VBO help...

Started by
7 comments, last by Ahl 11 years, 7 months ago
Having trouble getting my VBOs to work. No compiler errors. No run time errors. Just nothing showing up on the screen. VA's work though. I'm wondering if my implementation is the issue. I'll try to explain as best I can...

All the model information gets stored in the following:


struct MyVertex
{
float x, y, z; // Vertex
float nx, ny, nz; // Normal
float u, v; // Texcoords
float a, r, g, b; // Color
float padding[4];
};


This is how I load the VBO:


for (Itter = Mesh_Multimap.begin(); Itter != Mesh_Multimap.end(); ++Itter)
{
VBO_Size += (*Itter).second->NumVerticies*sizeof(MyVertex);
}
glGenBuffersARB( 1, &VBO_ID );
glBindBufferARB( GL_ARRAY_BUFFER_ARB, VBO_ID );
glBufferDataARB( GL_ARRAY_BUFFER_ARB, VBO_Size, NULL, GL_STATIC_DRAW_ARB );
for (VBO_Vector_IT = VBO_Vector.begin(); VBO_Vector_IT != VBO_Vector.end(); ++VBO_Vector_IT)
{
glBufferSubDataARB( GL_ARRAY_BUFFER_ARB, VBO_Offset, (*VBO_Vector_IT)->NumVerticies*sizeof(MyVertex), (*VBO_Vector_IT)->VertexData );
(*VBO_Vector_IT)->VBO_Offset = VBO_Offset;
VBO_Offset += (*VBO_Vector_IT)->NumVerticies*sizeof(MyVertex);
}


Rendering:


glVertexPointer( 3, GL_FLOAT, 0, &(*RenderListIT)->RenderMesh->VBO_Offset );
glNormalPointer( GL_FLOAT, 0, &(*RenderListIT)->RenderMesh->VBO_Offset );
glTexCoordPointer( 2, GL_FLOAT, 0, &(*RenderListIT)->RenderMesh->VBO_Offset );


That's the point that's giving me pause. When just rendering with VAs I need to do the following:


glVertexPointer( 3, GL_FLOAT, sizeof(MyVertex), &(*RenderListIT)->RenderMesh->VertexData->x );
glNormalPointer( GL_FLOAT, sizeof(MyVertex), &(*RenderListIT)->RenderMesh->VertexData->nx );
glTexCoordPointer( 2, GL_FLOAT, sizeof(MyVertex), &(*RenderListIT)->RenderMesh->VertexData->u );
glColorPointer( 3, GL_FLOAT, sizeof(MyVertex), &(*RenderListIT)->RenderMesh->VertexData->a );


I had to specify what part of the struct to pull the appropriate data from. Am I supposed to be doing this with the VBOs somehow? What am I missing?
Advertisement
Looks very wrong that you're passing the address of the offset variables to the gl*Pointer functions. If the offset variables actually contains the offsets into the VBOs, then you need to pass their values and not their addresses.
Ooh, good catch. Didn't even see that. I converted it over from where I needed it for the VAs and forgot about that. Thank you.

Stuff is showing up now but it's all scrambled...
Yeah, still getting garbled models, even when just loading one. I have this sneaky suspician that it has to do with not being able (or not know how) to specift what part of my struct each pointer needs to be using...
It didn't occur to me first, but you have the same offset for all attributes. The vertex pointer needs the offset to the start of the vertex data, of course, and likewise for the other attribute arrays.

struct MyVertex
{
float x, y, z; // Vertex
float nx, ny, nz; // Normal
float u, v; // Texcoords
float a, r, g, b; // Color
float padding[4];
};


Keep in mind that this is my data structure. An array of this is what is being loaded into the buffer, so the offset will be the same for all of the vertexs/normals/texcoords/etc per model.

I figured out another problem. I had the stride set to 0. I set it to sizeof(MyVertex) and now my model is coming up properly. The textures are still wonky though. It's doing now what it was doing before when we figured out I had to do the whole &(*RenderListIT)->RenderMesh->VertexData->x bit.


struct MyVertex
{
float x, y, z; // Vertex
float nx, ny, nz; // Normal
float u, v; // Texcoords
float a, r, g, b; // Color
float padding[4];
};


Keep in mind that this is my data structure. An array of this is what is being loaded into the buffer, so the offset will be the same for all of the vertexs/normals/texcoords/etc per model.


No. If vertex data starts at offset 0 (the most common case in that scenario), then normal data starts at offset 0 + sizeof(float) * 3 and so on. Otherwise your normals are always your positions of the same vertex, your texture coordinates are just your (x, y) coordinates of the same vertex and your colors are the (x, y, z, nx) coordinates of the same vertex.

Setting the stride to sizeof(MyVertex) was correct but only half the problem.
You can do something like this:MyVertex dummy = NULL;

glVertexPointer (3, GL_FLOAT, sizeof (MyVertex), &dummy->x);
glNormalPointer (GL_FLOAT, sizeof (MyVertex), &dummy->nx);
glTexCoordPointer (2, GL_FLOAT, sizeof (MyVertex), &dummy->u);
glColorPointer (3, GL_FLOAT, sizeof (MyVertex), &dummy->a);


It looks awful but it will set up the offsets correctly for you, doesn't require any casting of a number to (void *), allows you to modify your struct (e.g by adding another element - maybe a second set of texcoords - somewhere in the middle of it) without breaking your VBO setup, and makes it easier to co-exist code for regular vertex arrays with VBOs. Just be careful to not do anything else with "dummy" aside from this as it will crash your program.

Generally I'd use an array of xyz[3] instead of individual x, y, z but that's just a stylistic preference.

Direct3D has need of instancing, but we do not. We have plenty of glVertexAttrib calls.


[quote name='Ahl' timestamp='1348017039' post='4981497']

struct MyVertex
{
float x, y, z; // Vertex
float nx, ny, nz; // Normal
float u, v; // Texcoords
float a, r, g, b; // Color
float padding[4];
};


Keep in mind that this is my data structure. An array of this is what is being loaded into the buffer, so the offset will be the same for all of the vertexs/normals/texcoords/etc per model.


No. If vertex data starts at offset 0 (the most common case in that scenario), then normal data starts at offset 0 + sizeof(float) * 3 and so on. Otherwise your normals are always your positions of the same vertex, your texture coordinates are just your (x, y) coordinates of the same vertex and your colors are the (x, y, z, nx) coordinates of the same vertex.

Setting the stride to sizeof(MyVertex) was correct but only half the problem.
[/quote]

Boom! That did it. Thank you very much!

This topic is closed to new replies.

Advertisement