- AMD released the videos and slides from the sponsored session at GDC23
- covers the new FidelityFX SDK, Real-time sparse distance fields for games, temporal upscaling as well as direct storage
- the presentation introduces the implementation of Brixelizer, an AMD real-time generator for sparse distance fields
- discusses what sparse distance fields are, how they can be generated
- followed by a discussion of how Brixelizer can be used to generate the necessary data
- the article introduces Spherical Gaussians based on the Spherical Harmonics covered in the last part
- presents how these two representations differ, strengths and weaknesses
- additionally introduces Ambient Cube encoding for diffuse lighting
- the second article of the series continues uncovering how Deep Neural Networks are implemented
- looking at aim to reconstruct the Intel Open Image Denoise open-source library
- explains the concept of Tensors, Convolution, and how these concepts are combined to form Convolutional Neural Networks
- the terrain video tutorial series introduces a continuous level of detail for geometric details
- explains the implementation of Geomipmapping, discussing the vertex number requirements and how to deal with edges between LODs
- implementation with OpenGL is provided
- the article uses the AMD and Nvidia dev tools to take a look at how they approach the BVH building process
- compares different hardware and presents performance comparisons
- Hogwarts Legacy and Cyberpunk 2077 are used as examples
Thanks to Angel Ortiz for support of this series.
Would you like to see your name here too? Become a Patreon of this series.
