Repository navigation
Conversation
The renderer in both images writes a VUI bitstream_restriction with max_num_reorder_frames 0 (ejc3/fcvm#1361). Without it Chrome's H.264 decoder held 7 or 8 frames (234 to 268 ms) before outputting each tile frame. The host image is revised from the 2c965afe image with main's recipe; the tile image from the 31f42d03 image with main's tile units. Both renderers are built from the same fleet/renderer source.
|
Important
This repository does not receive automatic reviews because it has fewer than 10 stars.
 |
ejc3
marked this pull request as ready for review
October 27, 2026 06:53
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
This was referenced Oct 27, 2026
ejc3
added a commit
that referenced
this pull request
Oct 27, 2026
Owner
Author
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Host and tile AMIs from ejc3/fcvm main 2e2181cf, whose renderer writes a VUI
bitstream_restrictionwithmax_num_reorder_frames0 (ejc3/fcvm#1361).Without it Chrome's H.264 decoder (macOS) assumed the level's MaxDpbFrames of reordering and held 7 or 8 frames, 234 to 268 ms, before outputting each tile frame. The page already rewrites the SPS for older renderers (ejc3/fcvm#1360, deployed); these images fix the stream itself.
fcvm-fleet-host-x86_64-2e2181cf-20261027, revised from the 2c965afe image with main's recipe (fleet/ami/host/build.sh buildandbake), store parts uploaded under its own prefix (17 PUTs, all 200). The build's warm VM reached its first frame in 18.9 s.fcvm-fleet-tile-x86_64-2e2181cf-20261027, revised from the 31f42d03 image with main's tile units, renderer image built from the samefleet/renderersource.fleet-renderer-testin that image on a T4: 84 checks passed, including the encoder's SPS at four sizes and three qualities.measure-boot.sh tileon the new tile AMI (on demand): first frame over HTTPS 67.6 s after the request (the README's earlier figures: 67.7 s and 83.0 s).measure-boot.sh hoston the new host AMI: GPUs on vfio-pci 58.4 s and the store restored from its own S3 prefix 64.4 s after the request (16 parts in 8.5 s, 25G apparent); one GPU VM to first frame in 11.5 s, restored from the build's pre-start snapshot (snapshot: hit).