{"title":"Topology-aware equipartitioning with coscheduling on multicore systems","authors":"Jan H. Schonherr, B. Juurlink, Jan Richling","doi":"10.1109/MUCOCOS.2013.6633602","DOIUrl":"https://doi.org/10.1109/MUCOCOS.2013.6633602","url":null,"abstract":"Over the last decade, multicore architectures have become omnipresent. Today, they are used in the whole product range from server systems to handheld computers. The deployed software still undergoes the slow transition from sequential to parallel. This transition, however, is gaining more and more momentum due to the increased availability of more sophisticated parallel programming environments, which replace the some-times crude results of ad-hoc parallelization. Combined with the ever increasing complexity of multicore architectures, this results in a scheduling problem that is different from what it has been, because features such as non-uniform memory access, shared caches, or simultaneous multithreading have to be considered. In this paper, we compare different ways of scheduling multiple parallel applications. Due to emerging parallel programming environments, we only consider malleable applications, i. e., applications where the parallelism degree can be changed on the fly. We propose a topology-aware scheduling scheme that combines equipartitioning and coscheduling. It does not suffer from the drawbacks of the individual concepts and also allows to run applications at different degrees of parallelisms without compromising fairness. We find that topology-awareness increases performance for all evaluated workloads. The combination with coscheduling is more sensitive towards the executed workloads. However, the gained versatility allows new use cases to be explored, which were not possible before.","PeriodicalId":397966,"journal":{"name":"2013 IEEE 6th International Workshop on Multi-/Many-core Computing Systems (MuCoCoS)","volume":"63 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-10-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125898799","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":"Automatic extraction of multi-objective aware parallelism for heterogeneous MPSoCs","authors":"D. Cordes, M. Engel, Olaf Neugebauer, P. Marwedel","doi":"10.1109/MUCOCOS.2013.6633599","DOIUrl":"https://doi.org/10.1109/MUCOCOS.2013.6633599","url":null,"abstract":"Heterogeneous MPSoCs are used in a large fraction of current embedded systems. In order to efficiently exploit the available processing power, advanced parallelization techniques are required. In addition to consider performance variances between heterogeneous cores, these methods have to be multi-objective aware to be useful for resource restricted embedded systems. This multi-objective optimization requirement results in an explosion of the design space size. As a consequence, efficient approaches are required to find promising solution candidates. In this paper, we present the first portable genetic algorithm-based approach to speed up ANSI-C applications by combining extraction techniques for task-level and pipeline parallelism for heterogeneous multicores while considering additional objectives. Using our approach enables embedded system designers to select a parallelization of an application from a set of Pareto-optimal solutions according to the performance and energy consumption requirements of a given system. The evaluation of a large set of typical embedded benchmarks shows that our approach is able to generate solutions with low energy consumption, high speedup, low communication overhead or useful trade-offs between these three objectives.","PeriodicalId":397966,"journal":{"name":"2013 IEEE 6th International Workshop on Multi-/Many-core Computing Systems (MuCoCoS)","volume":"22 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-10-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125173430","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":"Optimizing sparse matrix vector multiplication on emerging multicores","authors":"Orhan Kislal, W. Ding, M. Kandemir, I. Demirkiran","doi":"10.1109/MUCOCOS.2013.6633600","DOIUrl":"https://doi.org/10.1109/MUCOCOS.2013.6633600","url":null,"abstract":"After hitting the power wall, the dramatic change in computer architecture from single core to multicore/manycore brings us new challenges on high performance computing, especially for the data intensive applications. Sparse matrix-vector multiplication (SpMV) is one of the most important computations in this area, and has therefore received a lot of attention in recent decades. In contrast to the uniform/regular dense matrix computations, SpMV's irregular data access patterns with compact data structure for storage make the SpMV optimization more complex than optimizing regular/dense matrix computation. In this work, we look at the SpMV optimization problem in the context of emerging multicores from a different architecture conscious perspective, and propose an optimization strategy that has three key components: mapping, scheduling and data layout reorganization. Specifically, the mapping