We chose a Marching Cubes style algorithm as our base algorithm, opting for its ``embarrassingly parallel'' structure over the more elegant, but somewhat irregular, structure of a seed-growth method. The algorithm consists of moving linearly through the data, row by row, slice by slice, examining each cell for a possible intersection with the isosurface. When an intersected cell is located, the edge intersections are computed and triangles are generated to span those intersections. Viewed all together, these triangles represent a geometric piecewise linear set of surfaces where each surface either intersects the boundary of the volume or is closed.

- Predictable Data Traversal
- Cache-Ring Data Structure
- Algorithmic Performance
- Multiprocessor Implementation
- Edge Hashing

Scientific Computing and Imaging