Machine code optimization - improving executable object code

the long version

In many program development environments, certain significant optimization techniques cannot be performed on any program representation prior to the representation executed by the hardware. This thesis explores optimization techniques at the machine code level. These techniques are shown to yield a significant decrease in code space and a more modest improvement in execution speed. As a result, this dissertation describes two aspects of this area in parallel. It explores analysis and optimization techniques from a theoretical viewpoint. Some of these are new and some are extensions of techniques which have been applied in other phases of the compilation process. In addition, we report on the implementation of a production quality Machine Code Optimizer. This optimizer represents the first time that these techniques have been brought together in this fashion and at this level. The performancce of this optimizer substantiates the expected speed and space improvements on two target architectures.