- Overview
- A Realtime/Interactive Fractal Explorer with an OpenGL
interface and back end computation using MPI
- Problem
- Generally, a fractal has properties which inherently
prohibit digitization. That is, assigning a pixel of a fractal
image to be some value contradicts the property that fractals are
invariant to scaling. We can think of a fractal as a series of 2D
images where the region of interest rendered changes over time. As a
result, we are not trapped into assigning a pixel, which corresponds to
some interval in the plane, with a single ,likely, incorrect
value. Presuming
that the region of interest changes provides for an arbitrary
resolution over time.
- If we presume to present an interactive image of a
fractal then we must compute many regions of interest. Which
points in the plane are meaningful to iterate over? That is,
which points will map to a pixel value in some rendered image?
Can the values from previous renderings be reused? How many
iterations are needed? How should this be parallelized?
- Interactivity
- Zooming in/out
- change of resolution
- need to recompute the entire image
- will require precomputation
- Navigation
- translation in x and/or y plane
- slide pixels in rendered image
- compute borders
- Front end
- OpenGL GUI for displaying rendered image
- Interacts with MPI
- requests data for current region of interest
- data transfered via LAN ( bandwidth bottleneck? )
- Assumes 1024x1024 size
- User selects the region of interest interactively
- Back end
- parallel fractal image generation
- recieves region of interest from OpenGL
- renders requested frame
- preemptively renders future frames
- additive increase, exponential decrease of iteration
number ( like TCP )
|