Performance characterization of data mining benchmarks

Mekkat, Vineeth; Natarajan, Ragavendra; Hsu, Wei-Chung; Zhai, Antonia

doi:10.1145/1739025.1739040

Cited by 5 publications

(6 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…32 However, we have demonstrated that only a few benchmarks scale linearly when multiple threads of execution shares the same cores. 19 Our work shows the impact of data sharing and organization of last-level cache architecture can a®ect scalability. Previous work 16 has also pointed out the importance of communication overheads and resource utilization in the workload scalability.…”

Section: E®ect Of Prefetching On Scalabilitymentioning

confidence: 87%

“…In Sec. 4, we revisit the key observations on memory characteristics of data mining benchmarks as discussed in Mekkat et al 19 and in Secs. 5 and 6, we study the e®ect of compiler-directed prefetching on data mining benchmarks running in single-thread and multithread modes, respectively.…”

Section: Characteristics Of Nu-minebenchmentioning

confidence: 99%

“…In our prior work, 19 we examined the instruction distribution, memory hierarchy usage and stall distribution in order to study the basic characteristics of data mining applications. We used SPEC CPU 24 for comparison as SPEC has been widely accepted by computer architects as the chosen benchmark suite for measuring CPU performance.…”

Section: Characteristics Of Nu-minebenchmentioning

confidence: 99%

See 2 more Smart Citations

Effectiveness of Compiler-Directed Prefetching on Data Mining Benchmarks

Natarajan

Mekkat

Hsu

et al. 2012

J CIRCUIT SYST COMP

Self Cite

View full text Add to dashboard Cite

For today's increasingly power-constrained multicore systems, integrating simpler and more energy-efficient in-order cores becomes attractive. However, since in-order processors lack complex hardware support for tolerating long-latency memory accesses, developing compiler technologies to hide such latencies becomes critical. Compiler-directed prefetching has been demonstrated effective on some applications. On the application side, a large class of data centric applications has emerged to explore the underlying properties of the explosively growing data. These applications, in contrast to traditional benchmarks, are characterized by substantial thread-level parallelism, complex and unpredictable control flow, as well as intensive and irregular memory access patterns. These applications are expected to be the dominating workloads on future microprocessors. Thus, in this paper, we investigated the effectiveness of compiler-directed prefetching on data mining applications in in-order multicore systems. Our study reveals that although properly inserted prefetch instructions can often effectively reduce memory access latencies for data mining applications, the compiler is not always able to exploit this potential. Compiler-directed prefetching can become inefficient in the presence of complex control flow and memory access patterns; and architecture dependent behaviors. The integration of multithreaded execution onto a single die makes it even more difficult for the compiler to insert prefetch instructions, since optimizations that are effective for single-threaded execution may or may not be effective in multithreaded execution. Thus, compiler-directed prefetching must be judiciously deployed to avoid creating performance bottlenecks that otherwise do not exist. Our experiences suggest that dynamic performance tuning techniques that adjust to the behaviors of a program can potentially facilitate the deployment of aggressive optimizations in data mining applications.

show abstract

Section: E®ect Of Prefetching On Scalabilitymentioning

confidence: 87%

Section: Characteristics Of Nu-minebenchmentioning

confidence: 99%

Section: Characteristics Of Nu-minebenchmentioning

confidence: 99%

See 1 more Smart Citation

Effectiveness of Compiler-Directed Prefetching on Data Mining Benchmarks

Natarajan

Mekkat

Hsu

et al. 2012

J CIRCUIT SYST COMP

Self Cite

View full text Add to dashboard Cite

show abstract

“…Only through effective communication the team is in order to play their collective strengths. Information communication includes all the information to send, to receive, to transfer, and the interaction of this process [1,10]. In software development, all to obtain…”

