Formalizing Data Locality in Task Parallel Applications

Ceballos, Germán; Hägersten, Erik; Black-Schaffer, David

doi:10.1007/978-3-319-49956-7_4

Cited by 8 publications

(6 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Several recent works have focused on CRD profiles for predicting the performance of shared cache [21,33,76,79,89]. Recently, researchers attempted to use analytical model and sampling to speed up the performance prediction [13,47,68,70,71].…”

Section: Reuse Distance Analysis On Multicore Processorsmentioning

confidence: 99%

PPT-Multicore: Performance Prediction of OpenMP applications using Reuse Profiles and Analytical Modeling

Barai,

Arafa,

Badawy

et al. 2021

Preprint

View full text Add to dashboard Cite

We present PPT-Multicore, an analytical model embedded in the Performance Prediction Toolkit (PPT) to predict parallel application performance running on a multicore processor. PPT-Multicore builds upon our previous work towards a multicore cache model. We extract LLVM basic block labeled memory trace using an architecture-independent LLVM-based instrumentation tool only once in an application's lifetime. The model uses the memory trace and other parameters from an instrumented sequentially executed binary. We use a probabilistic and computationally efficient reuse profile to predict the cache hit rates and runtimes of OpenMP programs' parallel sections. We model Intel's Broadwell, Haswell, and AMD's Zen2 architectures and validate our framework using different applications from PolyBench and PARSEC benchmark suites. The results show that PPT-Multicore can predict cache hit rates with an overall average error rate of 1.23% while predicting the runtime with an error rate of 9.08%.

show abstract

Section: Reuse Distance Analysis On Multicore Processorsmentioning

confidence: 99%

PPT-Multicore: Performance Prediction of OpenMP applications using Reuse Profiles and Analytical Modeling

Barai,

Arafa,

Badawy

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Traditional graphics scheduling approaches have focused on keeping hardware resources busy [10]. Yet scheduling can have a drastic impact in the data locality properties of the applications as well [11], [12], [13]. Scheduling tasks that share data together can reduce bandwidth and improve performance (frame rate) as they will be able to keep reused data in smaller caches.…”

Section: I I S C H E D U L I N Gmentioning

confidence: 99%

Behind the Scenes: Memory Analysis of Graphical Workloads on Tile-Based GPUs

Ceballos

Sembrant

Carlson

et al. 2018

2018 IEEE International Symposium on Performance Analysis of Systems and Software (ISPASS)

Self Cite

View full text Add to dashboard Cite

“…We start with an example that shows how the overall performance of an application changes when executing with different schedules due to an increase in last-level cache misses (Section 2). We then propose a profiling tool and TaskInsight's data classification technique that allows to clearly di↵erentiate the schedules in terms of their data reuse patterns, using a data reuse graph as in [2] (Section 3). Later, we show how to connect this classification to changes in data reuse, changes in cache misses and changes in performance during the execution: first from the perspective of the private caches (temporal locality on a single-threaded execution, Section 4) and later from the shared caches (spatial locality on multi-threaded run, Section 5).…”

Section: A New Technique To Analyze Schedulers Based On Thementioning

confidence: 99%

“…At the same time, TaskInsight builds the application's (schedule-independent) reuse graph introduced in [2], and combines it with the co-running task sequence to compute the set of memory addresses used by each co-running set. This allows to model the sequence of co-running tasks over time, and use the analysis in Section 4 to analyze how much data was reused over time, but in the shared cache.…”

Section: Locality Of Shared Cachesmentioning

confidence: 99%

See 1 more Smart Citation

TaskInsight

Ceballos

Grass

Hugo

et al. 2017

Proceedings of the 8th International Workshop on Programming Models and Applications for Multicores and Manycores

Self Cite

View full text Add to dashboard Cite

Recent scheduling heuristics for task-based applications have managed to improve their by taking into account memoryrelated properties such as data locality and cache sharing. However, there is still a general lack of tools that can provide insights into why, and where, di↵erent schedulers improve memory behavior, and how this is related to the applications' performance. To address this, we present TaskInsight, a technique to characterize the memory behavior of di↵erent task schedulers through the analysis of data reuse between tasks. Task-Insight provides high-level, quantitative information that can be correlated with tasks' performance variation over time to understand data reuse through the caches due to scheduling choices. TaskInsight is useful to diagnose and identify which scheduling decisions a↵ected performance, when were they taken, and why the performance changed, both in single and multi-threaded executions. We demonstrate how TaskInsight can diagnose examples where poor scheduling caused over 10% di↵erence in performance for tasks of the same type, due to changes in the tasks' data reuse through the private and shared caches, in single and multi-threaded executions of the same application. This flexible insight is key for optimization in many contexts, including data locality, throughput, memory footprint or even energy e ciency.

show abstract

Formalizing Data Locality in Task Parallel Applications

Cited by 8 publications

References 18 publications

PPT-Multicore: Performance Prediction of OpenMP applications using Reuse Profiles and Analytical Modeling

PPT-Multicore: Performance Prediction of OpenMP applications using Reuse Profiles and Analytical Modeling

Behind the Scenes: Memory Analysis of Graphical Workloads on Tile-Based GPUs

TaskInsight

Contact Info

Product

Resources

About