nx_js_sota_bench.nx

nx_js_sota_bench.nx -- JS-ENGINE SOTA BENCHMARK (operator 2026-07-07: "we want to get to state of the art on everything ... july 7th state of the art use the nishi researcher and build it out if it cant give us that benchmark"). A SEPARATE, SOVEREIGN benchmark organ (benchmarks-exist-separate + sovereign-benchmark-not-thirdparty doctrine): measures OUR two engines (tree-walker + bytecode VM) on standard kernel FAMILIES and ranks the result against the RESEARCH-GROUNDED 2026 architecture ladder -- no wave, BEHIND rows printed as BEHIND. RESEARCH GROUNDING (banked 2026-07-07 by nx_jsvm_sota_research_fetch over sovereign TLS, in knowledge/fetched/jsvm_*.raw, corpus-ingested): - jsvm_ignition.raw / jsvm_sparkplug.raw / jsvm_maglev.raw / jsvm_turbofan.raw: V8's 2026 tier ladder = Ignition (bytecode interpreter) -> Sparkplug (baseline JIT straight off bytecode) -> Maglev (mid-tier SSA JIT) -> TurboFan (optimizing JIT). WE ARE at the Ignition tier. - jsvm_hidden_classes.raw + jsvm_inline_caching.raw: SOTA property access = hidden classes (shapes) + inline caches -> O(1) monomorphic loads; OUR obj_find is a LINEAR key scan. - jsvm_gc_tracing.raw: SOTA = generational/tracing GC; OUR arena never frees (honest cap). - jsvm_spidermonkey.raw / jsvm_jsc.raw: same pillar set independently (Baseline+Warp/Ion; LLInt+Baseline+DFG+FTL) -> the pillars below are THE 2026 architecture axis, not V8 trivia. - jsvm_jetstream.raw / jsvm_browser_bench.raw: the field scores on JetStream/Speedometer-class suites; kernels below are our sovereign micro-equivalents (arith/property/calls/strings/ alloc/callbacks). bellard.org (QuickJS reference numbers) was TLS-unreachable (status 0) -- the interpreter-class band below is therefore an ORDER-OF-MAGNITUDE band, stated as such. CLASS BANDS (order-of-magnitude, hot int loop iters/sec on this box): interpreter-class >= 5M · baseline-JIT-class >= 100M · optimizing-JIT-class >= 500M. expect_exit: 0 license_tier: ORIGINAL