The Parameterized Linear Array with a Reconfigurable Pipelined Bus System (LARPBS(p)) model [1] [2] (2005) is the most recent model published that incorporates a shared optical bus communication network. There are two principal features of the LARPBS(p) model that distinguish it from the earlier optical bus models: firstly, its bridging model definition similar to that of the BSP model enables a more detailed algorithm cost analysis than the other models, and secondly, the incorporation of several new communication primitives not defined in earlier models enables possible new algorithms. Historically, the early work on this project, 1999-2005, emphasized systems related aspects of optical bus parallel computer models in general, and the LARPBS (1996) model in particular. Since 2005, all subsequent research relates to systems aspects of the LARPBS(p) model. Collectively, this project has had three journal and six conference publications with additional manuscripts in preparation.

There are two principal publications. The first [1] (2005) introduces the LARPBS(p) model, details the principal model aspects, and proposes two matrix multiplication algorithms together with the associated cost analysis. The recent publication [2] (2008) summarizes the LARPBS(p) model and discusses details of all aspects of the current and on-going systems related research. These aspects are illustrated in the following figure.

LARPBS(p) Systems Research Work, see [2]

• Overview and Survey: An early overview of optical bus parallel computing models was published in [3]. A detailed survey of many optical bus parallel computing models appears in [4].,

• Architecture: The LARPBS(p) model defines a virtual machine abstraction which is similar (but which may be different) as many of the other optical bus models. It is common to consider a fiber-optic implementation of this machine abstraction. We have completed an optical power budget feasibility study for the fiber-based implementation. We have also considered the technical and architecture design of free-spaced-based implementations. And, we have considered extensions to space-based systems. Other publications include the following: preliminary details of the optical power budget study [5], further details of the free-space-based implementation [6], and further details of the modeling of free space systems incorporating computation, communication, command and control (C4) aimed at spacecraft and satellite cluster platforms [7].

• Communication The LARPBS(p) model defines its routing and addressing as based on the well-known coincident pulse technique. We have performed a preliminary analysis of the communication properties for this technique and combined with network traffic modeling, we have considered the difficult issues involved in optical bus collision detection and avoidance. We have proposed a simple collision avoidance algorithm. Other publications include the following: an initial network traffic analysis is discussed in [8].

• Algorithms: All of the algorithms defined for models similar to the LARPBS(p) are SIMD in nature, and where the virtual machine abstraction is similar to the LARPBS(p), are expected to be portable to our model. However, the LARPBS(p) virtual machine abstraction does not preclude MIMD algorithms. A MIMD algorithm for the farming (master-slave) parallel/distributed model has been published. This is significant for two reasons, first, it is (to our knowledge) the first MIMD-type algorithm published for optical bus models, and second, algorithms developed for the farming model could be ported to the LARPBS(p) model thereby enabling additional MIMD algorithms. Other publications include: A detailed analysis of a well known SIMD algorithm appears in [9].