4cf498b09d
Use queued audio duration instead of queued frame count to constrain latency
...
This avoids latency explosion when we negotiate 10 ms instead of 5 ms audio packets.
2026-05-10 21:53:09 -05:00
9d6218d1ff
Don't capture the cursor by default in windowed mode
...
Closes #1872
2026-05-09 13:11:45 -05:00
5faa5fece8
Don't manually override v4l2m2m buffer counts
...
This causes POLLERR when decoding from meson-vdec on the Libre
Computer Alta (and likely other Amlogic boards). Our requried buffer
counts were already lower than the default 16/20 configuration that
FFmpeg uses, so overriding them was just a memory usage optimization.
2026-04-20 00:25:21 -05:00
51f86caac3
Fix assert on exit using KMSDRM on AMDGPU
...
AMDGPU disallows commits that would leave a CRTC active without
the primary plane enabled. Instead of disabling the primary plane,
simply restore the original state of the plane to avoid this.
2026-04-19 23:17:28 -05:00
5d813ed3df
Fix clamping of signed range properties
2026-04-19 22:17:37 -05:00
0e50a9d457
Only enable D3D11/DXGI debug layer if Graphics Tools are installed
...
Closes #1857
2026-04-13 20:06:24 -05:00
a766c9cafc
Disable texture sharing on Qualcomm GPUs
...
See #1857
2026-04-11 20:14:14 -05:00
af03f57e08
Fix Vulkan decoder probing on Nvidia using KMSDRM
2026-04-09 22:11:08 -05:00
1fc1122c65
Consolidate test window creation code
2026-04-09 21:43:35 -05:00
94d47e95fe
Fix non-determinism during Vulkan decoder probing on KMSDRM
2026-04-09 21:37:25 -05:00
4ec682dc72
Remove errant assert to fix software decoding on Xe
2026-04-06 21:12:00 -05:00
c8d18f8a8e
Bring back DRM atomic test commit using new DRM master lock
...
The DRM master lock avoids race conditions between Qt taking master
to draw the GUI and the DRM renderer performing a test commit.
2026-04-06 19:25:03 -05:00
9c11dca454
Move DRM master state cleanup into cleanupRenderContext()
2026-04-06 19:24:04 -05:00
5676a076fd
Fix overlay plane selection on drivers without zpos properties
2026-04-01 22:37:12 -05:00
91ce7a41e5
Cache the result of SDL_GetNumDisplayModes()
...
It's not free when using SDL2-compat.
2026-03-28 21:57:51 -05:00
53a7680a38
Fix incorrect autoselection of SW AV1 over HW H.264 for SDR content
2026-03-28 20:19:11 -05:00
88b4a17fe1
Use AV_CODEC_RECEIVE_FRAME_FLAG_SYNCHRONOUS for decoder probing on FFmpeg 8.1
2026-03-28 16:30:28 -05:00
8326223220
Revert "Use EOS to force decoder test frame output"
...
It causes hangs on defective v4l2m2m decoders, and it's not always
well-supported by out-of-tree decoders either.
This reverts commit 66a30c66f3
2026-03-28 16:24:20 -05:00
4aa5ad71c0
Use av_vk_get_optional_device_extensions() on FFmpeg 8.1
2026-03-28 16:17:59 -05:00
209511b2dc
Revert Pinyin IME workaround which is ineffective on SDL3
...
The underlying SDL bug is now fixed upstream for SDL 3.4.2.
2026-02-16 15:00:30 -06:00
6f4ced5ed6
Assume GPUs on non-x86 platforms are slow by default
...
This can be overriden with GL_IS_SLOW=0 at runtime
for systems with a fast GPU, though it doesn't make
much difference either way for non-KMSDRM cases.
For slower GPUs, this enables important scaling and
rendering optimizations that can substantially
improve performance (especially on fillrate-limited
GPUs common in SBCs).
2026-02-05 23:17:23 -06:00
8795fb54f4
Fix double-free in Vulkan renderer when an overlay is disabled
...
Also a drive-by fix to avoid pl_tex copying
2026-02-04 23:05:01 -06:00
5020fc6f48
Do not reset the renderer on SDL_RENDER_TARGETS_RESET
...
We only need to do this on a SDL_RENDER_DEVICE_RESET where we lose the whole device.