component derives a suitable iteration-to-core mapping; the scheduling component determines the execution order of loop iterations assigned to each core in the target multicore architecture; and finally, the data layout reorganization component prepares multiple memory layouts for the source (input) vector customized for different row patterns. A distinguishing characteristic of our approach is that it is cache hierarchy aware, that is, all three components take the underlying cache hierarchy of the target multicore architecture into account, and therefore, the derived solution is, in a sense, customized to the target architecture. We evaluate the proposed strategy using 10 sparse matrices with two different multicore systems. Our experimental evaluation reveals that the proposed optimization algorithm brings significant performance improvements (up to 26.5%) over the unoptimized case.","PeriodicalId":397966,"journal":{"name":"2013 IEEE 6th International Workshop on Multi-/Many-core Computing Systems (MuCoCoS)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-10-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133127758","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":"One OpenCL to rule them all?","authors":"R. Dolbeau, F. Bodin, Guillaume Colin de Verdiere","doi":"10.1109/MUCOCOS.2013.6633603","DOIUrl":"https://doi.org/10.1109/MUCOCOS.2013.6633603","url":null,"abstract":"OpenCL is now available on a very large set of processors. This makes this language an attractive layer to address multiple targets with a single code base. The question on how sensitive to the underlying hardware is the OpenCL code in practice remains to be better understood. 1. This paper studies how realistic it is to use a unique OpenCL code for a set of hardware co-processors with different underlying micro-architectures. In this work, we target Intel® Xeon Phi™, NVIDIA® K20C and AMD® 7970™ GPU. All these accelerators provide at least support for OpenCL 1.1 and fit in the same high-end version of accelerator technology. To assess performance, we use OpenACC CAPS compiler to generate OpenCL code and use a moderately complex miniapplication, Hydro. This code uses 22 OpenCL kernels and was tailored to limit data transfers between the host and the accelerator device. To study how stable are the performance, we performed many experiments to determine the best OpenCL code for each hardware platform. This paper shows that, if well chosen, a single version of the code can be executed on multiple platforms without significant performance losses (less than 12%). This study confirms the need for auto-tuning technology to look for performance tradeoffs but also shows that deploying selftuning/adaptive code is not always necessary if the ultimate performance is not the goal.","PeriodicalId":397966,"journal":{"name":"2013 IEEE 6th International Workshop on Multi-/Many-core Computing Systems (MuCoCoS)","volume":"2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-10-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131737533","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}
Georgios Chatzopoulos, K. Kourtis, N. Koziris, G. Goumas
{"title":"Towards a compiler/runtime synergy to predict the scalability of parallel loops","authors":"Georgios Chatzopoulos, K. Kourtis, N. Koziris, G. Goumas","doi":"10.1109/MUCOCOS.2013.6633605","DOIUrl":"https://doi.org/10.1109/MUCOCOS.2013.6633605","url":null,"abstract":"Large classes of applications fail to scale well in CMPs due to contention in the memory subsystem. Assigning full core capacity to such applications is a clear resource waste. To support efficient and power-aware resource allocation policies, we need a prediction mechanism to provide information about the potential scalability of an application. In this paper we take an initial step towards building a scalability predictor, based on the utilization of information collected both during compile and runtime. Our approach is applied separately to each parallel-for region in the program and calculates an on-chip to off-chip activity ratio Sr, which then is associated to the scalability of the region (maximum speedup) with linear regression. Experimental results on two architectures using the Polybench suite demonstrate that our prediction model exhibits a good accuracy in predicting the scalability of various parallel-for regions.","PeriodicalId":397966,"journal":{"name":"2013 IEEE 6th International Workshop on Multi-/Many-core Computing Systems (MuCoCoS)","volume":"66 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-10-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133198371","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":"Algorithmic species revisited: A program code classification based on array references","authors":"C. Nugteren, R. Corvino, H. Corporaal","doi":"10.1109/MUCOCOS.2013.6633604","DOIUrl":"https://doi.org/10.1109/MUCOCOS.2013.6633604","url":null,"abstract":"The shift towards parallel processor architectures has made programming, performance prediction and code generation increasingly challenging. Abstract