Using the depth buffer for floating point rendertargets in OpenGL 2.1

Question

I would like to use the depth buffer to store the depth values of particles in eye space in a 2D texture by using OpenGL 2.1 / GLSL 1.2.

So far I found a way to use the colorbuffer

// create texture
glGenTextures(1, &g_hDepthTexture);
glBindTexture(GL_TEXTURE_2D, g_hDepthTexture);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA32F_ARB, g_windowWidth, g_windowHeight, 0, GL_RGBA, GL_FLOAT, 0);
// create framebuffer
glGenFramebuffersEXT(1, &g_hFBO);
glBindFramebufferEXT(GL_FRAMEBUFFER_EXT, g_hFBO);glFramebufferTexture2DEXT(GL_FRAMEBUFFER_EXT, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, g_hDepthTexture, 0);

However, I don't need the BGA components. Hence, I have tried to use the depth buffer, but the following code clamps each value in the texture to 0...1

// create texture
glGenTextures(1, &g_hDepthTexture);
glBindTexture(GL_TEXTURE_2D, g_hDepthTexture);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
glTexImage2D(GL_TEXTURE_2D, 0, GL_DEPTH_COMPONENT, g_windowWidth, g_windowHeight, 0, GL_DEPTH_COMPONENT, GL_FLOAT, 0);  // create framebuffer
glGenFramebuffersEXT(1, &g_hFBO);
glBindFramebufferEXT(GL_FRAMEBUFFER_EXT, g_hFBO);

glDrawBuffer(GL_NONE);
glReadBuffer(GL_NONE);

glFramebufferTexture2DEXT(GL_FRAMEBUFFER_EXT, GL_DEPTH_ATTACHMENT, GL_TEXTURE_2D, g_hDepthTexture, 0);

I would like to know how to use the depth buffer (probably how to choose the correct internal format / format) so that the texture values are not clamped.

score 1 · Answer 1 · answered Mar 05 '12 at 17:15

Normalized integer image formats are always clamped. That's why they're normalized integers. If you want an unclamped format, then you need floating point values.

I would suggest using an actual 1-channel floating-point image format, such as GL_R32F. Maybe GL_R16F, depending on how much precision I need. If you don't have GL 3.x hardware, you may be able to use GL_LUMINANCE32F_EXT, depending on what extensions are available.

BTW, if you're doing this for deferred rendering, don't bother. You can actually calculate the eye-space point directly from the regular depth buffer. Yes, really.

Using the depth buffer for floating point rendertargets in OpenGL 2.1

1 Answers1