OpenGL: Bind FBO's depth texture to a compute shader - c++

Question

I've been trying to render to an FBO and render two FBO's to the screen, but have been unsuccessfull to do a depth test at the merge of the two FBO's. I've tried merging the textures with an compute shader, but I am unable to read the values of the depth textures(all values are value 1, but depth test is working when I render to the FBO). Does anybody knows what I am doing wrong, or knows an other method to merge two FBO's?
This is how I create the FBO:
struct FBO {
uint color;
uint depth;
uint fbo;
};
FBO createFBO(int width, int height) {
FBO fbo;
fbo.color = createFBOTexture(width, height, false);
fbo.depth = createFBOTexture(width, height, true);
fbo.fbo = generateFramebuffer(fbo.color, fbo.depth);
return fbo;
}
uint createFBOTexture(int width, int height, bool isDepthBuffer) {
uint texture;
glGenTextures(1, &texture);
glBindTexture(GL_TEXTURE_2D, texture);
if (isDepthBuffer)
glTexImage2D(GL_TEXTURE_2D, 0, GL_DEPTH_COMPONENT16, width, height, 0, GL_DEPTH_COMPONENT, GL_FLOAT, NULL);
else
glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA8, width, height, 0, GL_RGBA, GL_FLOAT, NULL);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
int i = glGetError();
if (i)
std::cout << "Error while creating the FBO: " << gluErrorString(i) << '\n';
return texture;
}
uint generateFrameBuffer(uint color, uint depth)
{
int mipmapLevel = 0;
//Generate FBO
uint fbo;
glGenFramebuffers(1, &fbo);
glBindFramebuffer(GL_FRAMEBUFFER, fbo);
//Attatch textures to the FBO
glFramebufferTexture2D(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, color, mipmapLevel);
glFramebufferTexture2D(GL_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_TEXTURE_2D, depth, mipmapLevel);
//Error check
int i = glCheckFramebufferStatus(GL_FRAMEBUFFER);
if (i != GL_FRAMEBUFFER_COMPLETE)
std::cout << "ERROR: frambuffer is not ok, status: " << i << '\n';
else {
int i = glGetError();
if (i)
std::cout << "Error while creating the FBO: " << gluErrorString(i) << '\n';
}
glBindFramebuffer(GL_FRAMEBUFFER, 0);
return fbo;
}
This is how I render to the FBO:
glEnable(GL_DEPTH_TEST);
glDepthMask(GL_TRUE);
glBindFramebuffer(GL_FRAMEBUFFER, fbo_pointcloud.fbo);
glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
drawPointCloud(); //Here I draw points with the depreciated fixed pipeline
glFlush();
I have tried to bind the textures with glBindTexture(all values read where 1):
void MergeFrames::merge(TextureLoading::FBO tex0, TextureLoading::FBO tex1, GLuint result, int width, int height)
{
glUseProgram(shader);
//Bind depth textures
glActiveTexture(GL_TEXTURE0+0);
glBindTexture(GL_TEXTURE_2D, tex0.depth);
glActiveTexture(GL_TEXTURE0+1);
glBindTexture(GL_TEXTURE_2D, tex1.depth);
//Bind color textures
glBindImageTexture(2, tex0.color, 0, GL_FALSE, 0, GL_READ_ONLY, GL_RGBA8);
glBindImageTexture(3, tex1.color, 0, GL_FALSE, 0, GL_READ_ONLY, GL_RGBA8);
glBindImageTexture(4, result, 0, GL_FALSE, 0, GL_READ_WRITE, GL_RGBA8);
//Dispatch the shader
glDispatchCompute(ceilf(width / 16.0f), ceilf(height / 16.0f), 1);
}
The compute shader:
#version 430
uniform sampler2D depthTex0;
uniform sampler2D depthTex1;
uniform layout(rgba8) readonly image2D colorTex0;
uniform layout(rgba8) readonly image2D colorTex1;
uniform layout(rgba8) writeonly image2D mergedColor;
layout (local_size_x = 16, local_size_y = 16) in;
void main() {
ivec2 ID = ivec2(gl_GlobalInvocationID.xy);
ivec2 size = imageSize(colorTex0);
vec2 texCoords = vec2(float(ID.x)/float(size.x),float(ID.y)/float(size.y));
float depths[2];
vec4 colors[2];
depths[0] = texture2D(depthTex0, texCoords).x;
depths[1] = texture2D(depthTex1, texCoords).x;
colors[0] = imageLoad(colorTex0, ID);
colors[1] = imageLoad(colorTex1, ID);
int i = int(depths[1] > depths[0]);
imageStore(mergedColor, ID, colors[i]);
}
I have tried to bind the textures with glBindTexture(all values read where 0):
void MergeFrames::merge(TextureLoading::FBO tex0, TextureLoading::FBO tex1, GLuint result, int width, int height)
{
glUseProgram(shader);
glBindImageTexture(0, tex0.color, 0, GL_FALSE, 0, GL_READ_ONLY, GL_RGBA8);
glBindImageTexture(1, tex1.color, 0, GL_FALSE, 0, GL_READ_WRITE, GL_RGBA8);
glBindImageTexture(2, tex0.depth, 0, GL_FALSE, 0, GL_READ_ONLY, GL_R16);
glBindImageTexture(3, tex1.depth, 0, GL_FALSE, 0, GL_READ_ONLY, GL_R16);
glBindImageTexture(4, result, 0, GL_FALSE, 0, GL_READ_WRITE, GL_RGBA8);
glDispatchCompute(ceilf(width / 16.0f), ceilf(height / 16.0f), 1);
}
The compute shader:
#version 430
uniform layout(rgba8) readonly image2D colorTex0;
uniform layout(rgba8) readonly image2D colorTex1;
uniform layout(r16) readonly image2D depthTex0;
uniform layout(r16) readonly image2D depthTex1;
uniform layout(rgba8) writeonly image2D mergedColor;
layout (local_size_x = 16, local_size_y = 16) in;
void main() {
ivec2 ID = ivec2(gl_GlobalInvocationID.xy);
float depths[2];
vec4 colors[2];
colors[0] = imageLoad(colorTex0, ID);
depths[0] = imageLoad(depthTex0, ID).x;
colors[1] = imageLoad(colorTex1, ID);
depths[1] = imageLoad(depthTex1, ID).x;
int i = int(depths[1] < depths[0]);
imageStore(mergedColor, ID, colors[i]);
}
And this is how I bind the indices (the indices are differt in the version where I use sampler2D instead of image2D):
MergeFrames::MergeFrames()
{
shader = OpenGL::compileComputeShader("MergeFrames.comp");
glUseProgram(shader);
glUniform1i(glGetUniformLocation(shader, "colorTex0"), 0);
glUniform1i(glGetUniformLocation(shader, "colorTex1"), 1);
glUniform1i(glGetUniformLocation(shader, "depthTex0"), 2);
glUniform1i(glGetUniformLocation(shader, "depthTex1"), 3);
glUniform1i(glGetUniformLocation(shader, "mergedColor"), 4);
}

Answer 1

I got the code working, by copying the data from depth texture to a texture with format GL_RED after rendering .