Multi-model parallel programming in psyche

Abstract

Many different parallel programming models, including lightweight processes that communicate with shared memory and heavyweight processes that communicate with messages, have been used to implement parallel applications. Unfortunately, operating systems and languages designed for parallel programming typically support only one model. Multi-model parallel programming is the simultaneous use of several different models, both across programs and within a single program. This paper describes multi-model parallel programming in the Psyche multiprocessor operating system. We explain why multi-model programming is desirable and present an operating system interface designed to support it. Through a series of three examples, we illustrate how the Psyche operating system supports different models of parallelism and how the different models are able to interact.

Uniprocessor virtual memory without TLBs

“…SunOS aligns shared pages on extremely large virtual boundaries to ensure that aliases map to the same cache block [7]. 1 Single address space operating systems (SASOS) such as Opal [5] or Psyche [35] eliminate the need for virtual-address aliasing entirely: All shared data is referenced through global addresses, allowing pointers to be shared freely.…”

“…The sharing mechanism is defined by the page table. One can simplify virtual cache management by sharing memory via global addresses, a scheme used in many systems [5], [35] and shown to have good performance. Alternatively, one could share memory through virtual-address aliasing.…”