Problem / 问题
Context creation still takes about 2.27 ms after release cleanup optimization (PR #4). The retest records 778 layout replacements and 943 shape lookups per Context; layout replacement accounts for about 62.6% of sampled CPU periods within Context creation. Lazy builtin-property installation repeatedly copies and replaces progressively larger layouts.
Context 初始化逐项登记内建懒属性，反复复制布局、查询 shape 并销毁中间布局。优化目标是批量安装合适的方法表，保留现有懒初始化与所有权机制。
Implementation scope / 实现范围
Add an internal batch entry that validates a table, retains key roots, copies the existing layout once, appends descriptors in order and calls existing replace_layout once. Pilot a Date table without intermediate read dependencies; expand suitable Array/TypedArray tables only after validation. Preserve Date toUTCString/toGMTString alias identity and explicit materialization boundaries, property flags/order/name/length, realm ownership and cleanup.
不重写 Heap 所有权架构，不加入少量引用保留快路径。复用现有布局和引用管理，核对准备失败与发布后错误的契约。
Stack and validation / 分支与验证
Based on personal PR #4: Eric-Song-Nop#4, branch perf/release-cleanup-fast-path at 8eb2a1e. Implementation PR targets that branch in Eric-Song-Nop/quickjs-oxide. Issues remain upstream because the personal fork has Issues disabled.
Separately measure baseline, Date pilot and expanded tables. Cover lifecycle/single startup, the existing full V8-v7 suite and microbench matrix, representative phase timing, CPU samples and diagnostics. Preserve timeout/failure/regression evidence and avoid extrapolating Richards to all workloads. Run Rust tests, strict production Clippy, full Test262 before/after comparison and binary-boundary canaries without updating frozen baselines. All heavy work runs serially on Herdr/PocketLab. Deliver bilingual report and raw evidence.
Problem / 问题
Context creation still takes about 2.27 ms after release cleanup optimization (PR #4). The retest records 778 layout replacements and 943 shape lookups per Context; layout replacement accounts for about 62.6% of sampled CPU periods within Context creation. Lazy builtin-property installation repeatedly copies and replaces progressively larger layouts.
Context 初始化逐项登记内建懒属性，反复复制布局、查询 shape 并销毁中间布局。优化目标是批量安装合适的方法表，保留现有懒初始化与所有权机制。
Implementation scope / 实现范围
Add an internal batch entry that validates a table, retains key roots, copies the existing layout once, appends descriptors in order and calls existing replace_layout once. Pilot a Date table without intermediate read dependencies; expand suitable Array/TypedArray tables only after validation. Preserve Date toUTCString/toGMTString alias identity and explicit materialization boundaries, property flags/order/name/length, realm ownership and cleanup.
不重写 Heap 所有权架构，不加入少量引用保留快路径。复用现有布局和引用管理，核对准备失败与发布后错误的契约。
Stack and validation / 分支与验证
Based on personal PR #4: Eric-Song-Nop#4, branch perf/release-cleanup-fast-path at 8eb2a1e. Implementation PR targets that branch in Eric-Song-Nop/quickjs-oxide. Issues remain upstream because the personal fork has Issues disabled.
Separately measure baseline, Date pilot and expanded tables. Cover lifecycle/single startup, the existing full V8-v7 suite and microbench matrix, representative phase timing, CPU samples and diagnostics. Preserve timeout/failure/regression evidence and avoid extrapolating Richards to all workloads. Run Rust tests, strict production Clippy, full Test262 before/after comparison and binary-boundary canaries without updating frozen baselines. All heavy work runs serially on Herdr/PocketLab. Deliver bilingual report and raw evidence.