Abstract This paper describes the implementation of a Prolog‐based, instruction‐level simulator used in the design of a new processor. The adaptability of Prolog as a hardware description language is reported. Prolog was a good choice for purposes of rapid prototyping and as a means to derive high‐level executable specifications for the instruction set. The rapid software development strategy entailed a price to be paid in the form of a relatively slow performance. Methods to improve the speed of Prolog‐based simulators are suggested.

Empirical observations of computer operating systems have shown that operating systems are designed with one of two object oriented strategies: a process or a monitor oriented approach. This paper compares the two design approaches in a Concurrent Pascal environment. Resource manager programs that are implemented in conformity with each model are evaluated using software complexity measures and program performance measures. The average complexity of resource manager processes is 94 percent larger than the average complexity of resource manager monitors. The runtime synchronization overhead of the process model program is two-eight times higher than that of its counterpart.

Abstract Resource manager objects are the operating system entities that manage computer system resources. In this paper we extend Hoare's and Brinch Hansen's monitor concept, and introduce a resource monitor language construct that represents a resource manager object. Resource monitors provide requesting processes with implicit synchronization, and do not require separate processes to perform the resource access operations. The resource monitor construct is compared to current language constructs used to structure operating system software. For this purpose we use an evaluation methodology that combines software complexity measurements with program performance measurements. The evaluation itself is carried out in a Concurrent Pascal‐like programming environment. The current language constructs have a software complexity that is larger by 37 to 219 per cent over the resource monitor's complexity. The run‐time synchronization overhead of programs that use current language constructs is 1.43 to 2.75 times higher than the overhead of programs that use a resource monitor.