Faster export pipeline with parallel IO, cached offsets, and CPU RGBA→NV12#1786

Open

richiemcilroy wants to merge 17 commits into

mainfrom

Commits on May 8, 2026

perf(enc-ffmpeg): skip BufferedResampler when conversion is identity
richiemcilroy
committed
perf(rendering): add rgba_to_nv12_fast CPU conversion
richiemcilroy
committed
perf(rendering): parallelize segment meta load and add first_camera_duration
richiemcilroy
committed
test(rendering): verify lazy zoom precompute matches full precompute
richiemcilroy
committed
perf(rendering): cache clip offsets and accelerate NV12 readback conversion
richiemcilroy
committed
fix(rendering): order readback wait after submit in finish_encoder paths
richiemcilroy
committed
fix(audio): handle resampler flush loops and propagate decode EOF errors
richiemcilroy
committed
perf(audio): batch stereo mix using prepared linear gains
richiemcilroy
committed
perf(editor): render audio in-place with cached clip offsets
richiemcilroy
committed
perf(editor): load segment audio concurrently with decoder setup
richiemcilroy
committed
refactor(editor): use first_camera_duration for recording length bound
richiemcilroy
committed
chore(editor): re-export AudioSegment from crate root
richiemcilroy
committed
build(export): add cap-audio dev dependency for benchmarks
richiemcilroy
committed
perf(export): parallelize renderer setup with segment loading
richiemcilroy
committed
perf(export): speed NV12 pipeline and widen export frame channel
richiemcilroy
committed
chore(export): add CPU startup profiling example binary
richiemcilroy
committed
chore(export): add encoder benchmark example binary
richiemcilroy
committed