Option like generateLegacySharpCornerPatches for stencil tables
12 views
Skip to first unread message
Bruce M
Sep 27, 2023, 9:56:41 AM9/27/23
to OpenSubdiv Forum
Hello,
When using the patch table, we like to turn off generateLegacySharpCornerPatches in order to get rounded corners.
There doesn't seem to be a similar option for stencil tables, to get vertex weights for rounded corners. Why is this? Are the weights given today all that are needed in either case?
Bruce
Barry F
Oct 6, 2023, 5:33:32 PM10/6/23
to OpenSubdiv Forum
The short answer is that the weights that reflect these options are being accurately computed and factored into the StencilTable when it is assembled. So no additional options for the StencilTable are required.
If you're interested in why...
A PatchTable uses two sets of points: the first computed from adaptive refinement and a second set for additional points of patches needed at irregular features (e.g. the points of Gregory patches). When a PatchTable is constructed, it builds an internal StencilTable to derive this second set of points from the first -- referred to as the "local point StencilTable". Any relevant options are taken into account when computing these weights. This local StencilTable can also be accessed via the PatchTable's public interface for use in a couple of ways.
When assembling an external StencilTable to compute all points required by a PatchTable, a StencilTable for the refined points is first explicitly constructed (with a few options specific to StencilTable) and the local point StencilTable for the PatchTable is then separately appended to it. Since this append operation is a simple recombination of weights, any options that affected their original computation do not need repeating.
If you're interested in why...
A PatchTable uses two sets of points: the first computed from adaptive refinement and a second set for additional points of patches needed at irregular features (e.g. the points of Gregory patches). When a PatchTable is constructed, it builds an internal StencilTable to derive this second set of points from the first -- referred to as the "local point StencilTable". Any relevant options are taken into account when computing these weights. This local StencilTable can also be accessed via the PatchTable's public interface for use in a couple of ways.
When assembling an external StencilTable to compute all points required by a PatchTable, a StencilTable for the refined points is first explicitly constructed (with a few options specific to StencilTable) and the local point StencilTable for the PatchTable is then separately appended to it. Since this append operation is a simple recombination of weights, any options that affected their original computation do not need repeating.
If you are constructing a LimitStencilTable to compute points on the limit surface, the construction is a little different and there is a situation to be aware of relative to such options...
The construction of a LimitStencilTable takes an existing PatchTable as an argument -- but that argument is optional (for backward compatibility with very early usage). If you do not specify a pre-existing PatchTable, then one will be constructed temporarily, and that construction will use defaults for all options. So always create the PatchTable separately with any desired options when constructing a LimitStencilTable.
Hopefully that answer all questions, thanks.
Reply all
Reply to author
Forward
0 new messages