The cache performance and optimizations of blocked algorithms

Lam, Monica D.; Rothberg, Edward; Wolf, Michael E.

doi:10.1145/106972.106981

Cited by 605 publications

(132 citation statements)

References 8 publications

Supporting

Mentioning

119

Contrasting

Unclassified

Order By: Relevance

“…Improving locality has also become one of the goals of algorithmic design, in the interest of better using the memory hierarchy [423,28,334]. When considered at the granularity of pages, locality lies at the basis of the working set model and of paging algorithms [171,174,172].…”

Section: Spatial and Temporal Localitymentioning

confidence: 99%

Workload Modeling for Computer Systems Performance Evaluation

Feitelson

2015

191

131

View full text Add to dashboard Cite

Reliable performance evaluations require the use of representative workloads. This is no easy task since modern computer systems and their workloads are complex, with many interrelated attributes and complicated structures. Experts often use sophisticated mathematics to analyze and describe workload models, making these models difficult for practitioners to grasp. This book aims to close this gap by emphasizing the intuition and the reasoning behind the definitions and derivations related to the workload models. It provides numerous examples from real production systems, with hundreds of graphs. Using this book, readers will be able to analyze collected workload data and clean it if necessary, derive statistical models that include skewed marginal distributions and correlations, and consider the need for generative models and feedback from the system. The descriptive statistics techniques covered are also useful for other domains.

show abstract

Section: Spatial and Temporal Localitymentioning

confidence: 99%

Workload Modeling for Computer Systems Performance Evaluation

Feitelson

2015

191

131

View full text Add to dashboard Cite

show abstract

“…A more efficient way to distribute data among parallel processors is to use blocked partitioning, that is to divide in more than one dimension the data structures [9,21]. Blocked solutions are frequently adopted in numerical operations because this strategy enables a better memory access and cache use for the applications.…”

Section: Blocked Parallel Algorithmsmentioning

confidence: 99%

Parallel Lattice Boltzmann Method with Blocked Partitioning

Schepke

Maillard

Navaux

2009

Int J Parallel Prog

View full text Add to dashboard Cite

This paper presents and discusses a blocked parallel implementation of bi-and three-dimensional versions of the Lattice Boltzmann Method. This method is used to represent and simulate fluid flows following a mesoscopic approach. Most traditional parallel implementations use simple data distribution strategies to parallelize the operations on the regular fluid data set. However, it is well known that block partitioning is usually better. Such a parallel implementation is discussed and its communication cost is established. Fluid flows simulations crossing a cavity have also been used as a real-world case study to evaluate our implementation. The presented results with our blocked implementation achieve a performance up to 31% better than non-blocked versions, for some data distributions. Thus, this work shows that blocked, parallel implementations can be efficiently used to reduce the parallel execution time of the method.

show abstract

“…A number of groups are attempting to improve performance through architectural innovation [18]. Other groups are attacking the problem in software: either in the compiler-level through reordering instructions and prefetching [9,10], or through complex data layouts to improve cache performance [2,16].…”

Section: Related Workmentioning

confidence: 99%

“…Nonserial polyadic dynamic programming algorithms, on the other hand, pose unique challenges to improving cache performance due to their irregular data access patterns. These challenges are significantly different from those faced in the dense linear algebra problems, which are often easily handled using standard cache-friendly optimizations such as tilling or blocking [4,10]. Optimizations such as tiling can be applied to nonserial polyadic dynamic programming only after considering these specific details individually.…”

Section: Related Workmentioning

confidence: 99%

Cache oblivious algorithms for nonserial polyadic programming

2007

View full text Add to dashboard Cite

The nonserial polyadic dynamic programming algorithm is one of the most fundamental algorithms for solving discrete optimization problems. Although the loops in the nonserial polyadic dynamic programming algorithm are similar to those in matrix multiplication, the available automatic optimization techniques have little effect on this imperfect loop because of nonuniform data dependencies. In this paper, we develop algorithmic optimizations to improve the cache performance of the nonserial polyadic dynamic programming algorithm. Our algorithmic transformation takes advantage of the cache oblivious method by relaxing some dependencies in the standard iterative version. Based on the ideal cache model of the cache oblivious algorithm, the approximate bound of cache misses is given by ( n 3 Z L √ Z). We also found that the optimized algorithm with the cache oblivious approach is more sensitive to conventional optimization techniques such as tiling. Experimental results on several platforms show that the optimized algorithms improve the cache performance and achieves speedups of 2-10 times.

show abstract

The cache performance and optimizations of blocked algorithms

Cited by 605 publications

References 8 publications

Workload Modeling for Computer Systems Performance Evaluation

Workload Modeling for Computer Systems Performance Evaluation

Parallel Lattice Boltzmann Method with Blocked Partitioning

Cache oblivious algorithms for nonserial polyadic programming

Contact Info

Product

Resources

About