Jeff Squyres: MPI Progress

En el blog de Jeff Squyres hay otro interesante post sobre cómo se implementa MPI “por dentro”. Esta vez sobre la implementación de las funciones asincrónicas:

Many MPI implementations do not have fully asynchronous progress for all cases.  For example, having a progress thread running in the background can (severely) negatively impact latency and/or resource consumption.  For example, an asynchronous thread will require locks into the critical message passing code paths, potentially thrash caches, incur context switching costs, consume more resources, etc.  The effects of progress threads — particularly for short messages — are… complex, at best.