Commit Graph
781 Commits
Author SHA1 Message Date
Cameron Gutman ac296864ac Fix some error handling in the libplacebo renderer 2026-10-24 22:41:06 -05:00
Cameron Gutman d606709d83 Add some missing error handling 2026-10-24 19:54:47 -05:00
Cameron Gutman ba700d67ec Fix shadowed member variable with base class 2026-10-24 19:54:47 -05:00
Cameron Gutman 465c41b5c0 Fix uninitialized InitFailureReason field 2026-10-24 19:54:47 -05:00
Cameron Gutman 29a280eb38 Fix return type of EGLRenderer::loadAndBuildShader() 2026-10-24 19:46:14 -05:00
Cameron Gutman fb9daf85ab Keep the VBO alive while the VAO references it 2026-10-24 19:39:58 -05:00
Cameron Gutman df7db4018e Fix some EGL error handling paths 2026-10-24 19:39:58 -05:00
Cameron Gutman bfc6751f0f Fix fd leak in VAAPI KMSDRM codepath 2026-10-24 19:39:58 -05:00
Cameron Gutman 2b690d14ad Initialize VAAPIRenderer class members 2026-10-24 19:39:47 -05:00
Cameron Gutman 4158fa7823 Don't add EGL modifiers for VA images if they aren't set 2026-10-24 18:49:46 -05:00
Cameron Gutman 372887b8a2 Don't copy unpopulated EGLAttribs 2026-10-24 18:49:46 -05:00
Cameron Gutman 1826d4c232 Fix incorrect mmap() failure check 2026-10-24 18:49:46 -05:00
Cameron Gutman 40f83bd294 Avoid wedging VDPAU renderer if mixer creation fails 2026-10-24 18:49:46 -05:00
Cameron Gutman 9af0e9f539 Improve robustness and error handling of DXVsyncSource 2026-10-20 00:29:25 -05:00
Cameron Gutman e4db5c2c18 Check for pacing queue emptiness before dequeuing
This shouldn't actually be reachable when the queue
is empty, but we can check anyway to be safe.

Ref: #2054
2026-10-19 00:25:00 -05:00
Cameron Gutman f743c23663 Fix PixelShaderMinPrecision bitfield comparison 2026-09-07 00:07:10 -05:00
Cameron Gutman e9d77967a6 Replace Nvidia FL 11.1 check with driver version check
The bug is driver-specific rather than GPU-specific.
2026-09-06 22:53:24 -05:00
Cameron Gutman 31ce68f136 Log the GPU driver version in the D3D11VA renderer 2026-09-06 22:22:07 -05:00
Cameron Gutman b86b5f71e3 Allow separate devices for Vega+ and block them for Fermi
Fermi is known broken, Kepler and Maxwell v1 may work but are currently untested
2026-09-06 16:12:52 -05:00
Cameron Gutman 09873287ca Only use separate D3D11 devices on Intel and Nvidia
Preserve the texture array binding path on AMD FL11.1+, since it seems to work fine on Polaris with shared devices.

Fixes #2003
2026-09-06 15:21:11 -05:00
Cameron Gutman 54f8fed51d Don't try fallback driver paths with the host libva 2026-09-05 14:30:28 -05:00
Cameron Gutman 7cf8b46c79 Pass the correct renderer to EGLImageFactory
We should be using the EGL renderer's colorspace and color range
rather than the backend renderer's for fallbacks. The former is
what actually gets passed to the host PC during stream init.
2026-08-06 00:29:53 -05:00
Cameron Gutman c1623ff44e Switch the default renderer color range to full 2026-08-05 23:56:18 -05:00
Cameron Gutman b2f0828df8 Use requested color range as a fallback in isFrameFullRange() 2026-08-05 23:54:19 -05:00
Cameron Gutman 3f26217df7 Select color range based on plane COLOR_RANGE property values 2026-08-05 23:49:09 -05:00
Cameron Gutman 4570fba87d Add full range handling to SdlRenderer 2026-08-05 23:44:11 -05:00
Cameron Gutman a903c5cef2 Explicitly request limited range for renderers that require it 2026-08-05 23:25:30 -05:00
Cameron Gutman 490229be3f Fix GEM handle leak in DRM renderer
Huge thanks to @oliverlj for identifying this longstanding bug.

