Related
I am attempting to use a CUDA kernel to modify an OpenGL texture, but am having a strange issue where my calls to surf2Dwrite() seem to blend with the previous contents of the texture, as you can see in the image below. The wooden texture in the back is what's in the texture before modifying it with my CUDA kernel. The expected output would include ONLY the color gradients, not the wood texture behind it. I don't understand why this blending is happening.
Possible Problems / Misunderstandings
I'm new to both CUDA and OpenGL. Here I'll try to explain the thought process that led me to this code:
I'm using a cudaArray to access the texture (rather than e.g. an array of floats) because I read that it's better for cache locality when reading/writing a texture.
I'm using surfaces because I read somewhere that it's the only way to modify a cudaArray
I wanted to use surface objects, which I understand to be the newer way of doing things. The old way is to use surface references.
Some possible problems with my code that I don't know how to check/test:
Am I being inconsistent with image formats? Maybe I didn't specify the correct number of bits/channel somewhere? Maybe I should use floats instead of unsigned chars?
Code Summary
You can find a full minimum working example in this GitHub Gist. It's quite long because of all the moving parts, but I'll try to summarize. I welcome suggestions on how to shorten the MWE. The overall structure is as follows:
create an OpenGL texture from a file stored locally
register the texture with CUDA using cudaGraphicsGLRegisterImage()
call cudaGraphicsSubResourceGetMappedArray() to get a cudaArray that represents the texture
create a cudaSurfaceObject_t that I can use to write to the cudaArray
pass the surface object to a kernel that writes to the texture with surf2Dwrite()
use the texture to draw a rectangle on-screen
OpenGL Texture Creation
I am new to OpenGL, so I'm using the "Textures" section of the LearnOpenGL tutorials as a starting point. Here's how I set up the texture (using the image library stb_image.h)
GLuint initTexturesGL(){
// load texture from file
int numChannels;
unsigned char *data = stbi_load("img/container.jpg", &g_imageWidth, &g_imageHeight, &numChannels, 4);
if(!data){
std::cerr << "Error: Failed to load texture image!" << std::endl;
exit(1);
}
// opengl texture
GLuint textureId;
glGenTextures(1, &textureId);
glBindTexture(GL_TEXTURE_2D, textureId);
// wrapping
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_MIRRORED_REPEAT);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_MIRRORED_REPEAT);
// filtering
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR_MIPMAP_LINEAR);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
// set texture image
glTexImage2D(
GL_TEXTURE_2D, // target
0, // mipmap level
GL_RGBA8, // internal format (#channels, #bits/channel, ...)
g_imageWidth, // width
g_imageHeight, // height
0, // border (must be zero)
GL_RGBA, // format of input image
GL_UNSIGNED_BYTE, // type
data // data
);
glGenerateMipmap(GL_TEXTURE_2D);
// unbind and free image
glBindTexture(GL_TEXTURE_2D, 0);
stbi_image_free(data);
return textureId;
}
CUDA Graphics Interop
After calling the function above, I register the texture with CUDA:
void initTexturesCuda(GLuint textureId){
// register texture
HANDLE(cudaGraphicsGLRegisterImage(
&g_textureResource, // resource
textureId, // image
GL_TEXTURE_2D, // target
cudaGraphicsRegisterFlagsSurfaceLoadStore // flags
));
// resource description for surface
memset(&g_resourceDesc, 0, sizeof(g_resourceDesc));
g_resourceDesc.resType = cudaResourceTypeArray;
}
Render Loop
Every frame, I run the following to modify the texture and render the image:
while(!glfwWindowShouldClose(window)){
// -- CUDA --
// map
HANDLE(cudaGraphicsMapResources(1, &g_textureResource));
HANDLE(cudaGraphicsSubResourceGetMappedArray(
&g_textureArray, // array through which to access subresource
g_textureResource, // mapped resource to access
0, // array index
0 // mipLevel
));
// create surface object (compute >= 3.0)
g_resourceDesc.res.array.array = g_textureArray;
HANDLE(cudaCreateSurfaceObject(&g_surfaceObj, &g_resourceDesc));
// run kernel
kernel<<<gridDim, blockDim>>>(g_surfaceObj, g_imageWidth, g_imageHeight);
// unmap
HANDLE(cudaGraphicsUnmapResources(1, &g_textureResource));
// --- OpenGL ---
// clear
glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
// use program
shader.use();
// triangle
glBindVertexArray(vao);
glBindTexture(GL_TEXTURE_2D, textureId);