Section: Introductionmentioning

confidence: 99%

Application Research on Data Mining Methods in Information Communication Mode of Software Development

ye¹,

Zhang²

2012

IJEME

View full text Add to dashboard Cite

Smaller time loss and smoother information communication mode is the urgent pursuit of the software R&D enterprise. Information communication is difficult to control and manage and it needs more technical to support. Data mining is an intelligent way tried to analyze knowledge and laws which hidden in massive amounts of data. Data mining technology together with share repositories can improve the intelligent degree of information communication mode. In this paper, the framework of intelligent information communication mode which based on data mining technology and share repositories is advanced, and data mining model for information communication of software development is designed. In view of the extant single decision tree algorithm existence the characteristics that counting inefficient and its learning based on supervise, a new semisupervised learning algorithm three decision trees voting classification algorithm based on tri-training (TTVA) is proposed. This algorithm in training only requests a few labeled data, and can use massively unlabeled data repeatedly revision to the classifier. It has overcome the single decision tree algorithm shortcoming. Experiments on the real communicated data sets of software developmental item indicate that TTVA has the good identification and accuracy to the crux issues mining, and can apply to the decision analysis of the development and management of the software project. At the same time, TTVA can effectively exploit the massively unlabeled data to enhance the learning performance.Index Terms: Information communication mode of software development; data mining; share repositories; semi-supervised learning; Three decision trees voting classification algorithm based on Tri-training (TTVA) 71 Mode of Software Development information from outside their own activities belong to the scope of the information communication. Information communication data have some characteristics such as reconcilability, transitivity, can be transformative and can be stored and others, these characteristics create the conditions for the information collection and reuse and afford idea for the intelligent information communication mode to apply data mining technologies and sharing technologies.Data mining is the analysis of (often large) observational data sets from the database, data warehouse or other large repository incomplete, noisy, ambiguous, the practical application of random data to find unsuspected relationships and summarize the data that are both understandable and useful to the data owner. It is a means that data extraction, cleaning and transformation, analysis, and other treatment models, and automatically discovers the patterns and interesting knowledge hidden in large amounts of data, this helps us make decisions based on a wealth of data. Information communication mode of software development lies in how to collection, analysis, and mine out the hidden useful information in the various data from information communication between developers and the staff interaction with manag...

show abstract

“…With effective inter and intra team communication, team collective strength can be greatly realized. Generally speaking, such communication includes sending, receiving, transferring of software development information, and combination between these behaviors [1,10]. Information in software development is characterized as reconcilable, transitive, imprecise, structured and uncertain, which makes the analysis a challenge, and powerful tools are required.…”

Section: Introductionmentioning

confidence: 99%

Data Mining based Software Development Communication Pattern Discovery

Zhang¹,

Ye²,

Wang³

et al. 2010

IJMECS

View full text Add to dashboard Cite

Abstract-Smaller time loss and smoother communication pattern is the urgent pursuit in the software development enterprise. However, communication is difficult to control and manage and demands on technical support, due to the uncertainty and complex structure of data appeared in communication. Data mining is a well established framework aiming at intelligently discovering knowledge and principles hidden in massive amounts of original data. Data mining technology together with shared repositories results in an intelligent way to analyze data of communication in software development environment. We propose a data mining based algorithm to tackle the problem, adopting a co-training styled algorithm to discover pattern in software development environment. Decision tree is trained as based learners and a majority voting procedure is then launched to determine labels of unlabeled data. Based learners are then trained again with newly labeled data and such iteration stops when a consistent state is reached. Our method is naturally semi-supervised which can improve generalization ability by making use of unlabeled data. Experimental results on data set gathered from productive environment indicate that the proposed algorithm is effective and outperforms traditional supervised algorithms.

show abstract

Performance characterization of data mining benchmarks

Cited by 5 publications

References 17 publications

Effectiveness of Compiler-Directed Prefetching on Data Mining Benchmarks

Effectiveness of Compiler-Directed Prefetching on Data Mining Benchmarks

Application Research on Data Mining Methods in Information Communication Mode of Software Development

Data Mining based Software Development Communication Pattern Discovery

Contact Info

Product

Resources

About