Loading...
Please wait, while we are loading the content...
Similar Documents
A Scripting Interface for Swift / T Parallel Workloads Using Messaging Queues
| Content Provider | Semantic Scholar |
|---|---|
| Author | Collins, Michael Wozniak, Justin M. Collins, Matt |
| Copyright Year | 2016 |
| Abstract | Modern scientific computing applications require not only highly parallel high-performance computing workloads but also the expressiveness and simplicity of scripting languages. In this work we present an interface that allows external C and C++ programs to control a parallel workflow using Swift/T. This interface facilitates the use of existing C/C++ algorithm implementations to run distributed programming models in an inversion-of-control fashion without the requisite knowledge of scheduling and message passing. The implemented framework uses a messaging queue interface that requires minimal modification to existing libraries and applications. We evaluate this solution using a genetic algorithm in C++ to perform an adaptive parameter search on an agent-based model. We perform this evaluation on large-scale distributed computing machines and compare system utilization and computational runtime with existing frameworks. |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | http://datasys.cs.iit.edu/grants/BigDataX/2016/reports/tr-script.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
