Skip to content

History / Vulkan Learned Fallback Readiness

Revisions

  • wiki: Reconcile current implementation and repair navigation Repair the audit's 153 malformed navigation rows and two fragment targets, plus an additional unescaped PowerShell pipe in the validation guide. Add current status, stable/main availability, complete post-snapshot change records, source-derived controls and balanced evidence summaries. Reconcile graphics, audio/CPU and upstream research against b7140848 while preserving dated investigations and failed/withdrawn evidence. Add an offline publication checker with regression coverage for table, target, fragment, metadata and record-drift failures. The wiki Git repository is canonical and publication remains separate from source documentation. Agent-assisted: OpenAI Codex / GPT-6.

    @codex codex committed Nov 1, 2026
  • docs: publish September 26 development status

    @Mainkill1 Mainkill1 committed Sep 26, 2026
  • docs: record exact PR 228 validation Document the exact-head shader and pipeline priority implementation, public Windows/Linux preview, and controlled cold/warm Windows ABBA measurements. Record the narrow supported claim: fewer synchronous shader and pipeline blocking intervals without a demonstrated general frame-tail gain. Preserve Morrowind snapshot output as a correctness gate after the tester reported strange behavior. Add runner lessons for never-executed saved-test sources, waiting-upload head-of-line blocking, and immutable shared snapshot carriers with private per-run overlays. Signed-off-by: Mainkill1 <Mainkill1@gmail.com>

    @Mainkill1 Mainkill1 committed Sep 25, 2026
  • docs: record learned readiness validation rules

    @Mainkill1 Mainkill1 committed Sep 25, 2026
  • docs: clarify disconnected runner ownership

    @Mainkill1 Mainkill1 committed Sep 25, 2026
  • docs: record focused readiness suite lessons

    @Mainkill1 Mainkill1 committed Sep 25, 2026
  • docs: archive learned fallback readiness research Preserve the focused repair, accepted synthetic evidence, negative control, rejected attacks, review lessons, and superseded PR dispositions in the technical wiki.

    @Mainkill1 Mainkill1 committed Sep 25, 2026