Compiler Data Dependence Analysis via Static Single Assignment Graphs

The patent segments the complex memory area analysis into distinct components by introducing memory descriptors that separate define information, use information, and kill set information. This segmentation allows the compiler to analyze data dependence relationships in a structured manner, reducing the difficulty of detection and measuring while avoiding wasteful computation through targeted analysis of specific memory regions.

The patent introduces memory descriptors as intermediary structures that mediate between instructions and memory areas. These descriptors act as intermediaries that capture and organize information about memory definitions, uses, and kills, enabling efficient data dependence analysis without directly analyzing complex memory area overlaps, thus reducing both analytical difficulty and computational resource waste.

The patent performs preliminary action by pre-computing and storing define information, use information, and kill set information in memory descriptors during the control flow graph construction phase. This preliminary organization of memory analysis data enables precise data dependence relationships to be determined later without repeating expensive computational operations, thus maintaining high measurement precision while reducing computational overhead during execution.

The patent changes the parameters of memory analysis by transforming detailed memory area definitions into discrete memory descriptor parameters (define sets, use sets, kill sets). This parameter transformation allows precise tracking of data dependencies through set operations rather than continuous memory area analysis, achieving high precision with reduced computational overhead through efficient set-based calculations.