Font Size: a A A

Hardware-software co-synthesis of distributed embedded systems

Posted on:1997-09-30Degree:Ph.DType:Thesis
University:Princeton UniversityCandidate:Yen, Ti-YenFull Text:PDF
GTID:2468390014483274Subject:Engineering
Abstract/Summary:
Embedded computer systems use both off-the-shelf microprocessors and application-specific integrated circuits (ASICs) to implement specialized system functions. Examples include the electronic systems inside laser printers, cellular phones, microwave ovens, and an automobile anti-lock brake controller. While such systems have been widely applied in consumer products for decades, the design automation of embedded systems has not been well studied. Hand-crafted techniques were sufficient to design small, low-performance systems in the past. Today manual design effort grows rapidly with system complexity and becomes a bottleneck in productivity, so it is important to apply computer-aided design (CAD) methodology for embedded system design.; This thesis presents new algorithms to simultaneously co-design the hardware engine and application software architecture of an embedded system. By hardware engine, we mean a heterogeneous distributed system composed of several processing elements (PE), which are either CPUs or ASICs. By application software architecture, we mean the allocation and scheduling of processes and communication. New techniques such as fixed-point iterations, phase adjustment, and separation analysis are proposed to efficiently estimate tight bounds on the delay required for a set of multi-rate processes preemptively scheduled on a real-time reactive distributed system. Based on the delay bounds, a gradient-search co-synthesis algorithm with new techniques such as sensitivity analysis, priority prediction, and idle-PE elimination is developed to select the number and types of PEs in a distributed engine, and determine the allocation and scheduling of processes to PEs. New communication modeling is presented to analyse communication delay under interaction of computation and communication, allocate interprocessor communication links, and schedule communication.; Distributed computers are often the most cost-effective means of meeting the performance requirements of an embedded computing application. Most recent work on co-design has focused on only one-CPU architecture. This thesis is one of the first few attempts to co-synthesize multi-rate distributed embedded systems.
Keywords/Search Tags:Embedded, Systems, Distributed
Related items