Export pipeline speed#1601

Merged

richiemcilroy merged 27 commits intomainfrom

cursor/export-pipeline-speed-cbf4

Feb 15, 2026

Commits on Feb 15, 2026

feat: GPU NV12 export path + export-optimized encoder settings

cursoragent
and
richiemcilroy
committed
fix: handle RGBA fallback in NV12 export path

cursoragent
and
richiemcilroy
committed
fix: CPU RGBA→NV12 fallback when GPU conversion fails

cursoragent
and
richiemcilroy
committed
fix: flush pending NV12 frame at end of export render loop

cursoragent
and
richiemcilroy
committed
improve: robust RGBA fallback + encoder timing instrumentation

cursoragent
and
richiemcilroy
committed
improve: zero-alloc NV12 frame encoding via reusable frame buffer

cursoragent
and
richiemcilroy
committed
improve: use SIMD-optimized swscale for RGBA→NV12 CPU fallback

cursoragent
and
richiemcilroy
committed
fix: ensure screenshot always receives NV12 data

cursoragent
and
richiemcilroy
committed
test: add unit tests for NV12 export path helpers

cursoragent
and
richiemcilroy
committed
refactor: extract shared decode_segment_frames_with_retry helper

cursoragent
and
richiemcilroy
committed
clippy
richiemcilroy
committed
perf(video-decode): enable slice threading and increase decode thread count
richiemcilroy
committed
perf(export): remove unnecessary 200ms sleep before export
richiemcilroy
committed
style(desktop): collapse nested if into let-chain
richiemcilroy
committed
refactor(rendering): wrap frame data in Arc<Vec<u8>>
richiemcilroy
committed
perf(rendering): cache NV12 converter bind groups across frames
richiemcilroy
committed
perf(rendering): pipeline render with decode prefetch and latency hiding
richiemcilroy
committed
refactor(desktop): use render_immediate for non-export render calls
richiemcilroy
committed
perf(export): always use NV12 GPU path with dimension alignment
richiemcilroy
committed
perf(export): increase pipeline channel buffer sizes to 32
richiemcilroy
committed
refactor(export): use into_data for Arc frame extraction
richiemcilroy
committed
clippy
richiemcilroy
committed
clippy
richiemcilroy
committed
clippy
richiemcilroy
committed
clippy
richiemcilroy
committed
Fix AVAssetReader unwraps and add session notes
richiemcilroy
committed
Sync system audio start_time; minor rust refactors
richiemcilroy
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Export pipeline speed#1601

Export pipeline speed#1601
richiemcilroy merged 27 commits intomainfrom
cursor/export-pipeline-speed-cbf4

Commits on Feb 15, 2026

feat: GPU NV12 export path + export-optimized encoder settings

fix: handle RGBA fallback in NV12 export path

fix: CPU RGBA→NV12 fallback when GPU conversion fails

fix: flush pending NV12 frame at end of export render loop

improve: robust RGBA fallback + encoder timing instrumentation

improve: zero-alloc NV12 frame encoding via reusable frame buffer

improve: use SIMD-optimized swscale for RGBA→NV12 CPU fallback

fix: ensure screenshot always receives NV12 data

test: add unit tests for NV12 export path helpers

refactor: extract shared decode_segment_frames_with_retry helper

clippy

perf(video-decode): enable slice threading and increase decode thread count

perf(export): remove unnecessary 200ms sleep before export

style(desktop): collapse nested if into let-chain

refactor(rendering): wrap frame data in Arc<Vec<u8>>

perf(rendering): cache NV12 converter bind groups across frames

perf(rendering): pipeline render with decode prefetch and latency hiding

refactor(desktop): use render_immediate for non-export render calls

perf(export): always use NV12 GPU path with dimension alignment

perf(export): increase pipeline channel buffer sizes to 32

refactor(export): use into_data for Arc frame extraction

clippy

clippy

clippy

clippy

Fix AVAssetReader unwraps and add session notes

Sync system audio start_time; minor rust refactors