NagyZoltanPeter be6bfed619

docs(bench): same-session pre-phase-2 vs post-phase-3 speed+GC comparison ...

Back-to-back A/B on one machine: micro +85.9%, macro +56.6% msg/s;
decode/hash byte volumes halved. Peak heap (getMaxMem) -21%/-20.7%.
GC-sampling series shows retained-memory slope is workload-driven and
unchanged between versions; the transient-copy reduction is not
observable via occupied/total sawtooth under refc (deterministic frees,
reused arena) and manifests as a lower absolute heap level, not a
gentler slope.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

2026-07-15 09:16:11 +02:00

..

async_copy_analysis.md

docs: async copy analysis and 3-phase elimination plan

2026-07-15 04:07:59 +02:00

async_copy_fix_plan.md

docs: async copy analysis and 3-phase elimination plan

2026-07-15 04:07:59 +02:00

bench_baseline.md

docs(bench): Phase 3 results — micro 1.0 hash/msg @ ~2220 msg/s, macro 3/3 @ 356 msg/s

2026-07-15 05:52:52 +02:00

phase2_mutation_audit.md

docs(bench): Phase 2 checkpoint + mutation-site audit

2026-07-15 05:03:22 +02:00

plan_phase1_bench.md

docs: async copy analysis and 3-phase elimination plan

2026-07-15 04:07:59 +02:00

plan_phase2_wakumessage_ref.md

docs: async copy analysis and 3-phase elimination plan

2026-07-15 04:07:59 +02:00

plan_phase3_wakuenvelope.md

docs: async copy analysis and 3-phase elimination plan

2026-07-15 04:07:59 +02:00

speed_gain_pre2_post3.md

docs(bench): same-session pre-phase-2 vs post-phase-3 speed+GC comparison

2026-07-15 09:16:11 +02:00