This document presents our main contributions to the field of compilation, and more generally to the quest of performance of computing systems. It is structured by type of execution environment, from static compilation (execution of native code), to JIT compilation, and purely dynamic optimization. We also consider interpreters. In each chapter, we give a focus on the most relevant contributions. Chapter 2 describes our work about static compilation. It covers a long time frame (from PhD work 1995--1998 to recent work on real-time systems and worst-case execution times at Inria in 2015) and various positions, both in academia and in the industry. My research on JIT compilers started in the mid-2000s at STMicroelectronics, and is still ongoing. Chapter 3 covers the results we obtained on various aspects of JIT compilers: split-compilation, interaction with real-time systems, and obfuscation. Chapter 4 reports on dynamic binary optimization, a research effort started more recently, in 2012. This considers the optimization of a native binary (without source code), while it runs. It incurs significant challenges but also opportunities. Interpreters represent an alternative way to execute code. Instead of native code generation, an interpreter executes an infinite loop that continuously reads a instruction, decodes it and executes its semantics. Interpreters are much easier to develop than compilers, they are also much more portable, often requiring a simple recompilation. The price to pay is the reduced performance. Chapter 5 presents some of our work related to interpreters. All this research often required significant software infrastructures for validation, from early prototypes to robust quasi products, and from open-source to proprietary. We detail them in Chapter 6. The last chapter concludes and gives some perspectives.