{"title":"The effect of program optimization on trace cache efficiency","authors":"Derek L. Howard, Mikko H. Lipasti","doi":"10.1109/PACT.1999.807570","DOIUrl":"https://doi.org/10.1109/PACT.1999.807570","url":null,"abstract":"Trace cache, an instruction fetch technique that reduces token branch penalties by storing and fetching program instructions in dynamic execution order, dramatically improves instruction fetch bandwidth. Similarly, program transformations like loop unrolling, procedure inlining, feedback-directed program restructuring, and profile-directed feedback can improve instruction fetch bandwidth by changing the static structure and ordering of a program's basic blocks. We examine the interaction of these compile-time and run-time techniques in the context of a high-quality production compiler that implements such transformations and a cycle-accurate simulation model of a wide issue superscalar processor. Not surprisingly, we find that the relative benefit of adding trace cache declines with increasing optimization level, and vice versa. Furthermore, we find that certain optimizations that improve performance on a processor model without trace cache can actually degrade performance on a processor with trace cache due to increased branch history table interference. Finally, we show that the performance obtained with a trace cache of a given size can be obtained with a trace cache of about half the size by applying aggressive compiler optimization techniques.","PeriodicalId":427657,"journal":{"name":"1999 International Conference on Parallel Architectures and Compilation Techniques (Cat. No.PR00425)","volume":"23 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129723895","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":"Reducing cache conflicts by partitioning and privatizing shared arrays","authors":"Zhiyuan Li","doi":"10.1109/PACT.1999.807525","DOIUrl":"https://doi.org/10.1109/PACT.1999.807525","url":null,"abstract":"Parallelizing compilers for shared-memory multiprocessors typically generate fork-join programs in which multiple threads may access different sections of shared arrays. Each thread tends to access discontiguous addresses due to the gap between sections which belong to different arrays. Such interarray discontiguity often causes cache-set conflicts even when the private cache is large enough to hold the working set. A compiler technique, called array partition and privatization (or APP), is proposed to correct this problem. The APP technique partitions a shared array into sections, one for each computing thread, and then maps each section to the private stack of the corresponding thread. Interarray discontiguity is thus removed and the cache performance improved. This paper presents a compiler algorithm for APP and preliminary experimental results based on four benchmark programs running on an SGI Origin 2000 multiprocessor.","PeriodicalId":427657,"journal":{"name":"1999 International Conference on Parallel Architectures and Compilation Techniques (Cat. No.PR00425)","volume":"121 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121733495","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":"Memory system support for image processing","authors":"Lixin Zhang, J. Carter, Wilson C. Hsieh, S. Mckee","doi":"10.1109/PACT.1999.807426","DOIUrl":"https://doi.org/10.1109/PACT.1999.807426","url":null,"abstract":"Image processing applications tend to access their data non-sequentially and reuse that data infrequently. As a result, they tend to perform poorly on conventional memory systems due to high cache and TLB miss rates and are particularly sensitive to the growing latency of main memory. We analyze the memory performance of three image processing algorithms (volume rendering, image warping and image filtering) on both a conventional memory system and on the Impulse memory system. The Impulse memory system allows application software to control how, when, and where data are loaded into a conventional processor cache. It does this by letting software configure how the memory controller interprets the physical addresses exported by the processor, which enables an application to dynamically change how data are fetched. Sparse data can be accessed densely, which improves both cache and TLB utilization, and memory latency is hidden by prefetching data within the memory controller. We find that for these image processing codes, using an Impulse memory system yields speedups of 40% to 226% over an otherwise identical machine with a conventional memory system.","PeriodicalId":427657,"journal":{"name":"1999 International Conference on Parallel Architectures and Compilation Techniques (Cat. No.PR00425)","volume":"9 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129985803","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 the complexity of loop fusion","authors":"A. Darte","doi":"10.1109/PACT.1999.807510","DOIUrl":"https://doi.org/10.1109/PACT.1999.807510","url":null,"abstract":"Loop fusion is a program transformation that combines several loops into one. It is used in parallelizing compilers mainly for increasing the granularity of loops and for improving data reuse. The goal of this paper is to study, from a theoretical point of view, several variants of the loop fusion problem-identifying polynomially solvable cases and NP-complete cases-and to make the link between these problems and some scheduling problems that arise from completely different areas. We study among others, the fusion of loops of different types, and the fusion of loops when combined with loop shifting.","PeriodicalId":427657,"journal":{"name":"1999 International Conference on Parallel Architectures and Compilation Techniques (Cat. No.PR00425)","volume":"14 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130165944","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}
Tsuneo Nakanishi, K. Joe, C. Polychronopoulos, A. Fukuda
{"title":"The modulo interval: a simple and practical representation for program analysis","authors":"Tsuneo Nakanishi, K. Joe, C. Polychronopoulos, A. Fukuda","doi":"10.1109/PACT.1999.807422","DOIUrl":"https://doi.org/10.1109/PACT.1999.807422","url":null,"abstract":"In this paper, the modulo interval, an extension of the traditional interval on real numbers, and its useful mathematical properties are presented as a representation for program analysis. Only with two additional parameters to the interval on real numbers, namely the modulus and the residue, the modulo interval can represent information on programs having cyclicity such as loop indices, array subscripts etc., at reasonable complexity and more accuracy. Well-defined arithmetic and set operations on the modulo interval make implementation of compilers simple and reliable. Moreover, application of the modulo interval to program analysis for parallelizing compilers is discussed in this paper.","PeriodicalId":427657,"journal":{"name":"1999 International Conference on Parallel Architectures and Compilation Techniques (Cat. No.PR00425)","volume":"28 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128474427","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}
