Exploiting symmetries for optimal integrated code generation

High quality code generation for irregular architectures, such as Digital Signal Processors, is a complex task, as it requires a fully integrated approach that simultaneously deals with the most important tasks of instruction selection, register allocation and instruction scheduling. We show how a fully integrated approach for optimal code generation based on dynamic programming can benefit from exploiting symmetry properties of dependence graphs that decrease the combinatorial explosion of the solution space. We identify conditions when it suffices to explore only one of multiple possible branches that would lead to equivalent results with respect to code quality.