Method for parallelizing software graphics geometry pipeline rendering

Abstract

In a multiprocessor system, such as a graphics data processing system, a method is disclosed for processing input data in a data processor pipeline. The method includes steps of operating a main thread to store input data in an input buffer until the input buffer is full or the input data ends. If a child thread does not exist, the method creates a child thread and assigns the input buffer to the child thread for processing. If a child thread already exists, the method determines if the child thread can be assigned the input buffer and, if so, then assigns the input buffer to the child thread for processing. If the child thread cannot be assigned the input buffer for processing, the main thread assigns the input buffer to itself for processing in parallel with processing being performed by the child thread. The steps of assigning and determining employ local variables that are accessible to both the main thread and the child thread, and that do not require an operating system call to change and/or test. The method includes a further step of sending a processed data buffer to a subsequent processing stage of the pipeline, such as a rasterizer unit. In a preferred embodiment the step of sending is executed only by the child thread, whether the data buffer was processed by the child thread or by the main thread.