Run-time detection mechanism of nested call-loop structure to monitor the actual execution of codes

摘要

To enhance dependability is becoming hard because emerging multicore processors and ever-increasing code size of software make a computer system larger and more complex year by year. In this circumstance, how to handle and make effective use of multicore processors is essential for tackling these issues, and a mechanism that monitors the dynamic behavior of code execution is required to realize it. In this paper, we present a profiling scheme of nested call-loop structure using dynamic binary instrumentation. We evaluate our profiling scheme using an application benchmark, and show the advantages of our profiling compared with conventional profiling. We also show that our profiling can monitor the actual dynamic behavior of an application and detect the nested loop structures and their relative locations toward function calls. The results also indicate that our scheme has the ability to be adaptable for performance tuning and optimization process of multicore systems.