Developing embedded parallel image processing applications is usually a very hardware-dependent process, often using the single instruction multiple data (SIMD) paradigm, and requiring deep knowledge of the processors used. Furthermore, the application is tailored to a specific hardware platform, and if the chosen hardware does not meet the requirements, it must be rewritten for a new platform. We have proposed the use of design space exploration [9] to find the most suitable hardware platform for a certain application. This requires a hardware-independent program, and we use algorithmic skeletons [5] to achieve this, while exploiting the data parallelism inherent to low-level image processing. However, since different operations run best on different kinds of processors, we need to exploit task parallelism as well. This paper describes how we exploit task parallelism using an asynchronous remote procedure call (RPC) system, optimized for low-memory and sparsely connected systems such as smart cameras. It uses a futures [16]-like model to present a normal imperative C-interface to the user in which the skeleton calls are implicitly parallelized and pipelined. Simulation provides the task dependency graph and performance numbers for the mapping, which can be done at run time to facilitate data dependent branching. The result is an easy to program, platform independent framework which shields the user from the parallel implementation and mapping of his application, while efficiently utilizing on-chip memory and interconnect bandwidth.

In order to make the implementation of network components flexible and cost effective, it is required to use widely available technologies as the implementation platform. The distributed operating systems can be adopted as such a platform, because they allow to implement a network component using multiple computers connected through a local area network. In this paper, we focus on the Intelligent Network (IN) whose network components are modelled as Functional Entities (FEs), and describe an implementation method of FEs using distributed operating systems. Our method is summarized as follows: The remote procedure call (RPC) is used for the access transparent inter-process communication. The lightweight process mechanism is used for handling concurrent requests. CCF/SSF (Call Control Function/Service Control Function) and SDF (Service Data Function) are implemented as an SSF server and an SDF server, respectively. SCF (Service Control Function) is composed of a Service Dispatcher and a set of Service Executors. The Service Dispatcher accepts all the requests for IN call processing and dispatches them to appropriate Service Executors. Service Executors are created for the individual IN services and execute the service logics. SDF server and Service Executor may be replicated for load partitioning.This paper has also described the implementation of experimental system supporting "Freephone" service based on our method, and showed the performance evaluation of the experimental system in terms of the real-time and concurrent call processing of IN services. We used Mach and SUN OS as a platform for implementing the servers for FEs. The experimental system using four workstations shows that it can handle up to 170IN calls in one second with the additional response time of less than 200msec, which is small enough compared with the response time for the basic connection control. Those results prove that our method is feasible for implementing practical FEs.