Performance Analysis Of Heap Managers Using Binary Instrumentation

Abstract

The scope of this thesis was the development of a performance analysis methodology of heap managers using Pin from Intel, a dynamic binary instrumentation framework. This thesis focuses on fragmentation and time cost metrics, though it provides a generic framework through which additional measurements can be derived. The code has been tailored to the specific behavior of the regular glibc memory manager as well as NTUA / ICCS dmmlib, a dynamic memory management framework.