A hardware-driven profiling scheme for identifying program hot spots to support runtime optimization

摘要: This paper presents a novel hardware-based approach for identifying, profiling, and monitoring hot spots in order to support runtime optimization of general purpose programs. The proposed consists set tightly coupled hardware tables control logic modules that are placed the retirement stage processor pipeline removed from critical path. features design include rapid detection program after changes execution behavior, runtime-tunable selection criteria spot detection, negligible overhead during application execution. Experiments using several SPEC95 benchmarks, as well large WindowsNT applications, demonstrate promise design.