This also reverts a56efe5
2026-01-31 13:00:23 -06:00
a2122e17c6
Explicitly request linear sampling in StretchRect()
2026-01-31 12:48:03 -06:00
6bb7c8bf2a
Fix blurriness of DXVA2 overlay
...
Also make a drive-by fix for D3DTSS_COLOROP since we don't use vertex colors.
2026-01-31 12:44:50 -06:00
2dda65078f
Remove D3D9Ex focus loss hint workaround
...
SDL3 doesn't suffer this bug and keeping the default SDL3 auto behavior (minimize iff in exclusive mode) is a better UX.
2026-01-30 21:07:01 -06:00
a56efe5c36
Fix infinite renderer resets on SDL3 using direct3d renderer
2026-01-30 20:39:53 -06:00
b7a05d73b9
Update SDLRenderer capabilities to match SDL3 behavior
2026-01-30 20:38:53 -06:00
6d5b9e4706
Remove Windows dark mode hack since SDL3 supports it natively
2026-01-30 19:57:11 -06:00
792f762c15
Standardize on QString::toUtf8()
2026-01-26 21:49:08 -06:00
8c9c799f81
Fix Clazy non-pod-global-static warnings
2026-01-26 21:49:08 -06:00
d484ec3ac8
Fix Clazy range-loop-detach warnings
2026-01-26 21:49:08 -06:00
f5f06ae44e
Remove handling for Windows versions that are no longer supported
2026-01-26 20:04:15 -06:00
cd9aceb62c
Disable D3D11VA texture sharing on known broken GPUs
2026-01-26 19:37:43 -06:00
e4be57db5d
Remove superfluous Session checks in prepareToRender()
...
prepareToRender() is only called with a live Session.
2026-01-26 18:17:14 -06:00
64fea80ac9
Create the SystemProperties test window on the main thread
2026-01-25 17:10:04 -06:00
1e825c85e5
Optimize render to decode fence usage to reduce pipeline bubbles
2026-01-19 18:38:38 -06:00
a0a4c1ea83
Use decoder texture binding by default when using separate devices
...
This significantly improves performance on Ryzen 3300U and should generally perform equal or better everywhere.
This decoder->SRV path has been prone to driver bugs, so we may need to adjust this logic if driver issues crop up.
2026-01-19 16:46:50 -06:00
b5e7dec378
Fix D2R fence initial value
...
Fixes synchronization of the first decoded frame
2026-01-19 16:10:02 -06:00
adbd19e6fa
Remove the previous frame rendered fence
...
- Waiting on this fence prior to rendering introduces pipeline bubbles which can impact rendering performance on low-end GPUs
- Pacer simply waiting to release the previous frame until after the current frame is rendered already means any decoder stall should be very short
2026-01-19 16:01:50 -06:00
02a86167da
Disable separate decoder devices for non-Intel FL11.0 GPUs
2026-01-19 15:08:14 -06:00
a739c1cef8
Fix D3D11 fence sharing without monitored fence support
2026-01-19 14:53:23 -06:00
f6e08f8a43
Use separate render and decode D3D11Devices
...
See #1424
2026-01-19 14:25:08 -06:00
218ffc2d55
Set explicit scaling mode for all SDL textures
2026-01-17 02:34:54 -06:00
0f49dca4c0
Use GL_NEAREST when possible without degrading quality
2026-01-17 02:23:28 -06:00
30274f3ae8
Don't keep the dummy renderer alive for EGLRenderer's lifetime
...
At best, it doesn't do anything other than some queue some viewport
changes via SDL_RendererEventWatch(). At worst, it can race with
the render thread and cause trouble.
2026-01-17 01:06:41 -06:00
66a30c66f3
Use EOS to force decoder test frame output
2026-01-16 01:08:51 -06:00
198a208b44
Clamp property values when restoring them
2026-01-14 23:26:46 -06:00
d865c77975
Crop the incoming frames if they deviate slightly from the expected size
...
Fixes #1466
2026-01-11 20:58:48 -06:00
cd13910e6c
Add support for YUV 4:4:4 in DXVA2
...
We get it basically for free now that we let FFmpeg handle the heavy lifting.
2026-01-11 20:55:46 -06:00
Cameron Gutman