Cameron Gutman
9186feca80
Slightly simplify Metal shader
2024-08-22 01:18:39 -05:00
Cameron Gutman
df814fef4a
Implement support for HDR with software decoding on macOS
2024-08-22 00:05:26 -05:00
Cameron Gutman
0bb0d27d64
Implement YUV 4:4:4 decoding with D3D11VA on Intel GPUs
2024-07-30 21:12:11 -05:00
Cameron Gutman
1a1ce05959
Revert "Use a Texture2DArray in HLSL to match the SRVs"
...
This reverts commit d389f9a6e9bc0e06dec1b299a330be0640b75652.
2024-06-29 14:07:41 -05:00
Cameron Gutman
d389f9a6e9
Use a Texture2DArray in HLSL to match the SRVs
2024-06-27 19:46:27 -05:00
Cameron Gutman
94943d2865
Clamp chroma texcoords to avoid sampling alignment padding
...
This resolves #885 without the massive perf hit on lower end Intel GPUs.
2024-06-22 16:19:26 -05:00
Cameron Gutman
8959ddbeb6
Rewrite overlay support in Metal renderer to avoid additional subviews
...
Subviews cause our Metal layer to fall out of the direct composition path on Intel hardware
2024-02-11 00:15:07 -06:00
Cameron Gutman
db06239018
Rewrite the macOS renderer using CAMetalLayer
...
This allows v-sync to be disabled on macOS and lets us remove a whole bunch of old hacks.
Further optimizations of the new renderer are still needed.
2024-02-07 02:40:35 -06:00
Cameron Gutman
36db791516
Refactor D3D11 shaders to reduce duplicated code
2022-02-16 19:38:18 -06:00
Cameron Gutman
738f64c903
Use FP32 for position and texcoords
...
We need that extra precision for sampling large textures
2022-02-13 13:41:39 -06:00
Cameron Gutman
9a64c026ea
Use optimized pixel shaders for the normal SDR and HDR colorspaces
2022-02-09 20:46:02 -06:00
Cameron Gutman
b21131a466
Use FP16 in the D3D11 shaders
2022-02-08 21:43:25 -06:00
Cameron Gutman
1942e72474
Only bind the constant buffer once
2022-02-05 14:15:35 -06:00
Cameron Gutman
f256407789
Implement D3D11VA video rendering
2022-02-04 21:51:34 -06:00
Cameron Gutman
67612f607e
WIP: D3D11VA support
...
Overlays work, but drawing the actual video is unimplemented
2022-02-03 19:45:07 -06:00
Cameron Gutman
26d04b5f9c
Render DRM-PRIME frames as opaque images
...
We can't always assume they are NV12. Even if they _are_ NV12,
they may have DRM format modifiers that are incompatible with
simply assuming linear Y and UV buffers (such as tiling).
2021-03-21 23:23:12 -05:00
Cameron Gutman
402b6d1f0c
Begin refactoring for non-NV12 support in EGLRenderer
2021-03-21 19:20:52 -05:00
Cameron Gutman
e5bcc793e5
Implement overlay support in EGLRenderer
...
Fixes #405
2020-12-19 23:01:41 -06:00
Antoine Damhet
0a396f3112
Introduce a new FFMPEG frontend renderer: EGLRenderer
...
Right now this renderer works on X11 & Wayland with VAAPI as a backend.
Some rendering latency benchmarks on my `i7-10510U` (with
`intel-media-driver` 20.1.1 which cause a *huge* regression with the
SDL_Renderer):
| | X11 | Wayland |
| Before | 6.78ms | 22.50ms |
| EGLRenderer | 0.76ms | 00.77ms |
Signed-off-by: Antoine Damhet <antoine.damhet@lse.epita.fr>
2020-05-12 11:11:35 +02:00
