Different math results in openMP depended on number of processes

Georg Bisseling (Intel)
August 21, 2008 7:34 AM PDT
Rate
 
#1

The effect you observe has nothing to do with threading as such.

If you sum up floating point numbers then the result can depend heavily on the exact sequence of operations. When summing up an array of such diverging numbers it is a common practice to sort the array by ascending magnitude and then sum it up to minimize roundoff errors.

The different scheduling for the reduction in the program of course changes the sequence of operations.

I recommend http://docs.sun.com/source/806-3568/ncg_goldberg.html for a much better explanation than I could give.

 

Forum Statistics

4476 users have contributed to 24010 threads and 69943 posts to date.
In the past 24 hours, we have 41 new thread(s) 210 new posts(s), and 263 new user(s).

In the past 3 days, the most popular thread for everyone has been Catastrophic error The most posts were made to Intel License Manager for FlexLM under CentOS 5.2 The post with the most views is You can report them here if

Please welcome our newest member santib90