JUNE 18–22, 2017
FRANKFURT AM MAIN, GERMANY

Presentation Details
 
Name: (RP17) How Does Memory Affect the Performance of Tasks?
 
Time: Tuesday, June 20, 2017
08:35 am - 09:45 am
 
Room:   Substanz 1+2  
 
Breaks:07:30 am - 10:00 am Welcome Coffee
 
Presenter:   German Ceballos, Uppsala University
 
Abstract:  
Recent scheduling heuristics for task-based applications have managed to improve performance by taking into account memory-related properties such as data locality and cache sharing. However, there is still a general lack of tools that can provide insights into why different schedulers improve memory behavior, and how this is related to the applications' performance. To address this, we present TaskInsight, a methodology to characterize the memory behavior of different task schedulers through the analysis of data reuse between tasks. TaskInsight 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 affected 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 decisions caused 20% worse performance on average due to 40% more last level cache misses for tasks of the same type. The methodology also exposes how this effect was caused by changes in the tasks' data reuse through the private and shared caches because of scheduling. TaskInsight provides useful and flexible insight, which is key for optimization in many contexts such as data locality, throughput, memory footprint and even energy efficiency. This information can be fed-back to the runtime system to improve online scheduling, to the developers to change the application or even used for compile-time optimizations.

Authors:
German Ceballos, Uppsala University
Andra Hugo, Uppsala University
David Black-Schaffer, Uppsala University
  
 
Download

RP17_Ceballos.pdf (3724 KB)