| 1 | |
|---|
| 2 | Illuminator Distributed Visualization Library |
|---|
| 3 | |
|---|
| 4 | Adam Powell |
|---|
| 5 | |
|---|
| 6 | May 11, 2001 |
|---|
| 7 | |
|---|
| 8 | This is a quick hack put together to graph 3-D distributed arrays' iso-contour |
|---|
| 9 | surfaces using Geomview. It is not at all "complete", rather a first cut, for |
|---|
| 10 | example, it computes triangle locations in parallel, but then rather |
|---|
| 11 | inefficiently gathers all of the triangles in the entire system to one node for |
|---|
| 12 | rendering and viewing. A future version (someday) will redistribute the |
|---|
| 13 | triangles from the processor where the data live to the processor(s) where it |
|---|
| 14 | will be rendered, so the image can be divided into pieces and rendered in |
|---|
| 15 | parallel. |
|---|
| 16 | |
|---|
| 17 | Long-term, there are other nice 3-D viewers out there such as Data Explorer |
|---|
| 18 | from IBM, MayaVi, etc., some kind of unified front end to them would be nice. |
|---|
| 19 | Furthermore, with parallel array storage, much more powerful parallel |
|---|
| 20 | visualization concepts are possible. Like if we can assign to each point a |
|---|
| 21 | certain luminoscity and transparency, it's easy to integrate those values along |
|---|
| 22 | a line through the locally-stored part of the array to produce a total |
|---|
| 23 | transparency and luminoscity for that line in the local array, then just |
|---|
| 24 | combine that with other line integrals through the other local sections, and we |
|---|
| 25 | have generated a total image in parallel. So then we have to decide where to |
|---|
| 26 | put the lines, and do that with one line per pixel, with the lines converging |
|---|
| 27 | for persective, and two sets of such lines for stereo imaging, etc. Or |
|---|
| 28 | multiple sets of lines for flat 3-D displays... |
|---|
| 29 | |
|---|
| 30 | So there's lots of fun stuff we can do with this. But for now, this humble |
|---|
| 31 | beginning. Share and enjoy. |
|---|
| 32 | |
|---|
| 33 | BUGS: |
|---|
| 34 | |
|---|
| 35 | If an isoquant surface happens to exactly intersect one or more of the |
|---|
| 36 | vertices, the triangles on the adjacent tetrahedra may be generated with |
|---|
| 37 | coordinates "nan nan nan". This is a problem, and I'll try to figure out a |
|---|
| 38 | solution at some point. In the meantime, the workaround is to choose a |
|---|
| 39 | slightly different isoquant surface value to avoid the intersection (e.g. if |
|---|
| 40 | 1000 intersects a vertex, then try 999.999). |
|---|
