{"title":"A study on the performance of two-level exclusive caching","authors":"E. Musoll, M. Nemirovsky","doi":"10.5753/sbac-pad.1999.19771","DOIUrl":"https://doi.org/10.5753/sbac-pad.1999.19771","url":null,"abstract":"This work presents a study on the performance of a level-two cache configured as a victim storage for the evicted lines of the level-one cache. This two-level cache configuration, known as exclusive caching, is evaluated for a wide range of level-one and level-two sizes and associalivily degrees, and the miss ratios of both levels are compared to those of the two-level inclusive caching. Although the two-level exclusive strategy has lower miss ratios than the inclusive one by increasing the effective associativity and capacity, the replacement policy of the exclusive caching organization forces the invalidation of cnlries in the level-two cache, which reduces the benefits of having a victim level-two cache. The effect of these invalidalions on the overall performance of a level-two exclusive caching organization is evaluated. For typical two-level cache configurations in which the level-two cache is direct-mapped, the performance of the exclusive caching is as much as 60% better for code fetches and as much as 75% for data accesses.","PeriodicalId":267202,"journal":{"name":"Anais do XI International Symposium on Computer Architecture and High Performance Computing (SBAC-PAD 1999)","volume":"69 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122849706","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":"A Survey of Load Balancers in Modern Multi-Threading Systems","authors":"P. Kakulavarapu, J. N. Amaral","doi":"10.5753/sbac-pad.1999.19766","DOIUrl":"https://doi.org/10.5753/sbac-pad.1999.19766","url":null,"abstract":"Multithreaded architectures are a feasible approach to exploit parallelism in both regular and irregular applications. Today a large collection of multi-threading architectures with different threaded models, and implementation platforms are available. These architectures provide support for multithreading either at hardware level, with customized functional units, or at the software level, as emulators written in some high-level language. The later approach is usually prefered because of its favorable price tag, speed of development, and portability. In this article we review some of these architectures focusing in their capabilities to provide load balancing for irregular, data-parallel and recursive applications. The paper is anchored on a description of our own implementation of load balancers for EARTH - Efficient Architecture for Running Threads. Most of the multithreading architectures that we review are software emulations based on off-the-shelf hardware and compiler technologies.","PeriodicalId":267202,"journal":{"name":"Anais do XI International Symposium on Computer Architecture and High Performance Computing (SBAC-PAD 1999)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129279271","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":"Investigating the Relative Performance of Static and Dynamic Instruction Scheduling","authors":"D. Tate, G. Steven, P. Findlay","doi":"10.5753/sbac-pad.1999.19788","DOIUrl":"https://doi.org/10.5753/sbac-pad.1999.19788","url":null,"abstract":"There are two distinct groups of research into ILP. Those that strongly favour static instruction scheduling and those that favour dynamic instruction scheduling. This paper introduces powerful static and dynamic scheduling models and combines them within the framework of a single simulation environment. Both individual models achieve respectable speedups; dynamic schedullng significantly out-performs static scheduling when an idealised processor model with perfect branch prediction is used. However, when a realistic branch predictor is substituted, the roles are reversed, and static scheduling achieves the higher performance. Similarly, static scheduling performs better in the absence of branch prediction or when processor resources are restricted. Finally, we combine static scheduling with out-of-order instruction issue. Disappointingly, when an ideal out-of-order processor is used, scheduled code fails to match the performance of unscheduled code. Furthermore, with realistic branch predictlon, out-of-order issue fails to improve the performance of scheduled code.","PeriodicalId":267202,"journal":{"name":"Anais do XI International Symposium on Computer Architecture and High Performance Computing (SBAC-PAD 1999)","volume":"32 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126977138","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":"Using Operand Factorization to Compress DSP Programs","authors":"Ricardo Pannain, P. Centoducatte, G. Araújo","doi":"10.5753/sbac-pad.1999.19793","DOIUrl":"https://doi.org/10.5753/sbac-pad.1999.19793","url":null,"abstract":"We propose a method for compressing programs running on embedded DSPs. Program expression trees are decomposed into opcode and operand sequences called patterns. We show that DSP program patterns have exponential frequency distribution. Based on that, we encode patterns