Skip to content

Fix #589 - Remove decoding speed limitation #598

Merged
scottdraves merged 2 commits into
masterfrom
fix/589-current-cap-bottlenecks
Apr 28, 2026
Merged

Fix #589 - Remove decoding speed limitation #598
scottdraves merged 2 commits into
masterfrom
fix/589-current-cap-bottlenecks

Conversation

@pavlopylypchuk321

Copy link
Copy Markdown
Collaborator

No description provided.

@scottdraves scottdraves merged commit 6b28118 into master Apr 28, 2026
amillionbouncyballs added a commit to amillionbouncyballs/infinidream-client that referenced this pull request May 1, 2026
Conflict resolutions:

ContentDecoder.cpp — HW decode init restructured to accommodate all three
platforms under one unified block. Upstream added Windows D3D11VA support
(e-dream-ai#584/e-dream-ai#598) using #if USE_HW_ACCELERATION + #if WIN32. Our Linux VAAPI
block moved to an #elif LINUX_GNU since USE_HW_ACCELERATION=0 on Linux.
Mac VideoToolbox kept in the #else branch with our improved error handling
(av_buffer_ref, get_format, proper fallback logging).

Clip.cpp (constructor) — upstream adds m_spDecoder->SetD3D11Device() in
a #ifdef WIN32 block for D3D11VA frame sharing; our #ifdef LINUX_GNU frame
generation init retained below it. Both kept.

Clip.cpp (GrabVideoFrame) — upstream inlined the frame upload logic; we
kept our UploadFrameToTexture() call which handles D3D11VA HW frames,
RGB24 (RIFE path), and RGBA. Updated UploadFrameToTexture to also detect
isHWFrame for correct Windows D3D11VA BindFrame routing.

client.h (F1 help text) — added upstream's settings shortcut
(FULLSCREEN_MODIFIER_KEY-comma) using our BK() macro style.

client.h (GPU display) — kept our m_RuntimeDiagnostics.gpuFrameTimeMs
(already sourced from GetGPUFrameTimeMs()) but adopted upstream's
", %i%% app" label for the system GPU utilization string.

Co-Authored-By: Claude Sonnet 4.6 <[email protected]>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants