A robust stack folding approach for Java processors: an operand extraction-based algorithm

Data dependency in stack operations limits the performance of Java processors. To enhance Java's performance, existing literature suggests using stack operations folding. We extend this concept in a new folding algorithm that identifies principle operations in folding groups and extracts necessary operands from the bytecode queue. The proposed algorithm permits nested pattern folding and multiple issue of folding groups. Hence, the need for and therefore the limitations of a stack are eliminated. This paper discusses various aspects of the proposed algorithm and illustrates different folding scenarios as well as possible hazards. Benchmarking using SPECjvm98 shows excellent performance gains as compared to existing algorithms.