By Topic

Optimizing the Execution of Statistical Simulations for Human Evolution in Hyper-threaded Multicore Architectures

Sign In

Cookies must be enabled to login.After enabling cookies , please use refresh or reload or ctrl+f5 on the browser for the login options.

Formats Non-Member Member
$31 $13
Learn how you can qualify for the best price for this item!
Become an IEEE Member or Subscribe to
IEEE Xplore for exclusive pricing!
close button

puzzle piece

IEEE membership options for an individual and IEEE Xplore subscriptions for an organization offer the most affordable access to essential journal articles, conference papers, standards, eBooks, and eLearning courses.

Learn more about:

IEEE membership

IEEE Xplore subscriptions

4 Author(s)
Dias, R. ; Pontifical Catholic Univ. of Rio Grande do Sul (PUCRS), Porto Alegre, Brazil ; De Rose, C.A.F. ; Gomes, A.T.A. ; Fagundes, N.J.R.

Simulations of statistical models have been used to validate theories of past events in evolution of species. Studies concerning human evolution are important for understanding about our history and biodiversity. However, these approaches use complex statistical models, leading to high computational cost. The present paper proposes optimization techniques for Hyper-threaded multicore architectures to improve the computational performance of these simulations. Combining granularity studies and Hyper-threading optimization, we improved the performance of simulations in more than 30%, if compared with common parallel execution (default parallelization applied by users). The performance was evaluated using a complex example of human evolution studies [1]. For this example, our techniques enable the user to decrease the simulation execution time from 50 days (sequential runtime) to less than 5 days. In addition, the evaluation has been extended for simulations running on multiple multicore cluster nodes. Our measurements show a high Speed-up, close to theoretical maximum, being 129 times faster for 160 computational cores. This represents an efficiency of 81%.

Published in:

Parallel and Distributed Processing Symposium Workshops & PhD Forum (IPDPSW), 2012 IEEE 26th International

Date of Conference:

21-25 May 2012