glDrawElements(GL_TRIANGLES, 6, GL_UNSIGNED_INT, 0);
glBindVertexArray(0);
// glfw: swap buffers and poll i/o events
glfwSwapBuffers(window);
glfwPollEvents();
}
CUDA Kernel
The actual CUDA kernel is as follows:
__global__ void kernel(cudaSurfaceObject_t surface, int nx, int ny){
int x = blockIdx.x * blockDim.x + threadIdx.x;
int y = blockIdx.y * blockDim.y + threadIdx.y;
if(x < nx && y < ny){
uchar4 data = make_uchar4(x % 255,
y % 255,
0, 255);
surf2Dwrite(data, surface, x * sizeof(uchar4), y);
}
}
If I understand correctly, you initially register the texture, map it once, create a surface object for the array representing the mapped texture, and then unmap the texture. Every frame, you then map the resource again, ask for the array representing the mapped texture, and then completely ignore that one and use the surface object created for the array you got back when you first mapped the resource. From the documentation:
[…] The value set in array may change every time that resource is mapped.
You have to create a new surface object every time you map the resource because you might get a different array every time. And, in my experience, you will actually get a different one every so often. It may be a valid thing to do to only create a new surface object whenever the array actually changes. The documentation seems to allow for that, but I never tried, so I can't tell whether that works for sure…
Apart from that: You generate mipmaps for your texture. You only overwrite mip level 0. You then render the texture using mipmapping with trilinear interpolation. So my guess would be that you just happen to render the texture at a resolution that does not match the resolution of mip level 0 exactly and, thus, you will end up interpolating between level 0 (in which you wrote) and level 1 (which was generated from the original texture)…
It turns out the problem is that I had mistakenly generated mipmaps for the original wood texture, and my CUDA kernel was only modifying the level-0 mipmap. The blending I noticed was the result of OpenGL interpolating between my modified level-0 mipmap and a lower-resolution version of the wood texture.
Here's the correct output, obtained by disabling mipmap interpolation. Lesson learned!
Here is a description of the problem:
I want to render some VBO shapes (rectangles, circles, etc) to an off screen framebuffer object. This could be any arbitrary shape.
Then I want to draw the result on a simple sprite surface as a texture, but not on the entire screen itself.
I can't seem to get this to work correctly.
When I run the code, I see the shapes being drawn all over the screen, but not in the sprite in the middle. It remains blank. Even though it seems like I set up the FBO with 1 color texture, it still only renders to screen even if I select the FBO object into context.
What I want to achieve is these shapes being drawn to an off screen texture (using an FBO, obviously) and then render it on the surface of a sprite (or a cube, or we) drawn somewhere on the screen. Yet, whatever I draw, appears to be drawn in the screen itself.
The tex(tex_object_ID); function is just a short-hand wrapper for OpenGL's standard texture bind. It selects a texture into current rendering context.
No matter what I try I get this result: The sprite is blank, but all these shapes should appear there, not on the main screen. (Didn't I bind rendering to FBO? Why is it still rendering on screen?)
I think it is just a logistics of setting up FBO in the right order that I am missing. Can anyone tell what's wrong with my code?
Not sure why the background is red, as I clear it after I select the FBO. It is the sprite that should get the red background & shapes drawn on it.
/*-- Initialization -- */
GLuint texture = 0;
GLuint Framebuffer = 0;
GLuint GenerateFrameBuffer(int dimension)
{
glEnable(GL_TEXTURE_2D);
glGenTextures(1, &texture);
glBindTexture(GL_TEXTURE_2D, texture);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA8, dimension, dimension, 0, GL_RGBA, GL_UNSIGNED_BYTE, nullptr);
glGenFramebuffers(1, &Framebuffer);
glBindFramebuffer(GL_FRAMEBUFFER, Framebuffer);
glFramebufferTexture2D(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, texture, 0);
glDrawBuffer(GL_COLOR);
glReadBuffer(GL_COLOR);
glBindFramebuffer(GL_FRAMEBUFFER, 0);
if (glCheckFramebufferStatus(GL_FRAMEBUFFER) != GL_FRAMEBUFFER_COMPLETE)
console_log("GL_FRAMEBUFFER != GL_FRAMEBUFFER_COMPLETE\n");
return texture;
}
// Store framebuffer texture (should I store texture here or Framebuffer object?)
