oxc-project/oxc - CodSpeed

oxc

Blog Docs Changelog

Performance History

Latest Results

perf(semantic): preserve precomputed identifier hash throughout semantic analysis - Change `symbol_names` storage from `Vec<Atom>` to `Vec<Ident>` to preserve hash - Update `create_symbol` and `set_symbol_name` to accept `Ident` instead of `&str` - Use `clone_in` in scoping methods instead of reconstructing from `&str` - Update traverse context APIs to use `Ident<'_>` instead of `&str` - Use `Ident::new_const` for compile-time constant strings ("module", "Object") - Use `Ident::from(Atom)` instead of `Ident::from(atom.as_str())` - Modify `base54` to compute hash incrementally while building name, returning `Ident` directly - Update mangler to use `Ident` throughout, eliminating rehashing of mangled names Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

perf-all

3 hours ago

perf(semantic): preserve precomputed identifier hash throughout semantic analysis - Change `symbol_names` storage from `Vec<Atom>` to `Vec<Ident>` to preserve hash - Update `create_symbol` and `set_symbol_name` to accept `Ident` instead of `&str` - Use `clone_in` in scoping methods instead of reconstructing from `&str` - Update traverse context APIs to use `Ident<'_>` instead of `&str` - Use `Ident::new_const` for compile-time constant strings ("module", "Object") - Use `Ident::from(Atom)` instead of `Ident::from(atom.as_str())` - Modify `base54` to compute hash incrementally while building name, returning `Ident` directly - Update mangler to use `Ident` throughout, eliminating rehashing of mangled names Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

perf-all

5 hours ago

feat(benchmark): add pipeline benchmark for full compilation flow (#18484)

main

5 hours ago

feat(benchmark): add pipeline benchmark for full compilation flow Add a new benchmark that measures the complete compilation pipeline: allocate -> parse -> semantic -> transform -> minify -> mangle -> codegen -> drop Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

feat/pipeline-benchmark

5 hours ago

feat(oxc_str): upgrade Ident with precomputed hash for fast equality (#18400) ## Summary Upgrade `Ident` to store a precomputed hash for fast equality checks and efficient hash map operations. ### Changes - Change `Ident` layout from simple `&str` wrapper to `NonNull<u8>` + `len: u32` + `hash: u32` - Fast `PartialEq` between `Ident`s: compare length and hash first before doing full string comparison - `Hash` impl uses precomputed hash (no `Borrow<str>` - lookups must use `&Ident`) - Add `Ident::new()` and `Ident::len()` methods - Add specific `PartialEq` implementations for `&Ident`, `&str`, `String`, `Atom`, `&Atom` - Add `rustc-hash` dependency for `FxHasher` - Fix enum layout calculation in ast_tools to round size to alignment ### Layout - **64-bit**: 16 bytes, align 8 (ptr 8 + len 4 + hash 4) - **32-bit**: 12 bytes, align 4 (ptr 4 + len 4 + hash 4) Using two separate `u32` fields instead of a single `u64` keeps 4-byte alignment on 32-bit platforms, avoiding size increases in AST structs. 🤖 Generated with [Claude Code](https://claude.ai/code)

refactor/ident-precomputed-hash

6 hours ago

perf(semantic): preserve precomputed identifier hash throughout semantic analysis - Change `symbol_names` storage from `Vec<Atom>` to `Vec<Ident>` to preserve hash - Update `create_symbol` and `set_symbol_name` to accept `Ident` instead of `&str` - Use `clone_in` in scoping methods instead of reconstructing from `&str` - Update traverse context APIs to use `Ident<'_>` instead of `&str` - Use `Ident::new_const` for compile-time constant strings ("module", "Object") - Use `Ident::from(Atom)` instead of `Ident::from(atom.as_str())` - Modify `base54` to compute hash incrementally while building name, returning `Ident` directly - Update mangler to use `Ident` throughout, eliminating rehashing of mangled names Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

perf-all

6 hours ago

feat(oxc_str): upgrade Ident with precomputed hash for fast equality (#18400) ## Summary Upgrade `Ident` to store a precomputed hash for fast equality checks and efficient hash map operations. ### Changes - Change `Ident` layout from simple `&str` wrapper to `NonNull<u8>` + `len: u32` + `hash: u32` - Fast `PartialEq` between `Ident`s: compare length and hash first before doing full string comparison - `Hash` impl uses precomputed hash (no `Borrow<str>` - lookups must use `&Ident`) - Add `Ident::new()` and `Ident::len()` methods - Add specific `PartialEq` implementations for `&Ident`, `&str`, `String`, `Atom`, `&Atom` - Add `rustc-hash` dependency for `FxHasher` - Fix enum layout calculation in ast_tools to round size to alignment ### Layout - **64-bit**: 16 bytes, align 8 (ptr 8 + len 4 + hash 4) - **32-bit**: 12 bytes, align 4 (ptr 4 + len 4 + hash 4) Using two separate `u32` fields instead of a single `u64` keeps 4-byte alignment on 32-bit platforms, avoiding size increases in AST structs. 🤖 Generated with [Claude Code](https://claude.ai/code)

refactor/ident-precomputed-hash

8 hours ago

perf(coverage): simplify architecture and cache test data upfront (#18478) Local: 2.4x faster (31s → 13s) CI: saved 50s

main

8 hours ago

Active Branches

perf(all): measure overall performance for `Ident` change

last run

3 hours ago

-20%

feat(oxc_str): upgrade Ident with precomputed hash for fast equality

last run

6 hours ago

-4%

feat(linter/eslint): eslint/object-shorthand

last run

10 hours ago

-1%

© 2026 CodSpeed Technology

Home Terms Privacy Docs