Skip to content

docs: record kDisco PR #52 Kotlin 2.3.21 toolchain upgrade as performance-neutral for fast-sim#761

Open
bedaHovorka with Copilot wants to merge 2 commits into
toKdisco0.6.1from
copilot/evaluate-kdisco-pr-52
Open

docs: record kDisco PR #52 Kotlin 2.3.21 toolchain upgrade as performance-neutral for fast-sim#761
bedaHovorka with Copilot wants to merge 2 commits into
toKdisco0.6.1from
copilot/evaluate-kdisco-pr-52

Conversation

Copilot AI commented Jul 12, 2026

Copy link
Copy Markdown
Contributor

Issue #752 evaluated whether upgrading kDisco's Kotlin 2.1.10 → 2.3.21 (kdisco#52) would speed up fast-sim's ~1.85M-derivative-eval hot path. Result: performance-neutral — 341 ms vs 349 ms, within ±7% run-to-run noise.

Changes

  • docs/FAST_SIM_BENCHMARK.md — adds a supplementary section recording:
    • Measurement methodology (shuntingLoop 300, median of 7, native linuxX64 release binary)
    • Side-by-side results table (kDisco 0.6.0/Kotlin 2.1.10 vs 0.6.0-k2321-SNAPSHOT/Kotlin 2.3.21)
    • Conclusion: merge kdisco#52 on toolchain-currency grounds, not as a performance lever
    • Scope caveat: the dominant hot code (Train.Motor.derivatives) is compiled by interlockSim's own Kotlin version, not kDisco's — so kDisco's toolchain version is irrelevant to Zero-crossing event detection: dtMax=1e-3 forces 1.85M derivative evals/run (60-70% of fast-sim wall) #750
    • Link to kdisco CI run 29184671509 ✅ (branch to0.6.1, conclusion: success)

Copilot AI changed the title [WIP] Evaluate kDisco PR #52 performance-neutral findings docs: record kDisco PR #52 Kotlin 2.3.21 toolchain upgrade as performance-neutral for fast-sim Jul 12, 2026
Copilot AI requested a review from bedaHovorka July 12, 2026 07:57
@bedaHovorka bedaHovorka marked this pull request as ready for review July 12, 2026 08:04
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Evaluate kDisco PR #52 (Kotlin 2.3.21) as a fast-sim optimisation — measured: performance-neutral

2 participants