Content

•  
   Table of contents
•  
   1 Introduction
•  
  1.1 Current parallel programming standards
•  
   1.1.1 Polling in non-trivial parallel applications
•  
  1.2 Photorealistic image synthesis
•  
   1.2.1 Measures of photorealism
•  
   1.2.2 Photorealistic rendering systems
•  
   1.2.3 Light phenomena and their simulation
•  
   1.3 Outline of this thesis
•  
   2 Event-driven message passing
•  
   2.1 Non-trivial parallel applications
•  
   2.2 Development of parallel programming
•  
   2.2.1 Occam programming language
•  
   2.2.2 Transputer
•  
   2.2.3 Occam, Transputer and non-trivial parallel applications
•  
   2.3 Current message passing standards: PVM and MPI
•  
   2.4 Point-to-point message passing in PVM and MPI
•  
   2.4.1 Message assembling and sending
•  
   2.4.2 Message receiving and disassembling
•  
   2.5 Unifying framework for message passing
•  
   2.5.1 Components of the message passing framework
•  
   2.5.2 Application process
•  
   2.5.3 Basic message passing operations
•  
   2.5.4 Message passing system
•  
   2.5.5 Language binding
•  
   2.5.6 Operation binding
•  
  2.6 Threaded non-trivial PVM and MPI applications
•  
   2.6.1 Threads and thread-safety
•  
   2.6.2 Polling in threaded non-trivial PVM and MPI applications
•  
   2.6.3 Polling in communication libraries
•  
   2.6.4 Limits of active polling
•  
   2.6.5 Previous work related to thread-safety of PVM and MPI
•  
   2.6.6 Quasi-thread-safe PVM and MPI
•  
   2.6.7 Towards a complete thread-safety of PVM and MPI
•  
   2.7 TPL: Event-Driven Thread Parallel Library
•  
   2.7.1 Concept
•  
   2.7.2 Process startup and termination
•  
   2.7.3 Thread management
•  
   2.7.4 Message passing
•  
   2.7.5 Message handling and message callbacks
•  
   2.7.6 Message packing and unpacking
•  
   2.7.7 Error handling and debugging
•  
   2.7.8 Flow control
•  
   2.8 Efficiency benchmarks
•  
   2.8.1 ONE-SIDED THREADED PINGPONG
•  
   2.8.2 SYMMETRICAL THREADED PINGPONG
•  
   2.8.3 Summary of benchmarking results
•  
   2.9 Conclusions
•  
   2.9.1 Overlapping of Communication and Computation
•  
   3 Global illumination
•  
   3.1 Physics of light
•  
  3.2 3D modeling
•  
   3.2.1 Modeling of colour spectrum
•  
   3.2.2 Modeling of surface geometry
•  
   3.2.3 Modeling of surface materials
•  
   3.2.4 Modeling of light sources
•  
   3.2.5 Modeling of camera
•  
   3.3 The global illumination problem
•  
   3.3.1 Rendering equations
•  
   3.4 Approaches to the global illumination problem
•  
   3.4.1 Direct methods
•  
   3.4.2 Approximation methods
•  
   3.5 Conclusions
•  
   4 Ray tracing
•  
   4.1 The basic ray tracing algorithm
•  
  4.2 Sequential optimisation techniques
•  
   4.2.1 Bounding volumes
•  
   4.2.2 Bounding slabs
•  
   4.2.3 Light buffers
•  
   4.3 Persistence of Vision Ray Tracer
•  
  4.4 Parallel ray tracing
•  
   4.4.1 Existing approaches
•  
   4.4.2 Image space subdivision
•  
   4.4.3 Setting of parameters in the perfect load balancing algorithm
•  
   4.4.4 Distributed object database
•  
   4.4.5 Experiments
•  
   4.4.6 Further extensions and improvements
•  
   4.5 Conclusions
•  
   5 Radiosity
•  
   5.1 Southwell relaxation
•  
   5.1.1 Shooting radiosity algorithm
•  
   5.2 Form factor computation
•  
   5.2.1 Monte Carlo form factor computation
•  
   5.3 Discretisation of surface geometry
•  
   5.4 Illumination storage and reconstruction
•  
   5.5 Energy transfer
•  
   5.5.1 Shooting radiosity algorithm using the ray tracing shader
•  
   5.6 Visualisation
•  
  5.7 Experiments
•  
   5.7.1 Form factors
•  
   5.7.2 Experiments with the box scene
•  
   5.7.3 Experiments with large scenes
•  
   5.8 Conclusions
•  
   6 Summary
•  
   6.1 Towards portable 3D standards
•  
   A MPI progress rule tester
•  
  B Threaded pingpong benchmark
•  
   B.1 TPL 2.0
•  
   B.2 PVM 3.4
•  
   B.3 MPI (MPI 1, MPI 2)
•  
   List of figures
•  
   Bibliography