With the arrival of multicore chips as the commodity architecture for a wide range ofplatforms, there is a growing pressure to make parallel programming the norm. Unfortunately,most current programmers find parallel programming too complex. Therefore,we need tools, models, and architectures that make multiprocessors more programmable.One compelling way to improve programmability is to enable back-and-forth timetravel of multiprocessor execution. Such ability simplifies parallel code debugging, and ispossible using a technique called Deterministic Replay of Execution. This thesis presentsDeLorean, a novel hardware substrate for deterministic replay of multiprocessor systems.DeLorean advances the state of the art in that it enables high-speed recording and replayof parallel execution and in that its space requirements are orders of magnitude smallerthan those of current schemes.To be practical, DeLorean and other hardware replay systems need to support anenvironment with multiple parallel jobs running concurrently — some being recorded,others being replayed and possibly many others running without recording or replay. Toaddress this issue, this thesis presents Capo, a novel software-hardware interface for practicalhigh-speed multiprocessor replay. It also introduces the novel abstraction of ReplaySphere to separate the responsibilities of the hardware and software components of a replaysystem. Finally, this thesis describes CapoOne, a prototype of a deterministic multiprocessorreplay system that implements Capo using Linux running on simulated DeLorean hardware.
PDF
download
878987797
028-85220240