GLuint FramebufferHandle = GenerateFrameBuffer( 256 );
Standard OpenGL initialization code follows, memory is allocated, VBO's are created and bound, etc. This works correctly and there aren't errors in initialization. I can render VBOs, polygons, textured polygons, lines, etc, on standard double buffer with success.
Next, in my render loop I do the following:
// Possible problem?
// Should FramebufferHandle be passed here?
// I tried "texture" and "Framebuffer " as well, to no effect:
glBindFramebuffer(GL_FRAMEBUFFER, FramebufferHandle);
// Correct projection, just calculates the view based on current zoom
Projection = setOrthoFrustum(-config.zoomed_width/2, config.zoomed_width/2, -config.zoomed_height/2, config.zoomed_height/2, 0, 100);
View.identity();
Model.identity();
// Mini shader, 100% *guaranteed* to work, there are no errors in it (works normally on the screen)
shaderProgramMini.use();
//Clear frame buffer with blue color
glClearColor(0.0f, 0.0f, 1.0f, 1.0f);
glClear(GL_COLOR_BUFFER_BIT);// | GL_DEPTH_BUFFER_BIT);
// Set yellow to draw different shapes on the framebuffer
color = {1.0f,1.0f,0.0f};
// Draw several shapes (already correctly stored in VBO objects)
Memory.select(VBO_RECTANGLES); // updates uniforms
glDrawArrays(GL_QUADS, 0, Memory.renderable[VBO_RECTANGLES].indexIndex);
Memory.select(VBO_CIRCLES); // updates uniforms
glDrawArrays(GL_LINES, 0, Memory.renderable[VBO_CIRCLES].indexIndex);
Memory.select(VBO_2D_LIGHT); // updates uniforms
glDrawArrays(GL_LINES, 0, Memory.renderable[VBO_2D_LIGHT].indexIndex);
// Done writing to framebuffer
glBindFramebuffer(GL_FRAMEBUFFER, 0);
// Correct projection, just calculates the view based on current zoom
Projection = setOrthoFrustum(-config.zoomed_width/2, config.zoomed_width/2, -config.zoomed_height/2, config.zoomed_height/2, 0, 100);
View.identity();
Model.identity();
Model.scale(10.0);
// Select texture shader to draw what was drawn on offscreen Framebuffer / texture
// Standard texture shader, 100% *guaranteed* to work, there are no errors in it (works normally on the screen)
shaderProgramTexture.use();
// This is a wrapper for bind texture to ID, just shorthand function name
tex(texture); // FramebufferHandle; // ? // maybe the mistake in binding to the wrong target object?
color = {0.5f,0.2f,0.0f};
Memory.select(VBO_SPRITE); Select a square VBO for rendering sprites (works if any other texture is assigned to it)
// finally draw the sprite with Framebuffer's texture:
glDrawArrays(GL_TRIANGLES, 0, Memory.renderable[VBO_SPRITE].indexIndex);
I may have gotten the order of something completely wrong. Or FramebufferHandle/Framebuffer/texture object is not passed to something correctly. But I spent all day, and hope someone more experienced than me can see the mistake.
GL_COLOR is not an accepted value for glDrawBuffer
See OpenGL 4.6 API Compatibility Profile Specification, 17.4.1 Selecting Buffers for Writing, Table 17.4 and Table 17.5, page 628
NONE, FRONT_LEFT, FRONT_RIGHT, BACK_LEFT, BACK_RIGHT, FRONT, BACK, LEFT, RIGHT, FRONT_AND_BACK, AUXi.
Arguments to DrawBuffer when the context is bound to a default framebuffer, and the buffers they indicate. The same arguments are valid for ReadBuffer, but only a single buffer is selected as discussed in section.
COLOR_ATTACHMENTi
Arguments to DrawBuffer(s) and ReadBuffer when the context is bound to a framebuffer object, and the buffers they indicate. i in COLOR_ATTACHMENTi may range from zero to the value of MAX_COLOR_ATTACHMENTS minus one.
