[SOLVED] When exactly is the getMeshBufferCount() > 1

[cegprakash]

To get all the vertices in a mesh, I'm using

Code: Select all

S3DVertex* vertexArr = (S3DVertex*)mesh->getMeshBuffer(0)->getVertices();

I couldn't guess the scenario of

Code: Select all

mesh->getMeshBufferCount() > 1

When does this happen?

[luthyr]

If you were to import an mesh (such as OBJ) with multiple materials/textures, it will come in as separate meshbuffers because it would need to be drawn in 2 or more separate draw calls.

[CuteAlien]

Yeah - it's 1 meshbuffer per material.

[cegprakash]

what do you mean by a "material" and why is texture called as a material. Texture does not come along with OBJ.

[CuteAlien]

A material describes the way the polygons are rendered. In the simplest case it might be just 1 color or even a wireframe. A typical case is for example a single texture. But can be several textures, light-setttings or even a shader. Check SMaterial for some of the settings you can use to describe a material.

[cegprakash]

I have to get all the vertices in the mesh. I don't care about texture details or shader details.

[hybrid]

Still, your mesh could have more than one meshbuffer. For example if you have more than 65k polys and a loader supporting this. Or if your mesh type (and obj does) support groups. These are usually kept in Irrlicht as well, as they might be intended for further use. Say, if you store multiple parts of a mesh in separate groups. Or multiple objects.

[cegprakash]

I wrote this function to merge all the meshBuffers in a mesh. Hope it'll help someone someday

Code: Select all

updated code available below

[hendu]

Your function has no check for the 16-bit limit. If you have over USHRT_MAX verts your mesh will be corrupted. The index offset of subsequent buffers is also wrong.

[cegprakash]

I've fixed the indices part.. No idea how to fix the u16 part. My code will fail if total no. of vertices exceed 2^16-1

Code: Select all

updated code available below

[hendu]

It's still wrong if you have more than two meshbuffers.

[cegprakash]

why is it wrong?
I've added code to skip meshBuffers when the number of vertices/indices crosses u16 limit.

Code: Select all

updated code available below

[hendu]

Oh, my mistake, numVertices is not zeroed. Nevermind.

[cegprakash]

is it looking okay now?
Why isn't irrlicht allowing more than 2^16 indices/vertices in a mesh buffer? Will the rendering be too slow?

Even if it's slow, irrlicht is capable to draw lot of meshBuffers simultaneously. Why won't it allow u32?

[hendu]

Yes, it looks ok.

Irr does allow 32-bit indices, it's merely the default SMeshBuffer that does not. 16-bit indices should be used because they are faster and save VRAM, this is basic advice that all industry agrees on, and can be found in both AMD and Nvidia guides.

In other words, you are merely discouraged from using 32-bit indices, and indirectly meshes over 65k verts, because using them is against best practices.