[0103] Next, an embodiment of the distributed information processing is described below where an application with an unpredictable magnitude of load is image processing.
[0104] Image processing includes the afore-described geometry processing and rendering processing. Geometry processing defines a shape of a model (object) in an absolute coordinate space by parameters of an equation that represents vertex coordinates, and boundary lines and/or planes, and converts the defined model to a viewpoint-referenced screen space; it performs, among others, 3-D space coordinate transformation, lighting, and clipping. Rendering processing renders lines or planes on the image memory based on the data generated by the geometry processing, to produce image expression of the model of interest.
[0105] Because geometry processing is enormously compute-intensive, it is suitable as an application that is load balanced by a plurality of computers.
[0106] It is presumed here that image processing is intended to create data for displaying a model as shown in FIG. 8 onto a display screen 50. The model is created based on mesh data; for geometry processing, coordinate transformation is performed to vertex data on grid points as shown in FIG. 9.
[0107] Assuming that the main computer 1m has four sub processors 14, as shown in FIG. 2, each of which acts as a cluster element, respectively; each of the sub processors 14 is capable of performing coordinate transformation of 17.times.17 vertices including four blocks A of 8.times.8 data per unit time, then the data to be processed is worthy of six blocks, as shown in FIG. 9; thus, when the data is processed by the main computer 1m in one unit time, two blocks worth of processing capacity is deficient. Accordingly, it is necessary to consign geometry processing related to blocks C and D to other available computers 1c.
[0108] Then, the main computer 1m finds available computers 1c to balance the load in accordance with the following procedure....
[procedure not quoted]