Thsi means that glDrawBuffer(GL_COLOR); and glReadBuffer(GL_COLOR); will generate a GL_INVALID_ENUM error.
Try to use COLOR_ATTACHMENT0 instead.
Furthermore, glCheckFramebufferStatus(GL_FRAMEBUFFER), checkes the completeness of the framebuffer object which is bound to the target.
This means that
glCheckFramebufferStatus(GL_FRAMEBUFFER) != GL_FRAMEBUFFER_COMPLETE
has to be done before
glBindFramebuffer(GL_FRAMEBUFFER, 0);
Or you have to use:
glNamedFramebufferReadBuffer(Framebuffer, GL_FRAMEBUFFER);
Implementing some effect, I end up with 1 frame buffer associated to 1 texture, which holds my final scene. This texture is then applied on a fullscreen quad.
The result is what I expect as far as the effect goes, but I noticed that edges on the scene thus rendered, weren't smooth - presumably, because multi-sampling did not apply during render-to-framebuffer passes, as it does when I render directly to the screen buffer.
So my question is
How can I apply/use multi-sampling on this final texture, so that its content shows smooth edges?
EDIT: I have removed the original version of my code here, which was using
a classic FrameBuffer + Texture not multi-sampled. Below is the lastest,
following suggestions in the comments.
For now also, I'll focusing on getting the glBlitFramebuffer approach to work!
So my code now goes like so:
// Unlike before, finalTexture is multi-sampled, thus created like this:
glGenFramebuffers(1, &finalFrame);
glGenTextures(1, &finalTexture);
glBindFramebuffer(GL_FRAMEBUFFER, finalFrame);
glBindTexture(GL_TEXTURE_2D_MULTISAMPLE, finalTexture);
glTexImage2DMultisample(GL_TEXTURE_2D_MULTISAMPLE, 4, GL_RGBA, w, h, GL_TRUE);
glFramebufferTexture2D(GL_FRAMEBUFFER,
GL_COLOR_ATTACHMENT0,
GL_TEXTURE_2D_MULTISAMPLE,
finalTexture,
0);
// Alternative using a render buffer instead of a texture.
//glGenRenderbuffers(1, &finalColor);
//glBindRenderbuffer(GL_RENDERBUFFER, finalColor);
//glRenderbufferStorageMultisample(GL_RENDERBUFFER, 8, GL_RGBA, w, h);
//glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_RENDERBUFFER, finalColor);
glBindFramebuffer(GL_FRAMEBUFFER, 0);
// Then I introduced a new frame buffer to resolve the multi-sampling:
// This one's not multi-sampled.
glGenFramebuffers(1, &resolveFrame);
glGenTextures(1, &resolveTexture);
glBindFramebuffer(GL_FRAMEBUFFER, resolveFrame);
glBindTexture(GL_TEXTURE_2D, resolveTexture);
glTexImage2D (GL_TEXTURE_2D, 0, GL_RGBA, w, h, 0, GL_RGBA, GL_UNSIGNED_BYTE, NULL);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
glFramebufferTexture2D(GL_FRAMEBUFFER,
GL_COLOR_ATTACHMENT0,
GL_TEXTURE_2D,
resolveTexture,
0);
glBindFramebuffer(GL_FRAMEBUFFER, 0);
// Now a lot of code to produce a glowing effect, things like:
// 1. Generate 1 frame buffer with 2 color attachments (textures) - no multisampling
// 2. Render the 3D scene to it:
// - texture 0 receives the entire scene
// - texture 1 receives glowing objects only
// 3. Generate 2 frame buffers with 1 color attachment (texture) each - no multisampling
// - we can call them Texture 2 and texture 3
// 4. Ping-pong Render a fullscreen textured quad on them
// - On the first iteration we use texture 1
// - Then On each following iteration we use one another's texture (3,2,3...)
