Shader serialization and instance unrolling

Abstract

A graphics engine with shader unit thread serializing and instance unrolling functionality that executes multi-threaded shader logic in a single hardware thread is described. Hardware accelerated tessellation functionality is implemented utilizing programmable pipeline stages that allow custom, runtime configuration of graphics hardware utilizing programs compiled from a high level shader language that are executed using one or more shader execution cores. In one embodiment, multiple shader unit program threads are serialized to run in one hardware thread to allow a greater number of instructions to be executed on the shader cores and preserve hardware threads for primitive processing by other shader units.