Repository navigation
Jet Set Radio Future - Windows preview 2026.11.01.1
Pre-releaseRelease 2026.11.01.1 — retained Xbox optimization work
The original 27 architecture areas have been evaluated: 13 retained areas
(five previously integrated and eight later retained), ten tested candidates
not retained, three existing-baseline areas and one reuse evaluation.
R09 enhanced rendering is deferred and excluded from this release.
Retained changes include native draw/pipeline and memory ownership, immutable
preload file views, revision-qualified resource mirrors and scalar preparation,
pristine spatial-voice processing, bounded GP DSP caches/DMA, GPU packed-movie
uploads, owned replay/Session observations and one complete launch admission.
Selections remain game/revision-specific; experimental rejected paths are absent.
Individual earlier CPU/RAM/I/O observations concern separate packages and routes;
their percentages are not additive. No general FPS gain is claimed from the final combined capped-rate checks.
Original games and Dashboard files remain user-supplied. No original game bodies,
console keys, firmware, media or player saves are included. Retain existing data,
console identity and save folders when upgrading; saved preferences win.
JSRF remains an experimental development preview; assisted area-entry coverage
does not establish fair combat, complete routes or end-to-end completion.
Both supported NTSC-U revisions are included and selected by exact original XBE
identity. USA enables resource mirrors, pristine spatial processing, GP caches/
DMA and Session-page ownership; revision 4 retains those original paths. Both
input readers remain off. The two previously approved RPTK/RAID graffiti images
and their notices are retained; personal images are not included. Hold-Start
cutscene skipping remains optional/off. Known corner-angle decal checks remain
limited as documented; this release makes no new exhaustive visual claim.
Final USA native-save write and fresh-process original load passed, along with targeted rev4 correctness and owner visual/audio clearance. One native-resolution USA pair retained about 60 source updates/s while peak private committed memory fell by 1.286 GiB. The observed 19.46 to 13.75 CPU ms/source-swap difference is confounded by substantial host-load variation and is not an isolated CPU-performance claim. Rev4 was not performance-profiled.