This is likely responsible for many different issues with the DRM renderer randomly failing.
2026-07-26 15:22:15 -05:00
Cameron Gutman 8b18d58aa0 Add workaround for interoperability issue with out-of-tree FFmpeg zero-copy v4l2m2m patches and libplacebo
Fixes #1786
2026-07-08 23:40:08 -05:00
Cameron Gutman c9b03f80f8 Partially revert "Use GL_NEAREST when possible without degrading quality"
The changes to GL_NEAREST for overlay textures and disabling alpha blending in SDL renderer are left intact.

This reverts commit 0f49dca4c0.
2026-07-01 21:56:54 -05:00
Cameron Gutman 6e231778a2 Fix blocking during rendering rather than in waitToRender() on MoltenVK
MoltenVK does a lazy fetch of the drawable only when it's being modified, so waitToRender() is completely ineffective as is. Rendering a small empty overlay in waitToRender() forces the next drawable wait to happen in waitToRender(), before we pick the current frame to render. This removes one display refresh period of latency when using V-sync.
2026-06-25 20:01:29 -05:00
Cameron Gutman d997a0d0c8 Dynamically adjust libplacebo swapchain depth on macOS
Direct-to-display mode on some Macs (M4 Mac Mini) can retain our drawables, which cuts the frame rate in half if we're in double-buffered mode.

We don't want to use triple-buffered mode all the time because it increases latency when used if not required, so detect delayed presentation and enable it on the fly if needed.
2026-06-23 21:09:20 -05:00
Cameron Gutman 06bd8a730b Force sRGB Vulkan swapchain on macOS to fix washed out colors 2026-06-23 19:19:56 -05:00
Cameron Gutman 50364fd522 Don't set VK lock/unlock queue functions past libavutil 62
They are scheduled for removal at that ABI bump.
2026-06-21 14:01:47 -05:00
Cameron Gutman bea5355c66 Fix unused parameter warning on Windows and Linux 2026-06-21 14:01:47 -05:00
Cameron Gutman a9ad048252 Fix HDR rendering in libplacebo on macOS 2026-06-14 18:58:22 -05:00
Cameron Gutman 51cc5a8537 Remove libplacebo workaround for FFmpeg compiled without Vulkan support 2026-06-14 18:57:46 -05:00
Cameron Gutman 8c6b422040 Implement libplacebo on MoltenVK 2026-06-14 15:45:38 -05:00
Cameron Gutman ec75f0d8de Refactor libplacebo renderer to handle non-Vulkan hwcontexts 2026-06-14 12:20:54 -05:00
Cameron Gutman 97240cd5f2 Add workaround for libplacebo crash with Raspbian FFmpeg
Fixes #1409
2026-05-24 14:43:02 -05:00
Cameron Gutman 459dcd7b2c Fix incorrect error print 2026-05-16 16:10:47 -05:00
Cameron Gutman b33a252739 Fix incorrect apply() return value for legacy DRM clients 2026-05-16 16:09:57 -05:00
Cameron Gutman 87c3179f77 Fix incorrect failure check 2026-05-15 23:23:36 -05:00
Cameron Gutman 9cbba106ad Use the Metal renderer by default on all Macs
Whatever dGPU/eGPU rendering issues existed at the time I wrote that check seem to no longer happen on my RX 480 eGPU test setup running Ventura (our oldest supported OS).

It seems safe enough to enable now, especially since 7d6ce0b4c should catch broken systems and transparently fall back to AVSampleBufferDisplayLayer.

Fixes #1885
2026-05-14 00:45:57 -05:00
Cameron Gutman 7d6ce0b4cc Implement testRenderFrame() for the VTMetal renderer
This can be used to (hopefully) detect cases where importing the decoded frames onto the target GPU fails.

See #1885
2026-05-14 00:34:32 -05:00
Cameron Gutman 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
Cameron Gutman 5d813ed3df Fix clamping of signed range properties 2026-04-19 22:17:37 -05:00
Cameron Gutman 0e50a9d457 Only enable D3D11/DXGI debug layer if Graphics Tools are installed
Closes #1857
2026-04-13 20:06:24 -05:00
Cameron Gutman a766c9cafc Disable texture sharing on Qualcomm GPUs
See #1857
2026-04-11 20:14:14 -05:00
Cameron Gutman 94d47e95fe Fix non-determinism during Vulkan decoder probing on KMSDRM 2026-04-09 21:37:25 -05:00