On macOS the Horizon client renders offscreen (--offscreen) with audio off,
from the logged-in session; Linux keeps vscreen and the null ALSA device.
The vanilla side was already a headless dedicated server. --bin-dir is
resolved, since the client runs from the run folder. parity.py no longer
skips save_roundtrip off Linux; it stays local-only (no rendering over SSH).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A new chunk relights its eight neighbours, and every finished light job
dirties all sections of the 3x3 columns around it. Meshing waited only for
the column itself to be lit, so sections near the loading frontier were
meshed with half-settled light and again after each neighbour's relight:
about 2.25 meshes per section while a world loads. Wait until no column of the
3x3 is being lit or queued for light. The final meshes are the same; they are
built once.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Same rule as perf_bench: with status.streaming, the last sections are drawn
once no stage has held work for half a second, so the remeshing that
follows a load (fluids settling, random ticks) isn't counted as loading.
Without it, meshing still has to go quiet.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Builds that report status.streaming are loaded once chunk counts are
stable and no stage (generation, decoration, light, meshing) has held work
for half a second. Meshing going quiet also waited through the remeshing
that follows a load (fluids settling, random ticks: 1500 falling to 60
meshes/s for ~150 s at RD16), which is simulation, not loading. Older
builds keep the meshing-rate rule.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Sections wait for their neighbours' light before they are meshed, so the
chunk counts settled while edge sections were still undrawn.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Each game runs --repeats times (3) per render distance and writes one v1
report per distance (benchmarks/perf/<label>-rd<N>.json): figures as means
with standard deviation and 95% CI, per-flight frame entries with the fixed
histogram, the hardware without the host name. frames() and summary() match
crates/horizon-bench value for value (tested on the same inputs).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Each session rewrites benchmarks/launcher-<time>-session.json in the data
folder (report format v1): start to first frame, release list fetch,
install/update and prepare durations, Play to the game's first frame (the
game now logs its first frame), RSS, UI frame times, binary size.
horizon-launcher benchmark starts the launcher N times (cold, then warm),
switches pages, measures idle CPU and memory and the release list fetch,
and writes one report with means and 95% CIs. Nothing is uploaded.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
One versioned JSON report for every benchmark Horizon writes (the game's
--benchmark, the launcher's measurements, tools/perf_bench.py): the local log
a player can attach and, with consent, what the telemetry upload sends.
Summaries over repeats (mean, stdev, 95% CI) and fixed-edge frame-time
histograms, never raw frame streams; no host, user or path in uploads.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Horizon saved a chunk only once a player changed it and never saved proto
chunks, so leaving an area dropped decorated chunks while the final chunks
next to them stayed loaded (or were regenerated later). Coming back, those
chunks were generated again and their decoration was skipped beside a final
neighbour: whole chunks of a bastion lost their pieces, other chunks their
features. Like vanilla, generated chunks are now saved, proto chunks that
decoration wrote to are saved with their status (terrain or features) and
restored onto the chunk generated again, columns that finish after the player
left are saved, and saving waits for the decorations in flight.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The mesh loop dropped every dirty section of a loaded column outside the
render distance. Columns load up to three chunks past it, and a column
loaded from a save with its light runs no light job, so nothing asked for
its sections again: once the player came closer it stayed invisible until
a block in it changed. Those sections now wait and are put back when the
centre, radius or forced chunks change.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
`status` gains `streaming` (queued jobs, chunks generating, protos waiting for
decoration, decorating, finishing, lighting, meshing) and `dirty_light`, so a
slow load shows which stage it waits on.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Vanilla trims each line and drops only the hidden splash (hash 125780783);
a blank line stays a (blank) splash.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
PalettedContainer::pack hashed all 4096 values with SipHash to build the
palette. It now shares from_values' scan, which only switches to a map
past 16 distinct values. Same palette order, same output.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Every mesh job cloned the whole FaceVisibility (every state, every face
VoxelShape and the occlusion cache) and dropped it afterwards: about half
of the workers' busy time while a world loads. Each worker now keeps its
copy, like vanilla's per-thread occlusion cache, and clones a new one only
when the mesh tables change.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>