Directory/Cache API for Sharing Data in Distributed Memory Systems

The transition to the Exascale predicts the appearance of more complex architectures. Such systems may have a huge number of multi-core nodes, deep memory hierarchies and complex interconnect topologies. Efficiently programming them requires new adequate programming models and currently, the task-based models are perceived as a very promising candidate to start with; they abstract the notion of parallelism from the application developer and offer better perspectives for efficiently exploiting heterogeneous architectures. One of the tasks defined in the INTERTWinE project funded by the European Commission is the definition of a common, generic API for a directory/cache service for task-based runtime systems (like OmpSs, StarPU, etc.). The main purpose of the directory/cache is to provide a set of services that allow task-based runtime systems to efficiently run distributed applications, while being able to consistently manage data stored in distributed memory or in local caches. The directory/cache API allows the task-based runtimes to be completely independent from the physical representation of data and from the type of storage used, facilitating the access through the same interface to an extendable list of memory segment implementations (GASPI, MPI, etc). Moreover, applications may also use the directory/cache API directly. In this poster we try to highlight some of the relevant aspects related to the resulted directory/cache API and the architectural concept.