Operating Systems in Silicon and the Dynamic Management of Resources in Many-Core Chips

Abstract

This discussion paper explores the problems of operating systems support when implementing concurrency controls at the level of the instruction set in processors designed for multi- and many-core chips. It introduces the SVP model and its implementation in DRISC processors to a level of detail required to understand these problems. The major contribution of the paper is in analysing the issues faced in porting operating system functionality onto such processors. The paper covers the issues of job scheduling, dynamic resource management, memory protection and security. It provides examples in μTC (a language based on the SVP model) of how resource management and security issues are managed. It concludes that the implementation of concurrency controls in a processors instruction set is very disruptive. However, the author sees no alternatives if mainstream computing is ever to be served effectively by multi- and many core processors.

Concurrency engineering

“…Our use of this term is more abstract as we approach the algorithm from the perspective of capturing its specification, which may include concurrent operations but independently of any implementation issues, which make the algorithm non-portable. We believe that low-level solutions are required in order to eliminate, as far as is possible, the experimentation from algorithm engineering but this requires systems to be designed from the ground upwards, namely from the processor's ISA, which must abstract and embed explicit concurrency and manage this in a dynamic manner [6]. We will return to this in due course (see Section 5).…”

“…SVP has its roots in the microthread machine model originally proposed by Bolychevsky et al [26] and extended by Jesshope in [6,7]. It is a general model of concurrency with implementations at the ISA [8] and language [9][10] levels.…”

“…This may be compiled to an implementation, e.g. a microgrid of microthreaded processors [6], without knowledge of the number of processors to be used in executing it. This usage is deterministic and binary programs can be combined concurrently without inducing deadlock and distributed arbitrarily to clusters of processors at run time.…”