mirror of
https://git.triforceproject.com/triforce/source
synced 2026-11-19 16:20:01 -04:00
27 lines
1.5 KiB
Markdown
27 lines
1.5 KiB
Markdown
# FidelityFX SDK
|
|
|
|
AMD FidelityFX SDK, used for the FSR 3.1 upscaler's GPU sources. Unmodified upstream copy.
|
|
|
|
- Upstream: https://github.com/GPUOpen-LibrariesAndSDKs/FidelityFX-SDK
|
|
- Version: v1.1.4, commit `c6efa6bf7f2027b3ec94f28578bb5965eabb9e55`
|
|
- License: MIT, in `LICENSE.txt`. Keep it with any distribution.
|
|
|
|
What is here and why:
|
|
|
|
| | |
|
|
|---|---|
|
|
| `include/FidelityFX/gpu/` | The GPU-side headers the passes include: the common core, `fsr1/`, `spd/` (the single-pass downsampler the pyramids use), `fsr3upscaler/`, `frameinterpolation/` and `opticalflow/` |
|
|
| `shaders/fsr3upscaler/*.glsl` | The eight upscaler passes, as AMD ships them |
|
|
| `shaders/frameinterpolation/*.glsl` | The eleven frame-interpolation passes, from `sdk/src/backends/vk/shaders/frameinterpolation/` |
|
|
| `shaders/opticalflow/*.glsl` | The seven optical-flow passes frame interpolation runs on, from `sdk/src/backends/vk/shaders/opticalflow/` |
|
|
| `LICENSE.txt` | AMD's MIT notice |
|
|
|
|
No runtime library is needed, and none is included: the passes are compiled from these sources into
|
|
`gpu/vk/shaders/fsr3_*_comp_spv.h` by `scripts/fsr3_shaders.py`, which is also what
|
|
`tests/test_fsr3_shaders.py` runs to prove the checked-in headers still match. This is the SDK-source
|
|
route rather than AMD's FidelityFX API, which would load AMD's own Vulkan runtime and is Windows
|
|
only: these passes run on any GPU, on Windows and Linux alike.
|
|
|
|
Nothing in this folder is edited. A change belongs in a new folder with its own note, per
|
|
`CONTRIBUTING.md`.
|