Using structs with LWJGL in combination with SSBOs

[Lobo]

Hi,

I'm new here and I want to ask how I can bind buffers to structs in my vertex shader with LWJGL?

For example I have this struct in my VS

struct Particle
{
    vec4 pos;
    vec4 vel;
    float mass;
};

layout (std430, binding = 0) buffer ParticleBuffer
{
    Particle vertexPos[];
}; 
....

My question now is how can I will this struct with LWJGL? With c++ it is realtive simple I create just the same struct at C++ side, but in Java?
I cannot find anything about this topic in this forum, maybe I'm searching for the wrong keywords.

Actually I have 3 SSBOs instead of one struct:

layout (std430, binding = 0) buffer VertexPos
{
    vec4 vertexPos[];
}; 

layout (std430, binding = 1) buffer VertexVel
{
    vec4 vertexVel[];
};  

layout (std430, binding = 2) buffer VertexMass
{
    float vertexMass[];
}; 

That means I have to upload 3 buffers from Java to the VS with 3 binding calls:

positionBuffer.rewind();
		GL30.glBindBufferBase(GL43.GL_SHADER_STORAGE_BUFFER, 0, positionBufferHandle);
		GL15.glBufferSubData(GL43.GL_SHADER_STORAGE_BUFFER, 0, positionBuffer);
		
		velocityBuffer.rewind();
		GL30.glBindBufferBase(GL43.GL_SHADER_STORAGE_BUFFER, 1, velocityBufferHandle);
		GL15.glBufferSubData(GL43.GL_SHADER_STORAGE_BUFFER, 0, velocityBuffer);
		
		GL30.glBindBufferBase(GL43.GL_SHADER_STORAGE_BUFFER, 2, massBufferHandle);
		GL15.glBufferSubData(GL43.GL_SHADER_STORAGE_BUFFER, 0,massBuffer);

I think if it is possible to upload just one buffer (maybe all datas aligned) and just use one binding call it will run a little bit faster.

Would be great if anybody could help.

Thanks a lot!

Regards
Lobo

[spasi]

Yes, you can do that. You write all particle data in the same FloatBuffer and use a single call to upload it.

You must be careful with the field offsets and struct stride. It isn't easy in C/C++ either, because the alignment/padding rules for GLSL structs are not the same for C structs. The std layouts are fully defined in section 7.6.2.2 of the OpenGL spec. For packed and shared layouts, you must query the driver to get the correct offsets/stride.

For more info see section 4.4.5 of the GLSL spec and the OpenGL wiki.

[Lobo]

Thank you for the tip that everything can be uploaded in one float buffer. It works now :-)