Communications of the ACM
Mastering Multicore
Credit: Christine Daniloff / MIT News
Massachusetts Institute of Technology (MIT) researchers have developed software that makes computer simulations of physical systems run more efficiently on multicore chips. The system can break a simulation into much smaller chunks, which it loads into a queue. When a core finishes a calculation, it moves onto the next chunk in the queue, which saves the system from having to estimate how long each chunk will take to execute. Additionally, smaller chunks mean that the system can better handle the problem of boundaries.
The management system can divide a simulation into chunks that are so small that they can fit in the cache along with information about the adjacent chunks. So a core working with one chunk can rapidly update factors along the boundaries of adjacent chunks. Using existing management systems, the MIT team found that a 24-core machine ran 14 times faster than a single-core machine. However, the new management system ran the same machine 22 times faster.
The new system would allow individual machines within clusters to operate more efficiently as well.
From MIT News
View Full Article
Abstracts Copyright © 2010 Information Inc., Bethesda, Maryland, USA 
No entries found