Using Pre-Execution and Helper Threads for Speeding Up Data Intensive Applications.
Saved in:
| Title: | Using Pre-Execution and Helper Threads for Speeding Up Data Intensive Applications. |
|---|---|
| Authors: | Dudás, Ákos1 akos.dudas@aut.bme.hu, Juhász, Sándor1 juhasz.sandor@aut.bme.hu |
| Source: | Proceedings of the World Congress on Engineering 2011 Volume I. 2011, p329-334. 6p. |
| Subjects: | Central processing units, Threads (Computer programs), Computer software, Computer storage devices |
| Abstract: | Pre-execution is a new technique used in conjunction with simultaneous multithreading or multi-core CPUs to reduce memory latency. Executing a slice of a program in a software or hardware thread ahead of the normal execution resolves memory addresses and prefetches data into the caches. By doing so the latency of memory reads is reduced in the main thread. Data intensive applications can benefit from pre-execution even if thread level parallelism is not available because of shared (software) resources. Despite the simplicity of the idea several factors have to be considered when putting this principle into practice. This paper applies and customizes pre-execution to a particular problem of hash table based data transformation and through this example provides a parameterized software preexecution algorithm which can be applied to arbitrary programs and executed on everyday hardware. The most important factor to be considered is to continuously keep an optimal temporal distance between the pre-worker and the main thread, which should be implemented with introducing minimal amount of control and communication dependencies between them. This paper presents a mechanism for attaining this goal. [ABSTRACT FROM AUTHOR] |
| Copyright of Proceedings of the World Congress on Engineering 2011 Volume I is the property of International Association of Engineers (IAENG) and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.) | |
| Database: | Engineering Source |
Be the first to leave a comment!
