1. Large Scale Structure of the Universe Sameshan Perumal and Carl Hultquist

2. Motivation Two major repositories of galaxy information: the Sloan Digital Sky Survey and the 2dF Galaxy Redshift Survey. Two major repositories of galaxy information: the Sloan Digital Sky Survey and the 2dF Galaxy Redshift Survey. Galaxies group to form large-scale structures. Galaxies group to form large-scale structures. By studying these, astronomers hope to develop a better understanding of the devlopment of the universe. By studying these, astronomers hope to develop a better understanding of the devlopment of the universe. Due to large amounts of data, suitable software is needed to conduct these studies. Due to large amounts of data, suitable software is needed to conduct these studies.

3. The “Bath-Sponge” Prototype visualisation of 2dF data

4. Data Representation Image taken from 2dF Movie II

5. Data Representation Potentially huge volume of data to represent Potentially huge volume of data to represent Large overhead to process Large overhead to process Must be efficiently used by OpenGL Renderer Must be efficiently used by OpenGL Renderer Memory constraints must be considered: 100 000 galaxies  256 bytes ≈ 24 Mb Memory constraints must be considered: 100 000 galaxies  256 bytes ≈ 24 Mb

6. Data Representation Use minimal spanning trees Use minimal spanning trees Split space into “neighbourhoods” Split space into “neighbourhoods” Store only critical information Store only critical information Keep coordinate data close to OpenGL standard Keep coordinate data close to OpenGL standard

7. Structure Identification Attempt to extract structure from raw data Attempt to extract structure from raw data Density calculations offer quick approximation Density calculations offer quick approximation Percolating spheres are accurate Percolating spheres are accurate “Grow” sphere of given radius around galaxy “Grow” sphere of given radius around galaxy Compute intersections between neighbours Compute intersections between neighbours Connect intersecting galaxies Connect intersecting galaxies

8. Percolated Spheres

9. Density Calculations Image taken from 2dF Movie II

10. Surface Generation Uses results from Structure Identification Uses results from Structure Identification Construct surfaces around structures Construct surfaces around structures Triangulate surfaces Triangulate surfaces Pass data through to rendering engine Pass data through to rendering engine

11. Surface Generation I Image generated by SurfGen, astro-ph/020136 v1, 6 October 2002.

12. Surface Generation II Image generated by SurfGen, astro-ph/020136 v1, 6 October 2002.

13. Statistical Feedback Use generated surfaces Use generated surfaces Compute various statistical measures Compute various statistical measures Volume, Area, Density Volume, Area, Density Can be used to refine results Can be used to refine results Most useful to compensate for “Diminishing Data Density” Most useful to compensate for “Diminishing Data Density”

14. N-Body Simulations Simulate the evolution of the Universe Simulate the evolution of the Universe Uses finite number of bodies Uses finite number of bodies Evolution occurs within a bounded box Evolution occurs within a bounded box Attempt to analyse data using system Attempt to analyse data using system Possibly run N-Body simulations Possibly run N-Body simulations

15. Graphical User Interface (GUI) Allow for “flying” around the universe. Allow for “flying” around the universe. Clearly depict galaxies, large-scale structures and spheres of percolation. Clearly depict galaxies, large-scale structures and spheres of percolation. Allow for customisation of view: Allow for customisation of view: Choosing which types of objects should be displayed Choosing which types of objects should be displayed Choosing what method should be used to display objects Choosing what method should be used to display objects Allow for individual structures of galaxies to be “selected” and queried for data. Allow for individual structures of galaxies to be “selected” and queried for data. Allow for parameters affecting structure identification to be modified. Allow for parameters affecting structure identification to be modified.

16. GUI ― Aims Usable Usable Accessible Accessible Sound HCI principles Sound HCI principles Realistic and adaptive Realistic and adaptive Shading techniques Shading techniques User immersion User immersion Efficient Efficient Varying levels of detail Varying levels of detail Maintaining a target frame-rate Maintaining a target frame-rate Bill-boarding Bill-boarding Programmable vertex and pixel pipelines Programmable vertex and pixel pipelines

17. GUI ― Usability Accessible Accessible Cross-platform Cross-platform C++ C++ OpenGL OpenGL wxWindows toolkit for dialogs, buttons, etc. that are native to the environment wxWindows toolkit for dialogs, buttons, etc. that are native to the environment Sound HCI principles Sound HCI principles Familiar icons Familiar icons “Intuitive” interface “Intuitive” interface Possibly develop more than one interface (or make interface adaptive) and assess usability by means of a poll amongst several users. Possibly develop more than one interface (or make interface adaptive) and assess usability by means of a poll amongst several users.

18. GUI ― Realism and Adaptivity Shading techniques Shading techniques 3D surfaces 3D surfaces Probability of accurate structure identification Probability of accurate structure identification User immersion User immersion Use of Chromadepth TM Use of Chromadepth TM A pair of ChromaDepth lenses A ChromaDepth image of the Death Valley Earthquake Fault

19. GUI ― Efficiency (1) Varying levels of detail (Clark, 1976) Varying levels of detail (Clark, 1976) Images from Progressive Meshes by Hugues Hoppe

20. GUI ― Efficiency (2) Maintaining a target frame- rate (Funkhouser and Séquin, 1993) Maintaining a target frame- rate (Funkhouser and Séquin, 1993) Fixed level of interactivity Fixed level of interactivity Maximises level of detail by placing an upper bound on the cost of rendering a scene Maximises level of detail by placing an upper bound on the cost of rendering a scene Images that appear larger “contribute” more to the image.

21. GUI ― Efficiency (3) Bill-boarding Bill-boarding Left: explanation of bill-boarding. Courtesy of Lighthouse 3D’s Bill-boarding Tutorial. Right: an prototype developed by us that renders the 2dF galaxy data by drawing each galaxy as a bill-board, using the image of a sphere as a texture.

22. GUI ― Efficiency (4) Programmable vertex and pixel pipelines Programmable vertex and pixel pipelines Vertex shaders Vertex shaders “Morph” continuously between different levels of detail (Southern & Gain, 2002) “Morph” continuously between different levels of detail (Southern & Gain, 2002) General purpose GPU-based transformations rather than CPU-based General purpose GPU-based transformations rather than CPU-based Pixel shaders Pixel shaders Shading of surfaces Shading of surfaces

23. Outcomes Create library to: Create library to: Efficiently identify large-scale structures Efficiently identify large-scale structures Generate surfaces around structures Generate surfaces around structures Refine structures, provide statistical feedback Refine structures, provide statistical feedback Effective GUI which uses the above to: Effective GUI which uses the above to: Interactively visualise large-scale structures Interactively visualise large-scale structures Allow the user to manipulate structure identification Allow the user to manipulate structure identification View information about structures View information about structures