Editor playback optimization#1598

Merged

richiemcilroy merged 56 commits into

mainfrom

cursor/editor-playback-optimization-5f42

Feb 15, 2026

Commits on Feb 14, 2026

feat: add playback pipeline benchmark for measuring per-stage performance

cursoragent
and
richiemcilroy
committed
feat: add playback pipeline benchmark for measuring per-stage performance
richiemcilroy
committed
improve: optimize playback prefetch and decode parallelism

cursoragent
and
richiemcilroy
committed
improve: optimize playback prefetch and decode parallelism
richiemcilroy
committed
improve: batch GPU command submissions for YUV conversion and rendering

cursoragent
and
richiemcilroy
committed
improve: batch GPU command submissions for YUV conversion and rendering
richiemcilroy
committed
improve: pipeline GPU readback to overlap with next frame rendering

cursoragent
and
richiemcilroy
committed
improve: pipeline GPU readback to overlap with next frame rendering
richiemcilroy
committed
improve: tighten audio-video sync thresholds for better playback sync

cursoragent
and
richiemcilroy
committed
improve: tighten audio-video sync thresholds for better playback sync
richiemcilroy
committed
improve: avoid cloning CursorEvents per frame, use Arc sharing

cursoragent
and
richiemcilroy
committed
improve: avoid cloning CursorEvents per frame, use Arc sharing
richiemcilroy
committed
improve: reduce decoder/GPU timeouts and optimize GPU poll loop

cursoragent
and
richiemcilroy
committed
improve: reduce decoder/GPU timeouts and optimize GPU poll loop
richiemcilroy
committed
chore: format code with cargo fmt

cursoragent
and
richiemcilroy
committed
chore: format code with cargo fmt
richiemcilroy
committed
fix: flush pipelined GPU readback to capture last frame in exports

cursoragent
and
richiemcilroy
committed
fix: flush pipelined GPU readback to capture last frame in exports
richiemcilroy
committed
improve: add playback performance instrumentation logging

cursoragent
and
richiemcilroy
committed
improve: add playback performance instrumentation logging
richiemcilroy
committed
improve: send NV12 frames over WebSocket to reduce bandwidth ~62%

cursoragent
and
richiemcilroy
committed
improve: send NV12 frames over WebSocket to reduce bandwidth ~62%
richiemcilroy
committed
improve: add smart frame cache eviction for better memory usage

cursoragent
and
richiemcilroy
committed
improve: add smart frame cache eviction for better memory usage
richiemcilroy
committed
chore: format code with cargo fmt

cursoragent
and
richiemcilroy
committed
chore: format code with cargo fmt
richiemcilroy
committed
improve: batch camera layer YUV conversion into shared GPU encoder

cursoragent
and
richiemcilroy
committed
improve: batch camera layer YUV conversion into shared GPU encoder
richiemcilroy
committed
improve: optimize RGBA→NV12 conversion with better cache access pattern

cursoragent
and
richiemcilroy
committed
improve: optimize RGBA→NV12 conversion with better cache access pattern
richiemcilroy
committed
fix: address clippy warnings - collapsible if and abs_diff

cursoragent
and
richiemcilroy
committed
fix: address clippy warnings - collapsible if and abs_diff
richiemcilroy
committed
improve: use Arc<WSFrame> in watch channel to avoid per-frame deep clone

cursoragent
and
richiemcilroy
committed
improve: use Arc<WSFrame> in watch channel to avoid per-frame deep clone
richiemcilroy
committed
feat: add GPU RGBA→NV12 converter shader and pipeline infrastructure

cursoragent
and
richiemcilroy
committed
feat: add GPU RGBA→NV12 converter shader and pipeline infrastructure
richiemcilroy
committed
improve: GPU-side RGBA→NV12 conversion eliminates CPU conversion and reduces readback by 62%

cursoragent
and
richiemcilroy
committed
improve: GPU-side RGBA→NV12 conversion eliminates CPU conversion and reduces readback by 62%
richiemcilroy
committed
fix: address clippy warnings in NV12 converter (div_ceil, is_multiple_of)

cursoragent
and
richiemcilroy
committed
fix: address clippy warnings in NV12 converter (div_ceil, is_multiple_of)
richiemcilroy
committed
improve: add pipelined readback to GPU NV12 conversion path

cursoragent
and
richiemcilroy
committed
improve: add pipelined readback to GPU NV12 conversion path
richiemcilroy
committed
fix: add GPU error retry logic to render_nv12 matching RGBA render path

cursoragent
and
richiemcilroy
committed
fix: add GPU error retry logic to render_nv12 matching RGBA render path
richiemcilroy
committed
fix: handle RGBA fallback correctly in GPU NV12 conversion path

cursoragent
and
richiemcilroy
committed
fix: handle RGBA fallback correctly in GPU NV12 conversion path
richiemcilroy
committed
fix: revert to RGBA GPU rendering to fix full-resolution playback regression

cursoragent
and
richiemcilroy
committed
fix: revert to RGBA GPU rendering to fix full-resolution playback regression
richiemcilroy
committed

Commits on Feb 15, 2026