DMCA
ScalaExtrap: Trace-based communication extrapolation for SPMD programs (2011)
Cached
Download Links
| Venue: | In ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming |
| Citations: | 18 - 9 self |
BibTeX
@INPROCEEDINGS{Wu11scalaextrap:trace-based,
author = {Xing Wu and Frank Mueller},
title = {ScalaExtrap: Trace-based communication extrapolation for SPMD programs},
booktitle = {In ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming},
year = {2011}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
Performance modeling for scientific applications is important for assessing potential application performance and systems procurement in high-performance computing (HPC). Recent progress on communication tracing opens up novel opportunities for communication modeling due to its lossless yet scalable trace collection. Estimating the impact of scaling on communication efficiency still remains non-trivial due to execution-time variations and exposure to hardware and software artifacts. This work contributes a fundamentally novel modeling scheme. We synthetically generate the application trace for large numbers of nodes by extrapolation from a set of smaller traces. We devise an innovative approach for topology extrapolation of single program, multiple data (SPMD) codes with stencil or mesh communication. The extrapolated trace can subsequently be (a) replayed to assess communication requirements before porting an application, (b) transformed to auto-generate communication benchmarks for various target platforms, and (c) analyzed to detect communication inefficiencies and scalability limitations. To the best of our knowledge, rapidly obtaining the communication behavior of parallel applications at arbitrary scale with the availability of timed replay, yet without actual execution of the application at this scale is without precedence and has the potential to enable otherwise infeasible system simulation at the exascale level.
Keyphrases
spmd program trace-based communication extrapolation novel opportunity timed replay application trace high-performance computing innovative approach recent progress large number exascale level communication requirement scalable trace collection extrapolated trace topology extrapolation communication inefficiency potential application performance multiple data scientific application software artifact arbitrary scale infeasible system simulation communication tracing actual execution system procurement communication behavior scalability limitation execution-time variation mesh communication auto-generate communication benchmark various target platform single program communication efficiency parallel application
