Using Uniform Buffer Objects with libgdx?

[soundasleep]

Hello! I'm trying to work out how to use Uniform Buffer Objects with libgdx/lwjgl, and I can't seem to get the data to bind correctly.

Following the UniformBufferObjectsTest.java in #7037 I've put together the following code that should create a new float[] buffer, populate it, and then render it:

int programId = compiled.getShader().getHandle();

// Get the block index for the uniform block
int uniformBlockIndex = Gdx.gl30.glGetUniformBlockIndex(programId, "VertexData");
if (uniformBlockIndex < 0) {
  throw new IllegalStateException(String.format("found no uniformBlockIndex: %s", uniformBlockIndex));
}

// creates a simple FloatBuffer
float[] data = new float[4];
data[0] = 1;
data[1] = 1;
data[2] = 1;
data[3] = 1;
FloatBuffer buf = BufferUtils.newFloatBuffer(data.length);
buf.put(data);
System.out.println("buf[0] = " + buf.get(0)); // yes, the FloatBuffer is being set to 1.0f

// Create a temporary buffer to create a new handle to store buffers (?)
IntBuffer tmpBuffer = BufferUtils.newIntBuffer(16);
Gdx.gl30.glGenBuffers(1, tmpBuffer);
int bufferHandle = tmpBuffer.get(0);

// create the actual UBO using this handle
Gdx.gl30.glBindBuffer(GL30.GL_UNIFORM_BUFFER, bufferHandle);
int size = buf.capacity() * 4 /* a float is 4 bytes */;
Gdx.gl30.glBufferData(GL30.GL_UNIFORM_BUFFER, size, buf, GL30.GL_STATIC_DRAW);
Gdx.gl30.glBindBuffer(GL30.GL_UNIFORM_BUFFER, 0);

// Use the index to bind to a binding point, then bind the buffer
int bindingPoint = 0;
Gdx.gl30.glBindBuffer(GL30.GL_UNIFORM_BUFFER, bufferHandle);
Gdx.gl30.glUniformBlockBinding(programId, uniformBlockIndex, bindingPoint);
Gdx.gl30.glBindBufferBase(GL30.GL_UNIFORM_BUFFER, bindingPoint, bufferHandle);
Gdx.gl30.glBindBuffer(GL30.GL_UNIFORM_BUFFER, 0);

// fill the buffer? possibly not necessary?
Gdx.gl30.glBindBuffer(GL30.GL_UNIFORM_BUFFER, bufferHandle);
Gdx.gl30.glBufferSubData(GL30.GL_UNIFORM_BUFFER, 0 /* offset */, size /* size */, buf);
Gdx.gl30.glBindBuffer(GL30.GL_UNIFORM_BUFFER, 0);

// update the buffer data store (should only be in the render call)
Gdx.gl30.glBindBuffer(GL30.GL_UNIFORM_BUFFER, bufferHandle);
Gdx.gl30.glBufferSubData(GL30.GL_UNIFORM_BUFFER, 0, buf.capacity() * 4, buf);
Gdx.gl30.glBindBuffer(GL30.GL_UNIFORM_BUFFER, 0);

// we're still able to bind other uniforms
compiled.getShader().setUniformf("anotherValue", 0.5f);

// and then draw as normal
batch.draw(texture, area.x, area.y, area.width, area.height);

The fragment shader code I'm using to debug is:

#version 150

uniform sampler2D u_texture;

in vec2 v_texCoords; // provided by vertex shader

layout (std140) uniform VertexData {
  vec2[4] v_polygon;
};
uniform float anotherValue;

out vec4 fragColor;

void main()
{
  // we ignore the texture but libgdx asserts there's always u_texture: https://stackoverflow.com/q/27967788/39531
  vec4 sampledButIgnored = texture(u_texture, v_texCoords);

  if (v_polygon[0].x == 0) {
    // v_polygon is not being populated with any data :(
    fragColor = vec4(anotherValue /* shows we're getting normal uniforms */, 1.0, 0.0, 1.0);
  } else {
    // we have data in v_polygon, we don't get here
    fragColor = vec4(v_polygon[0].x, v_polygon[0].y, v_polygon[1].x, 1.0);
  }

  fragColor += vec4(0, 0, 0, sampledButIgnored.a);
}

As far as I can tell I've got everything set up correctly but data just isn't getting loaded into the UBO/shader. Any ideas would be appreciated.

(I'd love to have first-class support for UBOs in a future version of libgdx!)

[Tom-Ski]

You should rewind the buffer after putting in the data

FloatBuffer buf = BufferUtils.newFloatBuffer(data.length);
buf.put(data);
buf.rewind();

You are also populating 4 floats, but in your shader you have 8 floats, (4 vec 2s)

[soundasleep]

buf.rewind() absolutely helped!! Thank you! 🥳

After that I had a flickering problem but that was probably because I was creating a brand new buffer on every single frame. If instead I created a single UBO then the flickering stops.

Also note that the float[] needs to have padding to align correctly, i.e.

float[] data = new float[8];
data[0] = 0.5f; // first vec2 is at offset=0
data[1] = 0.5f;
data[4] = 0.5f; // second vec2 is at offset=16
data[5] = 0.5f;

FloatBuffer buf = BufferUtils.newFloatBuffer(data.length);
buf.put(data);
buf.rewind(); // essential!
int size = buf.capacity() * 4; // a float is 4 bytes