The Meiko Computing Surface

The Meiko Computing Surface was first demonstrated in July 1985 at the SIGRAPH Conference in San Francisco, and became commercially available in the third quarter of 1986. It is a modular and expandable system organised as a reconfigurable array of transputer-based computing elements, I/O elements, and storage elements. These elements are supported by a library of circuit boards, each optimised for a specific function.

A Computing Surface consists of a number of Modules each containing up to 40 boards housed in two 19-inch racks. All inter-board links within a Module are routed via the System backplane, and links between Modules are provided by special inter-Module link boards. There is no theoretical limit on the number of Modules, and hence transputers, that can be contained in a Computing Surface, although the upper bound on the inter-Module link wire length ultimately constrains the configurations which can be extended indefinitely. Modules contain a private power supply and use forced-air cooling, dissipating up to 3.1 kW each. The structure of the Computing Surface is illustrated in the figure.

The system backplane in each Module supports a Supervisor bus as well as the link connectivity, and this provides for low bandwidth communication between all computing elements in the system. It has a single Bus Master, which is nominally a Local Host board, and is capable of resetting and examining the internal state of all transputers, reporting errors, and configuring the link connectivity. Application software can also use the Supervisor bus as a communication pathway, and this could be particularly useful for transmitting debugging information.