Title :
Identifying Performance Bottlenecks in Work-Stealing Computations
Author :
Tallent, Nathan R. ; Mellor-Crummey, John M.
Author_Institution :
Rice Univ., Houston, TX, USA
Abstract :
Work stealing is an effective load-balancing strategy for multithreading, but when computations based on it underperform, traditional tools can´t explain why. To resolve a computation´s key performance obstacles, tools must pinpoint and quantify parallel idleness and overhead.
Keywords :
multi-threading; resource allocation; load-balancing strategy; multithreading; parallel idleness; parallel overhead; performance bottleneck identification; work-stealing computations; Concurrent computing; Multithreading; Call path profiling; HPCToolkit; Multicore processors; Multithreaded programming models; Performance analysis;
DOI :
10.1109/MC.2009.396
