Improvements to object-oriented programming methods allow programmers to direct execution of dynamically generated, parallel computing tasks on variety of platforms. Data objects provide their data location information handle, plurality of which can be combined by arithmetic operations, thus providing application specific, computing platform non-specific method to inform executing environment about data needed for each programming object. During execution this information is used to produce platform specific data structure to aid in optimal processing node assignment. A new calling method greatly reduces data exchanges between objects running as separate tasks. These improvements are particularly well suited to programming method and computing machine presented by parent of this divisional patent application. As parent invention utilization results in dynamic allocation of large number of tasks, assignment of processing nodes at runtime to those tasks, load balancing with various distributed resource access schemes are critical problems to solve.