Skip to content

perf(ltx23): cap PyAV libx264 encode threads at 12 #20

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
Johan-de-R merged 1 commit into from
Jul 9, 2026
Merged

perf(ltx23): cap PyAV libx264 encode threads at 12 #20

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 2 additions & 2 deletions examples/diffusers/Dockerfile.dreamverse
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -54,15 +54,15 @@ RUN . /opt/venv/bin/activate \
# the exact commit these patches were authored against, so they apply clean.
COPY patches/ltx23_gpu_worker_megacache.patch /tmp/megacache.patch
COPY patches/x264-threads-cap.patch /tmp/x264-threads-cap.patch
COPY patches/pyav-scenecut.patch /tmp/pyav-scenecut.patch
COPY patches/pyav-x264-tuning.patch /tmp/pyav-x264-tuning.patch
COPY patches/flash-attn-cute-fa4-tuple-fix.patch /tmp/flash-attn-cute-fa4-tuple-fix.patch
# NOTE: verify by grepping the patched source, NOT by importing fastvideo --
# importing pulls in triton, which fails to init its driver in a GPU-less build
# ("0 active drivers").
RUN SP=/opt/venv/lib/python3.12/site-packages \
&& patch -p1 -d "$SP" --forward --fuzz=0 < /tmp/megacache.patch \
&& patch -p1 -d "$SP" --forward --fuzz=0 < /tmp/x264-threads-cap.patch \
&& patch -p1 -d "$SP" --forward --fuzz=0 < /tmp/pyav-scenecut.patch \
&& patch -p1 -d "$SP" --forward --fuzz=0 < /tmp/pyav-x264-tuning.patch \
&& patch -p1 -d "$SP" --forward --fuzz=0 < /tmp/flash-attn-cute-fa4-tuple-fix.patch \
&& grep -q 'torch.compiler.load_cache_artifacts' "$SP/fastvideo/worker/gpu_worker.py" \
&& grep -q 'FASTVIDEO_X264_THREADS' "$SP/fastvideo/entrypoints/video_generator.py" \
Expand Down
10 changes: 9 additions & 1 deletion ...les/diffusers/patches/pyav-scenecut.patch → .../diffusers/patches/pyav-x264-tuning.patch
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,7 +2,7 @@ diff --git a/fastvideo/entrypoints/video_generator.py b/fastvideo/entrypoints/vi
index bfc1633..ef4f213 100644
--- a/fastvideo/entrypoints/video_generator.py
+++ b/fastvideo/entrypoints/video_generator.py
@@ -1010,6 +1010,12 @@ class VideoGenerator:
@@ -1010,6 +1010,20 @@ class VideoGenerator:
video_stream.options = {
"preset": "ultrafast",
"tune": "zerolatency",
Expand All @@ -12,6 +12,14 @@ index bfc1633..ef4f213 100644
+ # FASTVIDEO_X264_PRESET below) so it's opt-out per model without
+ # a rebuild. Measured ~50-100ms/clip (deepinfra, 2026-07-07).
+ "x264-params": f"scenecut={os.getenv('FASTVIDEO_X264_SCENECUT', '0')}",
+ # Cap x264 threads. B200 has no NVENC so this is CPU libx264;
+ # unbounded, x264 auto-detects all node cores (~288) -> thread
+ # thrash. Isolated encode-only sweep (deepinfra 2026-07-09, 121f
+ # 1080p) bottoms out ~10-12 threads (~0.42s) and gets WORSE past
+ # 16 (0.64s+); 12 = sweet spot AND well under a fair share. The
+ # ffmpeg-pipe path caps at 32 but is dead on B200 (no ffmpeg on
+ # PATH). Env-configurable, no rebuild to change.
+ "threads": os.getenv("FASTVIDEO_X264_THREADS_PYAV", "12"),
}

audio_stream = output.add_stream("aac", rate=sample_rate, layout=layout)
Loading