109 |
that window. |
that window. |
110 |
|
|
111 |
This method should be safe to use as the general case; it is |
This method should be safe to use as the general case; it is |
|
optimizations that should be the special case, used if necessary. |
|
112 |
|
optimizations that should be the special case, used if necessary. |
113 |
|
|
114 |
|
- What about the interaction of stenciling in OpenGL: We do want |
115 |
|
to use stencils for e.g. irregular edges, but |
116 |
|
1) there are only 8 bits in most implementations, and each clip |
117 |
|
takes one bit |
118 |
|
2) only one stencil operation can be specified. |
119 |
|
|
120 |
|
RESOLVED: We can agree that we shall support four levels of clipping |
121 |
|
in OpenGL and Vobs are free to use the four lower-end bits. |
122 |
|
Also, the bits are always set to 1 for pass, so |
123 |
|
the stencil functions NEVER, ALWAYS, EQUAL, GEQUAL, GREATER |
124 |
|
shall always work. |
125 |
|
|
126 |
|
- Can we implement clipping along with Vob Batching correctly? |
127 |
|
|
128 |
|
RESOLVED: Surprisingly, yes, with little problems. When clip() |
129 |
|
is called, all active batches need to be prepared to clip, |
130 |
|
**if** a vob is placed in them after that. |
131 |
|
|
132 |
|
- If we can't clip to a nonlinear CS in a particular case, for example |
133 |
|
because there is no stencil buffer, what should we do when asked |
134 |
|
to do that? Exception or silent failure with artifacts? |
135 |
|
|
136 |
|
- Should the clipping exception be specified more accurately to allow |
137 |
|
graceful recovery from the preceding condition? |
138 |
|
|
139 |
|
|
140 |
|
|
141 |
|
|