Performance OpenGL^®: Platform Independent Techniques

Hi, and thank's for your interest in Performance OpenGL, which was presented as a tutorial course at SIGGRAPH 2001. Most of this web page is still under development, but here's some various bits that may be useful:

• Original paper that's the basis of the notes presented at SIGGRAPH 2001. It's slightly out of date. I'll be working to make it more up-to-date soon.
• Course Slides presented at SIGGRAPH 2001.
• Perl script that can be used to process an OpenGL header file to do automatic and more accurate error checking. The script has some options for controlling the amount of output and checking done. They're listed at the top of the script. If you're curious what the output looks like, here's the output when run against my workstation.
• If you're curious how I generated my statistics, here's source files for the programs I authored.
  
  

  • Test for determining fastest geometry mode

    This test goes through each of the different methods that can be used to pass geometry into OpenGL (immediate mode, diplay lists, separate and interleaved vertex arrays accessed throught the funcitions glDrawArrays(), glArrayElememt(), and glDrawElements().

    This test is written using GLUT in C++, and should be quite portable.

    I didn't test glDrawRangeElements() as it only does some additional error checking on array indices. I personally think that passing bad indices into a function call is an error, but that's only my opinion.

    • Parameters headers file
    • Rendering Test functions
    • Test harness code
  
  

  • Tests for determing appropriate pixel depths

    This test, written for the X Window System (GLUT doesn't permit specific selection of visual depths), queries all reasonable pixel formats, and determines how fast pixels in all supported OpenGL formats can be rendered to the framebuffer.

    I haven't had an opportunity to convert this into a Microsoft Window's version of the code. If someone's interested, I'd appreciate the port of the code.

    • Code for pixel tests
    • Test harness code
  
  

  • Tests for determing affects of validation on mode switches

    This test merely renders a number of cubes (controlled by a command line switch) in a number of different methods (e.g. six quads, quad strip, etc.). The purpose of the test is to determine the affects of the OpenGL validation process on your choice of rendering methods and order.

    • Parameters header file
    • Rendering Test functions
    • Test harness code