Dynamic Tracing: Memoization of Task Graphs for Dynamic Task-Based Runtimes
| Autor: | Sean J. Treichler, Todd Warszawski, Michael Garland, Elliott Slaughter, Michael Bauer, Alex Aiken, Wonchan Lee |
|---|---|
| Rok vydání: | 2018 |
| Předmět: |
020203 distributed computing
Distributed database Computer science Memoization Computation 010103 numerical & computational mathematics 02 engineering and technology Parallel computing Dependence analysis Tracing Supercomputer 01 natural sciences Graph Dynamic programming 0202 electrical engineering electronic engineering information engineering Task analysis 0101 mathematics |
| Zdroj: | SC |
| DOI: | 10.1109/sc.2018.00037 |
| Popis: | Many recent programming systems for both supercomputing and data center workloads generate task graphs to express computations that run on parallel and distributed machines. Due to the overhead associated with constructing these graphs the dependence analysis that generates them is often statically computed and memoized, and the resulting graph executed repeatedly at runtime. However, many applications require a dynamic dependence analysis due to data dependent behavior, but there are new challenges in capturing and re- executing task graphs at runtime. In this work, we introduce dynamic tracing, a technique to capture a dynamic dependence analysis of a trace that generates a task graph, and replay it. We show that an implementation of dynamic tracing improves strong scaling by an average of 4.9× and up to 7.0× on a suite of already optimized benchmarks. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|