MXL shared-memory flows on the FFmpeg 8.x series - #37

Open

jpietek wants to merge 19 commits into

developfrom

mxl-ffmpeg8

Commits on Sep 21, 2026

Add the MXL demuxer/muxer patch to the FFmpeg 8.x series

matiaspl
and
claude
committed
e6957ce
Add MXL round-trip demo

matiaspl
and
claude
committed
466e56b
Fold the MXL Dockerfile into the shared mixer image

matiaspl
and
claude
committed
66fffc8
Bump MXL SDK from v1.1.0-beta-1 to v1.1.0

matiaspl
and
claude
committed
8f4e7cd
Warn on input options not consumed by libavformat

matiaspl
and
claude
committed
22b07a4
Read MXL flows straight onto the GPU, zero-copy

matiaspl
and
claude
committed
e8f65ee
demos/mxl: pace the writer, survive a kill, verify on real hardware

matiaspl
and
claude
committed
f04fcac
docs: tighten the MXL demo and patch-series notes

matiaspl
and
claude
committed
3ebd242
MXL on develop: patch 0010, pyplumber.mixer import, README merge leftovers

authored and
matiaspl
committed
131a9f8
demos/mxl: measure throughput, and document the reset assert

matiaspl
and
claude
committed
75f7859
deps/ffmpeg: keep MXL pts monotonic across a reset_on_drop reset

matiaspl
and
claude
committed
9208b54
demos/mxl: re-measure throughput on an idle host, drop the abort caveat

matiaspl
and
claude
committed
c7af92d
rescale_video: apply the flags parameter, pass matching frames through

matiaspl
and
claude
committed
b52a8f3
demos/mxl: per-node CPU accounting, and reader tails that skip the encoder

matiaspl
and
claude
committed
4f428b9
demos/mxl: scale and convert on the GPU, and price the PCIe round trip

matiaspl
and
claude
committed
4896108
hwaccel: pack CUDA 10-bit 4:2:2 frames straight into v210

matiaspl
and
claude
committed
6ab5df9
demos/mxl: pack on the GPU, and find MXL's copy at the bottom

matiaspl
and
claude
committed
2facce6
deps/ffmpeg: regenerate the MXL patch as 0011, re-verify both bases

matiaspl
and
claude
committed
2927271
demos/mxl: keep the measurement scripts that produced the tables

matiaspl
and
claude
committed
42c215f

Navigation Menu

MXL shared-memory flows on the FFmpeg 8.x series - #37

MXL shared-memory flows on the FFmpeg 8.x series#37
jpietek wants to merge 19 commits into
developfrom
mxl-ffmpeg8

Commits on Sep 21, 2026

Add the MXL demuxer/muxer patch to the FFmpeg 8.x series

Add MXL round-trip demo

Fold the MXL Dockerfile into the shared mixer image

Bump MXL SDK from v1.1.0-beta-1 to v1.1.0

Warn on input options not consumed by libavformat

Read MXL flows straight onto the GPU, zero-copy

demos/mxl: pace the writer, survive a kill, verify on real hardware

docs: tighten the MXL demo and patch-series notes

MXL on develop: patch 0010, pyplumber.mixer import, README merge leftovers

demos/mxl: measure throughput, and document the reset assert

deps/ffmpeg: keep MXL pts monotonic across a reset_on_drop reset

demos/mxl: re-measure throughput on an idle host, drop the abort caveat

rescale_video: apply the flags parameter, pass matching frames through

demos/mxl: per-node CPU accounting, and reader tails that skip the encoder

demos/mxl: scale and convert on the GPU, and price the PCIe round trip

hwaccel: pack CUDA 10-bit 4:2:2 frames straight into v210

demos/mxl: pack on the GPU, and find MXL's copy at the bottom

deps/ffmpeg: regenerate the MXL patch as 0011, re-verify both bases

demos/mxl: keep the measurement scripts that produced the tables