AMD has announced the availability of the source code for an updated implementation of FSR 2.2 (FidelityFX Super Resolution) supersampling technology, which uses spatial scaling and detail reconstruction algorithms to reduce image quality loss when upscaling and converting to higher resolution. The code is written in C++ and distributed under the MIT license. In addition to the basic API for the C++ language, the project provides support for DirectX 12 and Vulkan graphics APIs, as well as HLSL and GLSL shader languages. A set of examples and detailed documentation is provided.
FSR is used in games to scale output on high-resolution screens to achieve near-native resolution quality while preserving texture detail and sharp edges by reconstructing fine geometric and bitmap detail. With the help of settings, you can balance between quality and speed. The technology is compatible with various GPU models, including integrated chips.
The new version significantly improved the quality of generated images and worked to eliminate artifacts such as flicker and halo around fast moving objects. Changes have been made to the API, which may require changes to the code of applications that use the mask generation functionality. The "Debug API Checker" mechanism has been introduced to simplify the integration of FidelityFX Super Resolution with the application in debug builds (after enabling the mode, debug messages are transferred from the FSR runtime to the game, which makes it easier to diagnose problems that occur).
Source: opennet.ru