using a mix of a variable-length and fixed-length codewords. A decompression engine is proposed, which converts patterns into uncompressed instruction sequences. The experimental results reveal an average compression ratio of 67% for typical DSP programs running on TMS320C25 processor. This ratio includes an estimate of the size of the decompression engine.","PeriodicalId":267202,"journal":{"name":"Anais do XI International Symposium on Computer Architecture and High Performance Computing (SBAC-PAD 1999)","volume":"60 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116282446","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":"Using Compile-Time Granularity Information to Support Dynamic Work Distribution in Parallel Logic Programming Systems","authors":"I. Dutra, V. S. Costa, J. Barbosa, C. Geyer","doi":"10.5753/sbac-pad.1999.19797","DOIUrl":"https://doi.org/10.5753/sbac-pad.1999.19797","url":null,"abstract":"A very important component of a parallel system that generates irregular computational patterns is its work distribution strategy Scheduling strategies for these kinds of systems must be smart enough in order to dynamically balance workload while not incurring a very high overhead. Logic programs running on parallel logic programming systems are examples of irregular parallel computations. The two main forms of parallelism exploited by parallel logic programming systems are: and parallelism, that arises when several literals in the body of a clause can execute in parallel, and or-parallelism, that arises when several alternative clauses in the data base program can be selected in parallel. In this work we show that scheduling strategies for distributing and work and or-work in parallel logic programming systems must combine information obtained at compile-time with runtime information whenever possible, in order to obtain better performance. The information obtained at compile-time has two advantages over current implemented systems that use only runtime information: (1) the user does not need to adjust parameters in arder to estimate sizes of and-work and or-work for the programs; (2) the schedulers can use more accuratc estimates of sizes of and-work and or-work to make belter decisions at runtime. We did our experiments with Andorra-I, a parallel logic programming system that exploits both determinare and-parallelism and or-parallelism. In order to obtain compile-time granularity information we used the ORCA tool. Our benchmark set ranges from programs containing and parallelism only, or-parallelism only and a combination of both and-, and or-parallelism. Our results show that, when well designed, scheduling strategies can actually benefit from compile-time granularity information.","PeriodicalId":267202,"journal":{"name":"Anais do XI International Symposium on Computer Architecture and High Performance Computing (SBAC-PAD 1999)","volume":"53 3","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131659006","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":"A Taxonomy for Cache Memory Misses","authors":"J. L. Hamkalo, B. Cernuschi-Frías","doi":"10.5753/sbac-pad.1999.19773","DOIUrl":"https://doi.org/10.5753/sbac-pad.1999.19773","url":null,"abstract":"One way to understand the causes of cache memory misses is to use a classification for them. Usually statistical models such as the 3C model are used to make the classification. In the present work a new definition for the 3C model: compulsory, capacity and conflict misses are given. The corresponding operational definitions are given, which are based on the use of the LRU stack distances. The proposed model is called a deterministic 3C model or D3C. The D3C model classifies the memory references in an individual way, conforming a taxonomy, and then it is possible to analyze when a memory reference belongs to a given category. Also the passage of a given memory reference from one category to another when some cache parameter is modified may be studied. The D3C model does not present anomalies such as negative conflict miss rales, as in the 3C model. Several patterns for memory access are theoretically analyzed for the 3C and D3C models, showing that the results given by the D3C model are intuitive and have easy interpretation. The 3C model underestimates the conftict misses and overestimates the capacity misses when compared with the D3C model. This difference comes from the references that hit in the cache under study and miss in a fully associative cache of the same size with LRU replacement policy. The amount of these references was measured using SPEC95 benchmarks in trace driven simulations. It is shown that high percent values are obtained for these references for usual cache configurations, and therefore these references have an important participation in the cache statistics.","PeriodicalId":267202,"journal":{"name":"Anais do XI International Symposium on Computer Architecture and High Performance Computing (SBAC-PAD 1999)","volume":"68 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129997168","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":"HetNOS Threads: