02-10-2005 10:51 AM
I am looking for cases where there is built-in load imbalance between threads (unsolvable by such techniques as dynamic allocation).
Can anyone point me to programs/benchmarks that have this type of behavior?
(Functional or data-pipelined decomposition could have examples of this. In these decompositions, each thread is doing different work at certain parallel regions, so one thread typically finishes its work for that region before for the other.)