This project implements a renderer using OpenGL and C++.
The program loads the model as a gLTF. The model contains a few different types of textures, of which the renderer currently makes use of the diffuse and normal textures. In addition to the Sponza model there is also a skybox implemented using cubemaps.
The meshes and textures are fed through a series of different rendering passes:
-
Shadow Map: The scene contains a directional light providing sun and atmospheric lighting. In order for the sun to cast shadows a 1024x1024 pixel depth map is rendered from the perspective of the directional light. The light uses orthographic projection to transform and render the scene.
-
Geometry: Next, the scene gets rendered from the perspective of the camera, which can move freely through the scene. In this stage the meshes and textures are rendered without lighting into a geometry buffer that contains an attachment for albedo, one for world positions of the fragments, one for normals and a depth attachment.
-
Lighting: After creating the geometry buffer, it is passed through a shader that calculates the appropriate lighting based on the lights in the scene. Currently, this is limited to the one directional light. The shadow map generated earlier is also used in this pass. The shader used in this pass applies the Blinn-Phong lighting model. This model isn't super realistic and currently only takes diffuse and normal maps into account but still looks quite nice. I hope to implement PBR soon.
-
Post-processing: The steps taken so far use HDR and the linear color space, so this step is primarily responsible for tone-mapping the HDR colors into SDR colors and applying gamma correction. However, the post-processing step also applies a bloom effect. For this to work the lighting pass also generated a texture containing just the bright parts of the image which are then blurred by applying a gaussian blur multiple times using ping-pong frame buffers, which alternate between horizontal and vertical blurring, which increases performance at the cost of some memory overhead.
-
UI: This pass isn't strictly part of the render, but once the actual scene has been rendered Dear ImGUI is used to generate a basic user interface to change some of the settings used to render the scene, such as light colors and strengths, gamma correction, and more.
As evident by the render passes this renderer uses deferred shading instead of forward shading. This could be considered over-kill for such a simple scene with a single light source, but allows implementing many other screen-space effects in the future, such as SSAO. And of course this is a learning project, so using a simple forward renderer would be boring!
The camera can be move through the scene using the WASD keys to move forward, backward, left and right. While holding the right mouse button the mouse can be moved around to look around in the scene. The space bar and left CTRL key can be used to move up and down as well.
This project uses CMake as a build tool and vcpkg for dependency management.
Vcpkg is included as a git submodule, as such you should execute git submodule update --init --recursive
in the project directory to ensure everything is ready. CMake will automatically
invoke vcpkg to download and install the required dependencies. This can take a while
the first time.
You might need to install additional system dependencies to be able to compile the project.
On Ubuntu libxinerama-dev libxcursor-dev xorg-dev libglu1-mesa-dev pkg-config are required.
To build the project the following commands can be used:
$ mkdir build
$ cmake --preset Debug -B build
$ cd build
$ cmake --build --target sponza_scene --config Debug --preset Debug
These commands should work on Linux and Windows.