SymGrid-PAR
  ===========

  Abyd Al Zain


  The SymGrid-Par middle-ware orchestrates computational algebra
  components into a parallel application. SymGrid-Par components
  communicate using the Symbolic Computation Software
  Composability Protocol, SCSCP, which in turn builds on
  OpenMath, an XML-based data description format, designed
  specifically to represent computational mathematical objects.

  SymGrid-Par forms a key part of the SymGrid framework that
  provides high-performance Grid services. In this session, we
  focus on parallelism and the way we can exploit the
  capabilities of SymGrid-Par to provide a simple parallel API
  for heterogeneous symbolic components. SymGrid-Par builds on
  and extends our Grid-enabled implementation of GUM, a
  lightweight portable parallel runtime environment for the GpH
  extension of the standard functional language, Haskell. GUM
  provides various high-level parallelism services including
  support for ultra-light-weight threads; virtual shared-memory
  management; automatic scheduling; automatic thread placement;
  automatic datatype-specific (un)\-marshalling; implicit
  communication; load-based thread throttling and thread
  migration. It thus provides a flexible, adaptive environment
  for managing parallelism at various degrees of granularity. It
  has been ported to a variety of shared-memory and
  distributed-memory parallel machines, and more recently to
  Globus-based computational Grids.

  SymGrid-Par comprises two generic interfaces: the
  "Computational Algebra system to Grid middle-ware'' (CAG)
  interface links Computational Algebra Systems (CASs) to GUM;
  and the ``Grid middle-ware to Computational Algebra system''
  (GCA) interface conversely links GUM to these systems. The CAG
  interface is used by computational algebra systems to interact
  with GUM. GUM then uses the GCA interface to invoke remote
  computational algebra system functions, to communicate with the
  computational algebra systems etc. In this way, we achieve a
  clear separation of concerns: GUM deals with issues of thread
  creation/coordination and orchestrates the computational
  algebra system engines to work on the application as a whole;
  while each instance of the computational algebra system engine
  deals solely with execution of individual algebraic
  computations.

  During the SymGrid-Par sessions, we will present the ideas
  behind our parallel generic interface and we will practically
  demonstrate this.