Difference between revisions of "Level of Detail Wireframes"
Starseeker (talk | contribs) (LOD wireframes) |
(new layout, add references) |
||
Line 11: | Line 11: | ||
#Implement and test the API and wireframe schemes. | #Implement and test the API and wireframe schemes. | ||
− | Requirements | + | =References= |
+ | |||
+ | * src/librt/primitives/table.c | ||
+ | * src/librt/primitives/** | ||
+ | ** all primitives are in a subdirectory, each implement an rt_*_tess() and rt_*_plot() for mesh and wireframes respectively. | ||
+ | * include/raytrace.h | ||
+ | * include/rtgeom.h | ||
+ | |||
+ | =Requirements= | ||
* Familiarity with C | * Familiarity with C | ||
* (optional) Familiarity with MGED drawing and librt API | * (optional) Familiarity with MGED drawing and librt API | ||
− | |||
− |
Latest revision as of 09:13, 24 March 2011
Currently, the primary mode of geometry visualization in BRL-CAD's interactive tools is wireframes. This is relatively primitive compared to a modern shaded display, but has some advantages in that it is fast and general (CSG primitives with no surface representation can still generate wireframes).
While this approach is generally useful for CSG models, it runs into difficulties in several cases. Currently, primitives have no awareness of how they are being viewed, and thus generate only one wireframe for all situations. Hence, a close-up view of a sphere in wireframe mode can easily lead to an empty screen if the wireframe is zoomed out of view, despite the entire view being "filled" by the sphere. Conversely, a large BoT (bag of triangles) model may thousands upon thousands of edges, and the current drawing routines have no knowledge of which edges are hidden behind other geometry - this slows performance dramatically as MGED attempts to redraw huge line sets for each move of the geometry.
This project would implement a mechanism for each primitive to provide a "view aware" wireframe, instead of a generic one. A dense tiny BoT that takes up 1/200th of the view would know not to draw thousands of lines, and close-up of a sphere would know to provide some more detailed lines. Occlusion culling isn't necessary for this project (it's a difficult problem for implicit primitives) - the goal is to keep wireframes informative and interactive whatever the particular view/zoom is.
The particular tasks would be:
- Design a callback API for librt primitives to provide on request a view dependent wireframe
- For at least most of the common primitives, outline wireframe strategies for generating wireframes spanning very tiny to very large view situations.
- Implement and test the API and wireframe schemes.
References[edit]
- src/librt/primitives/table.c
- src/librt/primitives/**
- all primitives are in a subdirectory, each implement an rt_*_tess() and rt_*_plot() for mesh and wireframes respectively.
- include/raytrace.h
- include/rtgeom.h
Requirements[edit]
- Familiarity with C
- (optional) Familiarity with MGED drawing and librt API