representations of program code (i.e. classifications) have been introduced to address this challenge. An example is `algorithmic species', a memory access pattern classification of loop nests. It provides an architecture-agnostic structured view of program code, allowing programmers and compilers to take for example parallelisation decisions or perform memory hierarchy optimisations. The existing algorithmic species theory is based on the polyhedral model and is limited to static affine loop nests. In this work, we first present a revised theory of algorithmic species that overcomes this limitation. The theory consists of a 5-tuple characterisation of individual array references and their corresponding merging operation. Second, we present an extension of this theory named SPECIES+, providing a more detailed 6-tuple characterisation. With this, we are able to retain relevant access pattern information not captured by the original algorithmic species, such as column-major versus row-major matrix accesses. We implement both new theories into a tool, enabling automatic classification of program code.","PeriodicalId":397966,"journal":{"name":"2013 IEEE 6th International Workshop on Multi-/Many-core Computing Systems (MuCoCoS)","volume":"8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-10-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123885621","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":"Quantifying the performance impacts of using local memory for many-core processors","authors":"Jianbin Fang, H. Sips, A. Varbanescu","doi":"10.1109/MUCOCOS.2013.6633601","DOIUrl":"https://doi.org/10.1109/MUCOCOS.2013.6633601","url":null,"abstract":"Due to the increasing complexity of multi/many-core architectures (with their mix of caches and scratch-pad memories) and applications (with different memory access patterns), the performance of many workloads becomes increasingly variable. In this work, we address one of the main causes for this performance variability: the efficiency of the memory system. Specifically, based on an empirical evaluation driven by memory access patterns, we qualify and partially quantify the performance impact of using local memory in multi/many-core processors. To do so, we systematically describe memory access patterns (MAPs) in an application-agnostic manner. Next, for each identified MAP, we use OpenCL (for portability reasons) to generate two microbenchmarks: a “naive” version (without local memory) and “an optimized” version (using local memory). We further evaluate both of them on typically used multi-core and many-core platforms, and we log their performance. What we eventually obtain is a local memory performance database, indexed by various MAPs and platforms. To show how our database can be used to provide information on local memory usage, we demonstrate a scenario for performance prediction: given an application, its MAPs, and a platform, a query in the database can indicate the performance impact of using local memory for the given application. This indication can be used to either avoid the hassle of implementing optimizations with too little gain or, alternatively, give a rough prediction of the performance gain.","PeriodicalId":397966,"journal":{"name":"2013 IEEE 6th International Workshop on Multi-/Many-core Computing Systems (MuCoCoS)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-10-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130210638","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}
Lars F. Bonnichsen, Sven Karlsson, Christian W. Probst
{"title":"ELB-trees an efficient and lock-free B-tree derivative","authors":"Lars F. Bonnichsen, Sven Karlsson, Christian W. Probst","doi":"10.1109/MUCOCOS.2013.6633606","DOIUrl":"https://doi.org/10.1109/MUCOCOS.2013.6633606","url":null,"abstract":"As computer systems scale in the number of processors, scalable data structures with good parallel performance become increasingly important. Lock-free data structures promise such improved parallel performance at the expense of higher algorithmic complexity and higher sequential execution time overhead. All lock-free data structures are based on simple atomic operations that, though supported by modern processors, are expensive in execution time. We present a lock-free data structure, ELB-trees, which under certain assumptions can be used as multimaps as well as priority queues. Specifically it cannot store duplicate key-value pairs, and it is not linearizable. Compared to existing data structures, ELB-trees require fewer atomic operations leading to improved performance. We measure the parallel performance of ELB-trees using a set of benchmarks and observe that ELB-trees are up to almost 30 times faster than library multimap implementations.","PeriodicalId":397966,"journal":{"name":"2013 IEEE 6th International Workshop on Multi-/Many-core Computing Systems (MuCoCoS)","volume":"47 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-08-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130720665","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}