A Programming Model for Precise Computation Control satisfying Application specfic semantics

Abstract

Many applications require specialized semantics for precise computation con- trol of its components. The standard Linux kernel does not provide this since application semantics do not always map onto priority based scheduling and also because when a system is under high load, applications can be aected by the high CPU contention which makes the scheduler to not choose a particular ap- plication. This thesis proposes a customized programming model that provides a flexibly configurable scheduling framework for expressing application semantics as the solution. This programming model is called the State and Control variable programming model (SCPM). State and Control variable programming model (SCPM) allows the representation of the application threads and its semantics as part of the scheduling algorithm. In order to demonstrate such a programming model, this thesis uses a video pipeline application as its driving application.