The Research On The Visualization In Scientific Computing Of Remote Real-Time Tracing Visualization For The Parallel Application Program

Tracing and steering technologies is one of more important research fields of current Visualization in Scientific Computation. The remote real-time visualization technology for parallel programs provides support for on-line graphical display of a large amount of data produced by numerical simulation, and it allows programs to be steered during their execution. It can help end users observe and control the computing process effectively.In this paper, we discuss the remote real-time tracing visualization technology for parallel programs, and introduce an efficient Remote Visualization System (RVS). With little modification of the parallel application, users will be able to select dynamically data produced by a computing program for visualization on a graphics workstation as the program executes on a remote High-Performance Parallel Computer. Example solutions from this solver are presented to show that the RVS also offers an environment for effective remote real-time tracing visualization for the parallel program.In Chapter 1, we introduce the conception of Scientific Computing Visualization, its backgrounds and application domains, and the key role it plays in scientific researches. Meanwhile, we summarize some popular visualization system, including their shortcomings. In the end, we propose the model of Remote Visualization System oriented Parallel Program in the distributed visualization environment.In Chapter 2, we introduce system's architecture structure, server versus client functionality and client-server communication, and propose scientific computing visualization environment of RVS. RVS's design goal is to implement remote real-time tracing visualization based on minimum of program perturbation. RVS is a Network-oriented distributed visualization system in Client/Server mode in which the data and the image are at different computers at possibly different locations connected by a network. The program executes on a remote High-Performance Parallel Computer, and the client acts as full visualization.In Chapter 3, we introduce the design and implementation of RVS, including software implementation of application-supported module, visualization processing module and data transmission module. We illuminate data structure of field object, memory object, communication object and publisher object after that. Meanwhile, we also explain here how to call API to implementation.In Chapter 4, we expatiate the implementation theories of the remote access techniques for distributed data, and describe some important techniques including share data collecting in parallel environment, the position indices of distributed variables and scheduling synchronized memoryaccess.In Chapter 5, the sample application highlighted in this paper is Lared-I, a parallel program. We implement memory object's construction of Grid, Electron temperature, ion temperature and photon temperature by calling appropriate APIs. By adding remote on-line tracing function into General Graphics visualization system, we realize remote visualization of Grid, Electron temperature, ion temperature and photon temperature.In Chapter 5, we describe System Evaluation. It includes the time analysis of creating memory object and field object, the time analysis of remote access fixed scale data, and the time analysis of remote access un-fixed scale data. Meanwhile, we discuss monitoring latency due to conversion of byte-ordering.At end of the dissertation, the author draws a conclusion and puts forward some key problems and developing trends for future works.