Machine Code Optimization by Integer Linear Programming

Embedded systems are computer systems that interact with a physical
environment for measuring and controlling purposes. Typical
requirements are high performance, but also low cost and low power
consumption. Embedded processors are characterized by specialized,
irregular hardware architectures which make the task of generating
efficient machine code from high-level languages a difficult task.
Classical heuristical code generation algorithms usually fail to
generate code of satisfactory quality. Integer linear programming
techniques allow a homogeneous problem description for multiple code
generation tasks to be developed which can be used to compute optimal
solutions. Moreover, they can be used in generic systems where
optimizers for specific hardware architectures are generated from a
processor specification. This talk will give an outline of the code
generation problem, the integer programming modelling and its
application in a generic code optimization framework.


Alle InteressentInnen sind zu dem Vortrag herzlich eingeladen.

Die Veranstaltung wird gemeinsam von der Fachrichtung Informatik und der
International Max Planck Research School getragen.

---------------------------------------------------------------------------------------