Optimizing Indirect Memory References with milk

Kiriansky, Vladimir; Zhang, Yunming; Amarasinghe, Saman

doi:10.1145/2967938.2967948

Cited by 38 publications

(27 citation statements)

References 47 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Our experiments show that the algorithms achieve good parallel scalability and signicantly better performance than prior work. Future work includes extending graph optimizations for locality and scalability (e.g., [7,53,57,85,92,96,97,101]) to hypergraphs. numbers reported in the paper are obtained using Cilk Plus.…”

Section: Resultsmentioning

confidence: 99%

Practical parallel hypergraph algorithms

Shun¹

2020

Proceedings of the 25th ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming

View full text Add to dashboard Cite

While there has been signicant work on parallel graph processing, there has been very surprisingly little work on highperformance hypergraph processing. This paper presents a collection of ecient parallel algorithms for hypergraph processing, including algorithms for betweenness centrality, maximal independent set, k-core decomposition, hypertrees, hyperpaths, connected components, PageRank, and single-source shortest paths. For these problems, we either provide new parallel algorithms or more ecient implementations than prior work. Furthermore, our algorithms are theoretically-ecient in terms of work and depth. To implement our algorithms, we extend the Ligra graph processing framework to support hypergraphs, and our implementations benet from graph optimizations including switching between sparse and dense traversals based on the frontier size, edge-aware parallelization, using buckets to prioritize processing of vertices, and compression. Our experiments on a 72-core machine and show that our algorithms obtain excellent parallel speedups, and are signicantly faster than algorithms in existing hypergraph processing frameworks.

show abstract

Section: Resultsmentioning

confidence: 99%

Practical parallel hypergraph algorithms

Shun¹

2020

Proceedings of the 25th ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming

View full text Add to dashboard Cite

show abstract

“…It is the middle stage that utilizes the resources to the maximum capability by mapping the above software to the hardware [24], [48], [49].…”

Section: ) System Software Levelmentioning

confidence: 99%

Towards a Novel Framework for Automatic Big Data Detection

Ahmed

Ismail

2020

IEEE Access

View full text Add to dashboard Cite

Big data is a "relative" concept. It is the combination of data, application, and platform properties. Recently, big data specific technologies have emerged, including software frameworks, databases, hardware accelerators, storage technologies, etc. However, the automatic selection of these solutions for big data computations remains a non-trivial task. Presently, the big data tools are selected by analyzing the problem manually, or by using several performance prediction techniques. The manual identification is based on the data properties only, whereas the performance predictors only estimate basic execution metrics without linking them with big data (3Vs) thresholds. Hence, both ways of identification are mostly incorrect, which can lead to inefficient use of 3Vs optimizations, resulting into global inefficiency, reduced system performance, increasing power consumption, requiring greater effort on the part of the programming team, and misallocation of the hardware resources required for the task. In this regard, a novel framework has been proposed for automatic detection of 3Vs (Volume, Velocity, Variety) of big data, using machine learning. The detection is done through static code features, data, and platform properties, leading to relevant tool selection, and code generation, with minimal overheads, lesser programmer interventions, higher usability, and portability. Instead of handling each application with big data specialized solutions, or manually identifying the 3Vs, the framework can automatically detect and link the 3Vs to the relevant optimizations. Several standard applications have been tested using the proposed framework. In the case of volume, the average detection accuracy is up to 97.8% for seen and 95.9% for unseen applications. In the case of velocity, the average detection accuracy is up to 97.3% for seen and 92.6% for unseen applications. There is no margin of error in variety detection, as it has straightforward computations without any predictions. Furthermore, an airline recommendation system case study strengthens the effectiveness of the proposed approach.

show abstract

“…Irregular workloads are important and challenging: Data movement is particularly challenging on applications that access data in irregular and unpredictable patterns. These include many important workloads in, e.g., machine learning [32,58,68], graph processing [54,59], and databases [92].…”

Section: Data Movement Is a Growing Problemmentioning

confidence: 99%

Livia

Lockerman

Feldmann

Bakhshalipour

et al. 2020

Proceedings of the Twenty-Fifth International Conference on Architectural Support for Programming Languages and Operating Syste

View full text Add to dashboard Cite

In order to scale, future systems will need to dramatically reduce data movement. Data movement is expensive in current designs because (i) traditional memory hierarchies force computation to happen unnecessarily far away from data and (ii) processing-in-memory approaches fail to exploit locality. We propose Memory Services, a flexible programming model that enables data-centric computing throughout the memory hierarchy. In Memory Services, applications express functionality as graphs of simple tasks, each task indicating the data it operates on. We design and evaluate Livia, a new system architecture for Memory Services that dynamically schedules tasks and data at the location in the memory hierarchy that minimizes overall data movement. Livia adds less than 3% area overhead to a tiled multicore and accelerates challenging irregular workloads by 1.3× to 2.4× while reducing dynamic energy by 1.2× to 4.7×. CCS Concepts • Computer systems organization → Processors and memory architectures.

show abstract

Optimizing Indirect Memory References with milk

Cited by 38 publications

References 47 publications

Practical parallel hypergraph algorithms

Practical parallel hypergraph algorithms

Towards a Novel Framework for Automatic Big Data Detection

Livia

Contact Info

Product

Resources

About