmark application that records the execution performance and memory allocation efficiency.","PeriodicalId":353142,"journal":{"name":"2017 46th International Conference on Parallel Processing Workshops (ICPPW)","volume":"49 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121463011","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}
Ghaith Tarawneh, A. Mokhov, Matthew Naylor, Alexander D. Rast, S. Moore, David B. Thomas, A. Yakovlev, Andrew D. Brown
{"title":"Programming Model to Develop Supercomputer Combinatorial Solvers","authors":"Ghaith Tarawneh, A. Mokhov, Matthew Naylor, Alexander D. Rast, S. Moore, David B. Thomas, A. Yakovlev, Andrew D. Brown","doi":"10.1109/ICPPW.2017.35","DOIUrl":"https://doi.org/10.1109/ICPPW.2017.35","url":null,"abstract":"Novel architectures for massively parallel machines offer better scalability and the prospect of achieving linear speedup for sizable problems in many domains. The development of suitable programming models and accompanying software tools for these architectures remains one of the biggest challenges towards exploiting their full potential. We present a multi-layer software abstraction model to develop combinatorial solvers on massively-parallel machines with regular topologies. The model enables different challenges in the design and optimization of combinatorial solvers to be tackled independently (separation of concerns) while permitting problem-specific tuning and cross-layer optimization. In specific, the model decouples the issues of inter-node communication, node-level scheduling, problem mapping, mesh-level load balancing and expressing problem logic. We present an implementation of the model and use it to profile a Boolean satisfiability solver on simulated massively-parallel machines with different scales and topologies.","PeriodicalId":353142,"journal":{"name":"2017 46th International Conference on Parallel Processing Workshops (ICPPW)","volume":"223 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124429774","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}