// - Each time we apply a gaussian blur
// 5. Finally sum texture 0 and texture 3 (holding the last blur result)
// - For this we create a multi-sampled frame buffer:
// - Created as per code here above: finalFrame & **finalTexture**
// - To produce the sum, we draw a full screen texured quad with 2 sampler2D:
// - The fragment shader then computes texture0+texture3 on each pixel
// - finalTexture now holds the scene as I expect it to be
// Then I resolve the multi-sampled texture into a normal one:
glBindFramebuffer(GL_READ_FRAMEBUFFER, finalFrame);
glBindFramebuffer(GL_DRAW_FRAMEBUFFER, resolveFrame);
glBlitFramebuffer(0, 0, w, h, 0, 0, w, h, GL_COLOR_BUFFER_BIT, GL_NEAREST);
glBindFramebuffer(GL_READ_FRAMEBUFFER, 0);
glBindFramebuffer(GL_DRAW_FRAMEBUFFER, 0);
// And the last stage: render onto the screen:
glActiveTexture(GL_TEXTURE0);
glBindTexture(GL_TEXTURE_2D, resolveTexture);
drawFullScreenQuad( ... );
The resulting output is correct, meaning that I can see the scene with the desired glowing effect... But no apparent multi-sampling! :(
Note: I am starting to wonder, if I am using multi-sampling at the right stage - I will be experimenting on this - but any chance I should use it when rendering the initial 3D scene for the first time, on the initial FBOs? (the ones I refer to in the comments and I didn't want to post here to avoid confusion :s)
I added more detailed comments on what's going on before this last stage with final & resolve frame buffers.
You have: "step 5. Finally sum texture 0 and texture 3 (holding the last blur result) - For this we create a multi-sampled frame buffer". But this way multisampling will only apply to fullscreen quad.
"if I am using multi-sampling at the right stage" so the answer to your question is no, you need to use multisampling on another stage when you render a scene.
I have very similar setup with framebuffers (that one which is used to render the scene is multisampled) two output textures (for color info and for highlights which will later be blurred to achieve glow) and ping-pong framebuffers. I also use glBlitFramebuffer solution (also I use 2 blit calls for each color attachment, each one will go in own texture), have not found any way of making it render directly into framebuffer with attached texture.
If you want some code, this is solution that worked for me (it is in C# though):
// ----------------------------
// Initialization
int BlitFrameBufferHandle = GL.GenFramebuffer();
GL.BindFramebuffer(FramebufferTarget.Framebuffer, BlitFrameBufferHandle);
// need to setup this for 2 color attachments:
GL.DrawBuffers(2, new [] {DrawBuffersEnum.ColorAttachment0, DrawBuffersEnum.ColorAttachment1});
// create texture 0
int ColorTextureHandle0 = GL.GenTexture();
GL.BindTexture(TextureTarget.Texture2D, ColorTextureHandle0);
GL.TexParameter(TextureTarget.Texture2D, TextureParameterName.TextureMinFilter, (int) TextureMinFilter.Linear); // can use nearest for min and mag filter also
GL.TexParameter(TextureTarget.Texture2D, TextureParameterName.TextureMagFilter, (int) TextureMagFilter.Linear);
GL.TexParameter(TextureTarget.Texture2D, TextureParameterName.TextureWrapS, (int) TextureWrapMode.ClampToEdge);
GL.TexParameter(TextureTarget.Texture2D, TextureParameterName.TextureWrapT, (int) TextureWrapMode.ClampToEdge);
// for HRD use PixelInternalFormat.Rgba16f and PixelType.Float. Otherwise PixelInternalFormat.Rgba8 and PixelType.UnsignedByte
GL.TexImage2D(TextureTarget.Texture2D, 0, PixelInternalFormat.Rgba16f, Width, Height, 0, PixelFormat.Rgba, PixelType.Float, IntPtr.Zero);
GL.FramebufferTexture2D(FramebufferTarget.Framebuffer, FramebufferAttachment.ColorAttachment0, TextureTarget.Texture2D, ColorTextureHandle0, 0);
// create texture 1
int ColorTextureHandle1 = GL.GenTexture();
