David N. Wilner

Software considerations for real-time RISC

Although reduced-instruction-set-computer (RISC) computers promise extremely high computing speeds, they present important and difficult challenges to the software designer, particularly for real-time use. With the SPARC microprocessor and the VxWorks implementation for that processor used as examples, some of those challenges are described and solutions offered. The most significant recent evolution in real-time computing is the shift from complex-instruction-set-computer (CISC) to RISC architecture. The minimalist RISC architecture runs very fast by running very simply and minimizing interactions outside of the CPU. This allows optimizations such as extensive pipelining and parallel instruction execution, which are difficult to implement in CISC machines. Also, although CISC machines perform many complex functions in hardware, RISC machines off-load these functions to software. Software can optimize these complex functions, thereby avoiding the use of microcoded instructions, which may actually be slower. Although RISC was not developed specifically for real time, RISC chips can perform extremely well in a real-time environment.<<ETX>>

Multiprocessing: an extension of distributed, real-time computing

The treatment of multiprocessing in VxWorks is described. In multiprocessing, two or more CPUs are connected to communicate, share data, and cooperate on an application. VxWorks treats multiprocessing as an extension or special case of distributed control. VxWorks allows loosely coupled communications in a multiprocessing system, as well as tightly coupled communications. The different communications strategies can be examined on two levels-the low-level hardware connection and the high-level software connection model. Software provides applications with a model for sending data. Data packets or data streams are the models which provide simple data transfer for loosely coupled communications. The VxWorks approach treats multiprocessing as an extension of a distributed system since both require the same facilities-process-to-process communications, remote procedure calls, and shared objects.<<ETX>>