Research On Static Single Assignment Optimization Technique Based On GCC

Static single assignment is an optimization compiler technique based on GCC. Lengauer-Tarjan is a fast algorithm for finding dominators in a flowgraph during the implementation of static single assignment. When function EVAL is used in Lengauer-Tarjan, it is required to execute calls and returns during the calling processes, and there are many useless calls to EVAL. To solve these problems, at first, the GCC optimization architecture, the main optimizer's features, the optimization compiler techniques, the optimization work flow, the static single assignment technique, the dominator-tree generation technique and its realizing algorithm Lengauer-Tarjan, the path compressed technique and its realizing function EVAL are analyzed in detail. Then an algorithm to solve the problems, which is called Real-Time Best-Point, is presented. Real-Time Best-Point is proved with validities in theory. The analysis in theory shows that, the Lengauer-Tarjan based on Real-Time Best-Point is more efficient than EVAL-based Lengauer-Tarjan in a good many circumstances.