Programação Multithreaded Distribuída","authors":"Vinicius Mantay, Cláudio Fernando Resin Geyer","doi":"10.5753/sbac-pad.1999.19781","DOIUrl":"https://doi.org/10.5753/sbac-pad.1999.19781","url":null,"abstract":"O pacote de programação HetNOS Threads visa a união de duas importantes tecnologias de desenvolvimento de aplicações paralelas e distribuídas, que são os pacotes de programação multithreaded e a utilização de programação distribuída em redes de computadores com alta velocidade de comunicação. O modelo de programação adotado tem por objetivo oferecer ao desenvolvedor desta classe de aplicações ferramentas para a exploração de paralelismo tanto de alta quanto de baixa granulosidade. Essas metas são alcançadas com a combinação de um pacote de programação multithreaded, as Pthreads, com o sistema operacional de rede HetNOS. São oferecidas primitivas de comunicação entre processos e threads distribuídos via troca de mensagens, manipulação remota destas entidades e variáveis de sincronização distribuídas, de maneira a proporcionar as funcionalidades necessárias ao desenvolvimento de aplicações paralelas distribuídas aliandofacilidade de uso, programação intuitiva e desempenho eficiente.","PeriodicalId":267202,"journal":{"name":"Anais do XI International Symposium on Computer Architecture and High Performance Computing (SBAC-PAD 1999)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129089263","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":"Mirador - Uma Ferramenta para Monitoramento e Gerenciamento de Multicomputadores","authors":"L. A. Araújo, Onofre Trindade Júnior","doi":"10.5753/sbac-pad.1999.19779","DOIUrl":"https://doi.org/10.5753/sbac-pad.1999.19779","url":null,"abstract":"Technological advances in computer memory, communication networks and central processing units have made the use of computer networks, as parallel machines, a cost-effective, high-performance solution. To the user's point of view, it is very important the availability of tools for the monitoring and management of this class of system. The entire machine must look as a single system and not a collection of independent machines. This paper presents the Mirador: a graphic tool for the monitoring and management of multicomputers, including those ones based on local area network technology. The Mirador can be used from any machine in the network running a WWW browser, with support to the Java language. The tool does not depend on any specific computer platform. Several tasks can be performed with the use of the Mirador, including: task removal, node interaction, monitoring of memory occupation and processor load and detection of hardware malfunctions.","PeriodicalId":267202,"journal":{"name":"Anais do XI International Symposium on Computer Architecture and High Performance Computing (SBAC-PAD 1999)","volume":"148 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123397552","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":"A Robust Causal Order Protocol","authors":"George Lima, R. Macêdo","doi":"10.5753/sbac-pad.1999.19791","DOIUrl":"https://doi.org/10.5753/sbac-pad.1999.19791","url":null,"abstract":"This paper presents a robust causal order protocol implemented for the BCG platform (Reliable Group Communication Base) developed at LaSiD/UFBA (Distributed System Laboratory at UFBA). It was designed using a symmetric message recovery approach in which any process of the group can detect and retransmit missing messages providing a reliable message recovery mechanism. Its algorithm, based on certain properties and structures of the BCG, makes the proposed protocol more flexible than similar protocols published to date. Data collected from experiments are also reported.","PeriodicalId":267202,"journal":{"name":"Anais do XI International Symposium on Computer Architecture and High Performance Computing (SBAC-PAD 1999)","volume":"10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121487069","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":"Simultaneous Speculation Scheduling","authors":"A. Unger, T. Ungerer, E. Zehendner","doi":"10.5753/sbac-pad.1999.19787","DOIUrl":"https://doi.org/10.5753/sbac-pad.1999.19787","url":null,"abstract":"Simultaneous Speculation Scheduling (S3) is a combined compiler and architecture technique to control multiple path execution. lt can be applied for dual path branch speculation in case of unpredictable branches and for multiple path speculative execution of loop iterations. Loop-carried dependences are handled by data dependence prediction. Architectural requirements are a minimal form of multithreaded processor architecture and three new instructions (fork, sync, wait). Simulation results show performance gains of up to 40% over purely static scheduling techniques by applying the S3 technique to branches in kernel sections of SPECint95 benchmark programs.","PeriodicalId":267202,"journal":{"name":"Anais do XI International Symposium on Computer Architecture and High Performance Computing (SBAC-PAD 1999)","volume":"105 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134604134","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}