[Bf-committers] Triangulate modifier

[Brecht Van Lommel]

Tessfaces contain both quads and triangles. The render engine can
store quads natively (uses less memory) and has its own splitting
algorithm which is different than what OpenGL does. Perhaps it would
not be a bad thing to add an option to Mesh datablocks to disable that
smart render engine splitting and have it the same as the viewport,
and enable it by default even.

I don't know how important the smarter splitting it is in practice, in
Cycles I intentionally did not add it because it can pop in animation
and be inconsistent with the viewport, did not see complaints about it
so far.

Brecht.

On Wed, Nov 14, 2012 at 1:00 AM, CoDEmanX <codemanx at gmx.de> wrote:
> Alright, maybe what i had on mind isn't possible in your case.
>
> As Blender has a tessface cache, I thought it should be used throughout
> the application for everything that requires tessellated mesh geometry.
> Polygon and tessface data can be accessed at the same time, so this is
> basically non-destructive as well. What is missing IMO is a checkbox to
> toggle display of tessfaces (we already see tesselated faces, but they
> aren't outlined etc.)
>
> A modifier might offer an extra ability of editing the triangulated mesh
> non-destructively, not sure if possible, but if so, it would be great to
> be able to adjust triangulation in certain spaces (tessfaces don't make
> up "nice" triangles)
>
>
> Am 14.11.2012 00:25, schrieb metalliandy:
>> The point of the modifier is not how it is triangulated, it is that the
>> triangulation is non destructive and can be visibly toggled on and off
>> by the user at will and without the need for undo steps :)
>>
>> Also that quote you was from me, not Morten and I was 100% accurate.
>>
>> Morten replied and said in his reply to me
>>
>>> This is not exactly the issue. Let me clarify.
>>> The mikktspace part is done at quad level if there are quads and will
>>> always give the same
>>> tangent spaces at vertex level no matter what the order of vertices or
>>> faces is.
>>> This is its very strong advantage which other tangent space calculators
>>> can't do.
>>>
>>> After mikktspace has already run the problem occurs at a later stage
>>> because before rendering/baking triangulation has to have been performed.
>>> So the issue is eventhough the tangent spaces at the vertices aren't
>>> impacted by how quads are triangulated (since triangulation happens later)
>>> the interpolated tangent space is heavily impacted by which diagonal is
>>> picked. Similar to how Gouraud shading is heavily impacted by which
>>> diagonal is picked. So eventhough tangent space is consistent at the
>>> vertices you get problems if baker and render don't pick the same
>>> triangulation
>>> due to differences in the interpolated tangent space.
>>>
>>> I should mention I am using "render triangulation" loosely here since it
>>> can happen in a lot of places though for a game it would typically happen
>>> somewhere in the engine's tools pipeline. Anyway the advantage to this
>>> modifier is it enforces a consistent triangulation which will be received
>>> by both baker and renderer (incl. Blender's renderer). Further, this is
>>> achieved without causing permanent triangulation.
>>> It's a sensible practice which is already used by artists in other modeling
>>> products.
>>>
>>> Morten M.
>>
>> I hope that clears everything up :)
>>
>> -Andy
>>
>> On 13/11/2012 23:12, CoDEmanX wrote:
>>> Sounds to me like there is an inconsistency somewhere in Blender, which
>>> causes the difference in mesh geometry between what baker uses and the
>>> stored/exported data... Shouldn't everything in Blender use tessfaces,
>>> if triangles are required?
>>>
>>> "When baking with an all quad mesh, the triangulation that is done
>>> internally in Morten's TS might not match the manual triangulation that
>>> the user applies on export [...]" - Morten Mikkelsen
>>>
>>> There shouldn't be "manual triangulation", as export scripts can use
>>> tessfaces to get triangulated geometry. Maybe here's the inconsistency?
>>> The "Triangulate Faces" operator (former Quads to Tris) might create
>>> triangles different from the tessfaces. I wonder how Morten's TS does
>>> triangulation, hm, maybe I missed your point! (sorry if so)
>>>
>>>
>>> Am 10.11.2012 19:32, schrieb Antony Riakiotakis:
>>>> Exactly, the point is not display, but data availability for baking. A
>>>> {0,1,2} {2,1,3} scheme is also being considered for the modifier.
>>>> _______________________________________________
>>>> Bf-committers mailing list
>>>> Bf-committers at blender.org
>>>> http://lists.blender.org/mailman/listinfo/bf-committers
>>>>
>>> _______________________________________________
>>> Bf-committers mailing list
>>> Bf-committers at blender.org
>>> http://lists.blender.org/mailman/listinfo/bf-committers
>>>
>>
>> _______________________________________________
>> Bf-committers mailing list
>> Bf-committers at blender.org
>> http://lists.blender.org/mailman/listinfo/bf-committers
>>
> _______________________________________________
> Bf-committers mailing list
> Bf-committers at blender.org
> http://lists.blender.org/mailman/listinfo/bf-committers