No label defined (Q4392293)

From MaRDI portal

!

This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:

scientific article; zbMATH DE number 1160045
Language Label Description Also known as
default for all languages
No label defined
    English
    No label defined
    scientific article; zbMATH DE number 1160045

      Statements

      0 references
      8 June 1998
      0 references
      parallel programming
      0 references
      distributed shared memory
      0 references
      Rthreads
      0 references
      Parallel programming is divided into two distinct paradigms: the structured-memory paradigm as used in multiprocessor machines and the message-passing paradigm as used in distributed systems. In the latter, every processor has its own local memory and communicates by sending messages through a network. In coding algorithms for a distributed system, it is the responsibility of the programmer to ensure that every node will be in possession of the necessary data at the correct time. This is why the shared-memory paradigm presents less of a departure from sequential programming habits. In Distributed Shared Memory (DSM) systems, an additional software layer in a distributed system emulates a shared memory, so that the programmer is relieved of duties regarding the distribution of data. In order to reduce the communication overhead, a DSM system employs local buffers; the consistency model describes when the state of these is consistent with the remote data. Problems with existing DSM systems are: static consistency models take a defensive approach to buffer updating, entailing a communication overhead that is still too high; heterogeneous networks are not supported; portability is lacking. The present dissertation describes Rthreads, a DSM system that allows memory accesses to be organized in a flexible way according to the needs of each separate algorithm, thus reducing the communication overhead to the most. In particular, Rthreads provides special access methods that allow combining data objects in dynamically changing groups. Rhtreads supports heterogeneous networks; multiprocessor workstations are seamlessly integrated because Rhthreads is compatible with POSIX local threads. High portability is achieved by implementing network communication on top of well-known existing communication systems.
      0 references

      Identifiers