GL.BindTexture(TextureTarget.Texture2D, ColorTextureHandle1);
GL.TexParameter(TextureTarget.Texture2D, TextureParameterName.TextureMinFilter, (int) TextureMinFilter.Linear);
GL.TexParameter(TextureTarget.Texture2D, TextureParameterName.TextureMagFilter, (int) TextureMagFilter.Linear);
GL.TexParameter(TextureTarget.Texture2D, TextureParameterName.TextureWrapS, (int) TextureWrapMode.ClampToEdge);
GL.TexParameter(TextureTarget.Texture2D, TextureParameterName.TextureWrapT, (int) TextureWrapMode.ClampToEdge);
GL.TexImage2D(TextureTarget.Texture2D, 0, PixelInternalFormat.Rgba16f, Width, Height, 0, PixelFormat.Rgba, PixelType.Float, IntPtr.Zero);
GL.FramebufferTexture2D(FramebufferTarget.Framebuffer, FramebufferAttachment.ColorAttachment1, TextureTarget.Texture2D, ColorTextureHandle1, 0);
// check FBO error
var error = GL.CheckFramebufferStatus(FramebufferTarget.Framebuffer);
if (error != FramebufferErrorCode.FramebufferComplete) {
throw new Exception($"OpenGL error: Framwbuffer status {error.ToString()}");
}
int FrameBufferHandle = GL.GenFramebuffer();
GL.BindFramebuffer(FramebufferTarget.Framebuffer, FrameBufferHandle);
// need to setup this for 2 color attachments:
GL.DrawBuffers(2, new [] {DrawBuffersEnum.ColorAttachment0, DrawBuffersEnum.ColorAttachment1});
// render buffer 0
int RenderBufferHandle0 = GL.GenRenderbuffer();
GL.BindRenderbuffer(RenderbufferTarget.Renderbuffer, RenderBufferHandle0);
GL.RenderbufferStorageMultisample(RenderbufferTarget.Renderbuffer, 8, RenderbufferStorage.Rgba16f, Width, Height);
GL.FramebufferRenderbuffer(FramebufferTarget.Framebuffer, FramebufferAttachment.ColorAttachment0, RenderbufferTarget.Renderbuffer, RenderBufferHandle0);
// render buffer 1
int RenderBufferHandle1 = GL.GenRenderbuffer();
GL.BindRenderbuffer(RenderbufferTarget.Renderbuffer, RenderBufferHandle1);
GL.RenderbufferStorageMultisample(RenderbufferTarget.Renderbuffer, 8, RenderbufferStorage.Rgba16f, Width, Height);
GL.FramebufferRenderbuffer(FramebufferTarget.Framebuffer, FramebufferAttachment.ColorAttachment1, RenderbufferTarget.Renderbuffer, RenderBufferHandle1);
// depth render buffer
int DepthBufferHandle = GL.GenRenderbuffer();
GL.BindRenderbuffer(RenderbufferTarget.Renderbuffer, DepthBufferHandle);
GL.RenderbufferStorageMultisample(RenderbufferTarget.Renderbuffer, 8, RenderbufferStorage.DepthComponent24, Width, Height);
GL.FramebufferRenderbuffer(FramebufferTarget.Framebuffer, FramebufferAttachment.DepthAttachment, RenderbufferTarget.Renderbuffer, DepthBufferHandle);
// check FBO error
var error = GL.CheckFramebufferStatus(FramebufferTarget.Framebuffer);
if (error != FramebufferErrorCode.FramebufferComplete) {
throw new Exception($"OpenGL error: Framwbuffer status {error.ToString()}");
}
// unbind FBO
GL.BindFramebuffer(FramebufferTarget.Framebuffer, 0);
// ----------------------------
// Later for each frame
GL.BindFramebuffer(FramebufferTarget.Framebuffer, FrameBufferHandle);
// render scene ...
// blit data from FrameBufferHandle to BlitFrameBufferHandle
GL.BindFramebuffer(FramebufferTarget.ReadFramebuffer, FrameBufferHandle);
GL.BindFramebuffer(FramebufferTarget.DrawFramebuffer, BlitFrameBufferHandle);
// blit color attachment0
GL.ReadBuffer(ReadBufferMode.ColorAttachment0);
GL.DrawBuffer(DrawBufferMode.ColorAttachment0);
GL.BlitFramebuffer(
0, 0, Width, Height,
0, 0, Width, Height,
ClearBufferMask.ColorBufferBit, BlitFramebufferFilter.Nearest
);
// blit color attachment1
GL.ReadBuffer(ReadBufferMode.ColorAttachment1);
GL.DrawBuffer(DrawBufferMode.ColorAttachment1);
GL.BlitFramebuffer(
0, 0, Width, Height,
0, 0, Width, Height,
ClearBufferMask.ColorBufferBit, BlitFramebufferFilter.Nearest
);
// after that use textures ColorTextureHandle0 and ColorTextureHandle1 to render post effects using ping-pong framebuffers ...
