New Techniques for Adaptive Program Optimization
Sapienza Università Editrice, 30 giu 2020 - 204 pagine
Living organisms are adapted to their environment. Modern compilers and runtime systems for computer software are no different: as part of a continuous optimization process, they can adapt the execution cycle of a program to the workload it operates on.
This thesis brings novel ideas to the software optimization domain. It illustrates methodological and practical contributions that advance the state of the art for performance profiling techniques and adaptive runtime designs, backed by promising experimental results on industrial-strength benchmarks. Part of the results has been presented in flagship programming language venues.