The current trends in computer architecture suggest that in the future it will be common to use computer networks where some of the nodes are multiprocessors - either shared memory multiprocessors or distributed multiprocessors. In such an environment a parallel application need not be confined to a single parallel machine. Instead it can be composed of several components each executing on a different node of the network to yield the best possible results. Each such component can be viewed as a shared resource providing services to the other components of the parallel application and the interactions between the components can be implemented as remote procedure calls. The issues in programming the remote components of a parallel application as shared resources are similiar to the issues in programming shared resources in general such as file servers and database systems. In both cases it is important to implement a shared resource as a parallel system so that the speed-up possible in parallel applications is not compromised due to delays in accessing the resource. This project is to develop an object oriented parallel programming system called ALPS using which a parallel application can be implemented efficiently on the multiple nodes of a network of parallel machines. The ALPS programming systems consists of a strongly typed high-level language and a kernel that forms the run-time support for the ALPS language and which can also be used from other languages such as C, Pascal and Fortran. The ALPS system is currently being implemented on 16-node transputer network. In the future it is plan to implement it on a 12-processor Encore Multimax.