Just implemented a bloom effect myself, faced the same aliased edges on the resulting image and faced the exactly same issues. Hence sharing my experience here.
Aliasing happens when you render the lines with OpenGL - e.g. edges of a triangle or a polygon, since OpenGL draws "diagonal" (or simply put non-straight) lines on the screen using quite simple (yet fast) algorithms.
That being said, if you want to anti-alias something - that would be a 3D shape, not a texture - it is just a plain image after all.
Off-topic: in order to fix aliasing on an image you would apply the similar technique, but you would need to figure out where the "edges" are on the image and then follow the same algorithm per "edge" pixel. "Edge" (in quotes) since they are just ordinary pixels from the image perspective and being an edge is just extra context we humans attach to those pixels.
With that out of our way, the thing with two image attachments is actually a nice optimization - you do not need to render your entire scene twice to different framebuffers. But you will pay the price of copying the data from each multi-sampled framebuffer attachment to a separate non-multisampled texture for post-processing.
A bit off-topic: performance-wise, I think this is exactly the same (or within a very small threshold) - rendering an entire scene twice, to two separate framebuffers with two separate multi-sampled attachments (as inputs for the post-processing) and then copying each of them separately to two separate non-multisampled textures.
So the last step before you can apply your (any) post-processing to the multi-sampled scene is to convert each multi-sampled render result to non-multisampled texture - so that your shaders work with plain sampler2D.
It would be something similar to this:
glBindFramebuffer(GL_READ_FRAMEBUFFER, bloomFBOWith2MultisampledAttachments);
glBindFramebuffer(GL_DRAW_FRAMEBUFFER, temporaryFBOWith1NonMultisampledAttachment);
// THIS IS IMPORTANT
glReadBuffer(GL_COLOR_ATTACHMENT0);
glDrawBuffer(GL_COLOR_ATTACHMENT0);
glBlitFramebuffer(0, 0, windowWidth, windowHeight, 0, 0, windowWidth, windowHeight, GL_COLOR_BUFFER_BIT, GL_NEAREST);
// bloomFBOWith2MultisampledAttachments is still bound
glBindFramebuffer(GL_DRAW_FRAMEBUFFER, blurFramebuffer1);
// THIS IS IMPORTANT
glReadBuffer(GL_COLOR_ATTACHMENT1);
glDrawBuffer(GL_COLOR_ATTACHMENT0);
glBlitFramebuffer(0, 0, windowWidth, windowHeight, 0, 0, windowWidth, windowHeight, GL_COLOR_BUFFER_BIT, GL_NEAREST);
glBindFramebuffer(GL_READ_FRAMEBUFFER, 0);
glBindFramebuffer(GL_DRAW_FRAMEBUFFER, 0);
Given you are rendering your scene to two attachments in one framebuffer, you will then need to copy from each of those multi-sampled attachments to non-multi-sampled textures and use them for additive rendering and blurring, correspondingly.
If you don't mind messy code and the use of globjects for OpenGL APIs abstraction, here's my entire bloom solution with anti-aliasing.
And few screenshots:
The first screenshot does not use a framebuffer to render to, so the lines are really smooth.
The second screenshot is the first implementation of a bloom effect (available as a separate CMake project).
Aliasing is more visible on longer distances, so the third screenshots shows a bit more of a scene - the edges look really stairs-like.
The last two screenshots show the bloom effect with anti-aliasing applied.
Note how lantern only has somewhat low-resolution texture, hence aliased lines, whilst the paper has its edges smoothed out by anti-aliasing.
I'm trying to read out the pixels of a texture that only has a depth component, however glReadPixels gives me an array where every value = 1.
