Clarify Huax-1839's Metal renderer credit Attribute the original shader compiler, rewrites, pipeline cache and Metal state tables to Huax-1839 and the later command layer and frame encoder to Wooloomooloo2. Link Huax-1839's first public code snapshot in the mirrored credits and renderer docs.
Render adapter M2 step 1 done (runtime 1dda5fb)
Render adapter: M1 done (resources made in Metal)
Render adapter: A6 done; performance: mutable-format images
Render adapter: A5 done
Render adapter: A4 done Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Render adapter: A3 done Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Render adapter: A2 done Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Render adapter: A1 done Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Render adapter: A0 done Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Render adapter: approved specification Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Document character lighting collector fix Mirror the runtime stability and Metal renderer conclusions after the native 1440p copied-save checks.
Mirror runtime stability and MetalFX notes Update the wiki with the GX worker and shader cache fixes, native renderer status, MetalFX measurements and contributor credit.
Document Metal recorder lifetime fix Mirror the runtime stability and native renderer findings after the shader lifetime fix.
Record native 1080p crash scope
Mirror the docs: one page per doc, generated from the repository The wiki now mirrors the repository's docs (rewritten 2026-10-20): the same pages and words, with links between them as wiki links. Pages without a doc any more are removed. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Native renderer: the Metal frame path with the recompiled CPU code
Native renderer: the route change (N5 before N4's command stream) and the status throughout
Progress 2026-10-20: settings sync and menu, Metal renderer (experimental), notices over the game only
Record guarded opening-video results and release QA Sync the runtime docs: confirmed shared-state reset, ten guarded completions versus three faults in five controls, measured CPU cost, uncapped pair and isolated test helper. Keep N4 paused and independent GPU assertions under observation. Co-Authored-By: OpenAI Codex <codex@openai.com>
Record release opening-video stability comparison Mirror the five-pair release/main result, earlier trial and assertion counts, and N4's pause. Co-Authored-By: OpenAI Codex <codex@openai.com>
Record passing N4 resource CPU and Linux gates Mirror the capped comparisons, passing Linux build, and next resource-lifetime work. Co-Authored-By: OpenAI Codex <codex@openai.com>
Record first N4 buffer slice and A/B gates Mirror the opt-in resource-description seam and its measured CPU limitation, with N4 progress at ten percent. Co-Authored-By: OpenAI Codex <codex@openai.com>
Record N4 contract, baseline and SRT follow-up Mirror the approved command-stream migration and hand-port policy, the measured Vulkan reference, and the ten-run opening-video comparison. Add the renderer-port decision log to the wiki navigation. Co-Authored-By: OpenAI Codex <codex@openai.com>
Refine N4 backend contract Mirror the runtime design for POD command recording, hazard ownership, flat bindings, CPU gates, Linux CI and pending upstream merge choice. Co-Authored-By: OpenAI Codex <codex@openai.com>
Document N1 trace and N4 backend design Keep the native renderer roadmap, measured clear and tessellation results, plume evaluation, and backend interface proposal in step with the runtime docs. Co-Authored-By: OpenAI Codex <codex@openai.com>
Document fixed quad triangle experiment Record the opt-in path, alternating clinic results, remaining CPU and pacing regression, and N1 acceptance gate to match the runtime roadmap. Co-Authored-By: OpenAI Codex <codex@openai.com>
Record deferred-clear causes and untraced pairs Document the upscaler redirect fix, reason counts at 16,384 clears, and three alternating 90-second off/on pairs. Keep the N1 progress bar at 25% while tessellation and copy planning remain open. Co-Authored-By: OpenAI Codex <codex@openai.com>
Track N1 fixed tessellation preparation Add 20-block N0–N6 progress bars and the four N1 acceptance gates. Record the validated fixed quad shader work and keep N1 at 25% until a triangle draw path passes visual and trace comparisons.
Record N1 tessellation and copy-batching findings Keep the native renderer and macOS performance wiki pages in step with the runtime docs. The tested copy-pair grouping showed no repeatable gain; the tessellated light path requires a shader-stage rewrite.