Daniel Ortega, Ivan Martel, V. Krishnan, E. Ayguadé, M. Valero
{"title":"Quantifying the benefits of SPECint distant parallelism in simultaneous multithreading architectures","authors":"Daniel Ortega, Ivan Martel, V. Krishnan, E. Ayguadé, M. Valero","doi":"10.1109/PACT.1999.807436","DOIUrl":"https://doi.org/10.1109/PACT.1999.807436","url":null,"abstract":"We exploit the existence of distant parallelism that future compilers could detect and characterise its performance under simultaneous multithreading architectures. By distant parallelism we mean parallelism that cannot be captured by the processor instruction window and that can produce threads suitable for parallel execution in a multithreaded processor. We show that distant parallelism can make feasible wider issue processors by providing more instructions from the distant threads, thus better exploiting the resources from the processor in the case of speeding up single integer applications. We also investigate the necessity of out-of-order processors in the presence of multiple threads of the same program. It is important to notice at this point that the benefits described are totally orthogonal to any other architectural techniques targeting a single thread.","PeriodicalId":427657,"journal":{"name":"1999 International Conference on Parallel Architectures and Compilation Techniques (Cat. No.PR00425)","volume":"6 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114970543","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":"Performance characteristics of a network of commodity multiprocessors for the NAS benchmarks using a hybrid memory model","authors":"F. Cappello, Olivier Richard","doi":"10.1109/PACT.1999.807434","DOIUrl":"https://doi.org/10.1109/PACT.1999.807434","url":null,"abstract":"Multiprocessors and high performance networks offer the opportunity to construct CLUster of MultiProcessors (CLUMPs) and use them as parallel computing platforms. The distinctive feature of the CLUMPs over traditional parallel computers is their hybrid memory model (message passing between the nodes and shared memory inside the nodes). We investigate the performance characteristics of a cluster of biprocessor PCs for the NAS 2.3 parallel benchmark using a programming model based on MPI for message passing between biprocessor nodes and OpenMP for shared memory inside biprocessor nodes. The paper provides several contributions. These include: speed-up measurements of a cluster of biprocessor PCs over a cluster of uniprocessor PCs using the hybrid memory model; a detailed analysis of the speed-up results from a breakdown of the benchmarks execution time; and a performance comparison of a commodity CLUMP with some high performance parallel computers.","PeriodicalId":427657,"journal":{"name":"1999 International Conference on Parallel Architectures and Compilation Techniques (Cat. No.PR00425)","volume":"9 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124143213","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":"Linux alighted: down to earth clusters","authors":"Beau Vrolyk","doi":"10.5555/520793.825740","DOIUrl":"https://doi.org/10.5555/520793.825740","url":null,"abstract":"Sky high costs and through the roof servicing of traditional specialized supercomputing has put immense pressure for earth-bound pricing and clustering solutions. Vectors and scalable UNIX-based solutions are not going to disappear, but over time Linux clusters will likely do for supercomputing what the Apple II did for personal computing: bring down costs, ease use, and make high-end computing power commercially viable and consumer available. This movement is in its adolescence, but maturity is rapidly approaching. This talk will discuss what is being done to integrate today's real solutions with tomorrow's promise.","PeriodicalId":427657,"journal":{"name":"1999 International Conference on Parallel Architectures and Compilation Techniques (Cat. No.PR00425)","volume":"36 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127554449","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":"Lower bounding techniques for the multiprocessor scheduling problem with communication delay","authors":"S. Fujita, Tadanori Nakagawa","doi":"10.1109/PACT.1999.807534","DOIUrl":"https://doi.org/10.1109/PACT.1999.807534","url":null,"abstract":"This paper proposes two techniques for obtaining a sharp lower bound for the multiprocessor scheduling problem (MSP) with nonnegligible communication delay. In the proposed techniques, we apply the notion of inevitable communication delay to obtain a nontrivial lower bound on the scheduling length. The effectiveness of the derived bound is evaluated by conducting experiments on several randomly generated instances. By the results of the experiments, it is shown that the proposed techniques generate a very sharp lower bound that is at least 97.5% of an upper bound, when the number of processors is not very small (e.g., at least 10) and the maximum communication cost is not very large (e.g., less than or equal to the minimum execution cost).","PeriodicalId":427657,"journal":{"name":"1999 International Conference on Parallel Architectures and Compilation Techniques (Cat. No.PR00425)","volume":"26 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133343689","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":"Control-flow speculation through value prediction for superscalar processors","authors":"José González, Antonio González","doi":"10.1109/PACT.1999.807406","DOIUrl":"https://doi.org/10.1109/PACT.1999.807406","url":null,"abstract":"In this paper, we introduce a new branch predictor that predicts the outcomes of branches by predicting the value of their inputs and performing an early computation of their results according to the predicted values. The design of a hybrid predictor comprising our branch predictor and a correlating branch predictor is presented. We also propose a new selector that chooses the most reliable prediction for each branch. This selector is based on the path followed to reach the branch. Results for immediate updates show a significant improvement with respect to a conventional hybrid predictor for different size configurations. In addition, the proposed hybrid predictor with a size of 8 KB achieves the same miss ratio as a conventional one of 64 KB. Performance evaluation for a dynamically-scheduled superscalar processor, with realistic updates, shows a speed-up of 11% despite its higher latency (up to 4 cycles).","PeriodicalId":427657,"journal":{"name":"1999 International Conference on Parallel Architectures and Compilation Techniques (Cat. No.PR00425)","volume":"93 5 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126030782","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}