Texture / Framebuffer creation:
GLuint frameBuffer;
glGenFramebuffers(1,&frameBuffer);
glBindFramebuffer(GL_FRAMEBUFFER,frameBuffer);
GLuint texture;
glGenTextures(1,&texture);
glBindTexture(GL_TEXTURE_2D,texture);
glTexImage2D(GL_TEXTURE_2D,0,GL_DEPTH_COMPONENT,width,height,0,GL_DEPTH_COMPONENT,GL_FLOAT,0);
glFramebufferTexture(GL_FRAMEBUFFER,GL_DEPTH_ATTACHMENT,m_depthTexture,0);
Reading from the texture:
glBindFramebuffer(GL_FRAMEBUFFER,frameBuffer);
float *depths = new float[width *height];
glReadPixels(0,0,width,height,GL_DEPTH_COMPONENT,GL_FLOAT,&depths[0]);
// glGetError reports no errors, but every value inside 'depths' is 1.
delete[] depths;
I didn't include the actual rendering to the texture, since I know that that works as it should.
This is what the depth texture looks like when I draw it on my main screen framebuffer:
It's definitely not empty, so why is it telling me the depth is 1 for all pixels?
Since you're using a desktop OpenGL version I suggest you use glGetTexImage instead of glReadPixels.
I have an FBO object with a color and depth attachment which I render to and then read from using glReadPixels() and I'm trying to add to it multisampling support.
Instead of glRenderbufferStorage() I'm calling glRenderbufferStorageMultisampleEXT() for both the color attachment and the depth attachment. The frame buffer object seem to have been created successfully and is reported as complete.
After rendering I'm trying to read from it with glReadPixels(). When the number of samples is 0 i.e. multisampling disables it works perfectly and I get the image I want. when I set the number of samples to something else, say 4, the frame buffer is still constructed OK but glReadPixels() fails with an INVALID_OPERATION
Anyone have an idea what could be wrong here?
EDIT: The code of glReadPixels:
glReadPixels(0, 0, width, height, GL_RGBA, GL_UNSIGNED_BYTE, ptr);
where ptr points to an array of width*height uints.
I don't think you can read from a multisampled FBO with glReadPixels(). You need to blit from the multisampled FBO to a normal FBO, bind the normal FBO, and then read the pixels from the normal FBO.
Something like this:
// Bind the multisampled FBO for reading
glBindFramebufferEXT(GL_READ_FRAMEBUFFER_EXT, my_multisample_fbo);
// Bind the normal FBO for drawing
glBindFramebufferEXT(GL_DRAW_FRAMEBUFFER_EXT, my_fbo);
// Blit the multisampled FBO to the normal FBO
glBlitFramebufferEXT(0, 0, width, height, 0, 0, width, height, GL_COLOR_BUFFER_BIT, GL_NEAREST);
//Bind the normal FBO for reading
glBindFramebufferEXT(GL_FRAMEBUFFER_EXT, my_fbo);
// Read the pixels!
glReadPixels(0, 0, width, height, GL_RGBA, GL_UNSIGNED_BYTE, pixels);
You can't read the multisample buffer directly with glReadPixels since it would raise an GL_INVALID_OPERATION error. You need to blit to another surface so that the GPU can do a downsample. You could blit to the backbuffer, but there is the problem of the "pixel owner ship test". It is best to make another FBO. Let's assume you made another FBO and now you want blit. This requires GL_EXT_framebuffer_blit. Typically, when your driver supports GL_EXT_framebuffer_multisample, it also supports GL_EXT_framebuffer_blit, for example the nVidia Geforce 8 series.
//Bind the MS FBO
glBindFramebufferEXT(GL_READ_FRAMEBUFFER_EXT, multisample_fboID);
//Bind the standard FBO
glBindFramebufferEXT(GL_DRAW_FRAMEBUFFER_EXT, fboID);
//Let's say I want to copy the entire surface
//Let's say I only want to copy the color buffer only
//Let's say I don't need the GPU to do filtering since both surfaces have the same dimension
glBlitFramebufferEXT(0, 0, width, height, 0, 0, width, height, GL_COLOR_BUFFER_BIT, GL_NEAREST);
//--------------------
//Bind the standard FBO for reading
glBindFramebufferEXT(GL_FRAMEBUFFER_EXT, fboID);
glReadPixels(0, 0, width, height, GL_BGRA, GL_UNSIGNED_BYTE, pixels);
Source: GL EXT framebuffer multisample