Talk:Bulk synchronous parallel

possible mistake?

I just read Leslie Valiant's original paper "A Bridging Model For Parallel Computation" in it, in the section "The BSP Model" the author writes :

"A computation consists of a sequence of supersteps. In each superstep, each component is allocated a task consisting of some combination of local computation steps, message transmissions and (implicitly) message arrivals from other components. After each period of L time units, a global check is made to determine whether the superstep has been completed by all the components." Therefore, i think it is very clear that there are two phases and not three:

phase 1 (L time steps) local computations and message passing
phase 2 Barrier check

Vflouris (talk) 07:56, 14 March 2012 (UTC)vflouris[reply]

You're right. The supersteps have three components, but the computation & communication are not necessarily separated in time. Eijkhout (talk) 16:05, 20 August 2012 (UTC)[reply]

barriers

Global barriers are really a bad idea; have been for 20 years, and are becoming more so. I referenced a paper by myself showing work on methods that try to remove synchronization points from algorithms. If this is objectionably I can find other references. Eijkhout (talk) 16:26, 20 August 2012 (UTC)[reply]