perf(ecmascript): avoid heap-allocating Math.min/max/imul operands#23941
Merged
camc314 merged 4 commits intoJul 5, 2026
Conversation
Merging this PR will not alter performance
Comparing Footnotes
|
linyiru
force-pushed
the
perf/ecmascript-math-variadic-arena
branch
from
June 29, 2026 15:31
c09ccc4 to
62c4a52
Compare
linyiru
marked this pull request as ready for review
June 29, 2026 15:34
camc314
reviewed
Jun 29, 2026
linyiru
force-pushed
the
perf/ecmascript-math-variadic-arena
branch
from
July 4, 2026 03:06
b0e170c to
d65f2ee
Compare
`try_fold_math_variadic` constant-folds `Math.min/max/imul(...)` by collecting the operand values into a `Vec<f64>`, which is a system allocation on every foldable call — a hot minifier path. The evaluation context already exposes the arena (`ConstantEvaluationCtx: GetAllocator`), so collect into an `ArenaVec` instead. The list is transient and only iterated/indexed, so behaviour is unchanged; the system allocation is gone. `cargo allocs` (minifier system allocations): kitchen-sink.tsx 2576 -> 2312 (-264) antd.js 1040 -> 870 (-170) pdf.mjs 2569 -> 2476 (-93) App.tsx 78 -> 72 (-6)
Addressing review (don't put the transient operand list in the arena). A stack `SmallVec` also beats `Vec::with_capacity(args.len())`: most `Math.min/max/imul` calls bail early when an arg isn't a constant, and `Vec::new()` doesn't allocate on that path while `with_capacity` allocates upfront — so it would *increase* allocations. `SmallVec` is zero-alloc on both the early-bail and the small-fold paths, and never touches the arena. cargo allocs (minifier system allocations, vs main; arena counts unchanged): kitchen-sink.tsx 2576 -> 2312 antd.js 1040 -> 870 pdf.mjs 2569 -> 2476
camc314
force-pushed
the
perf/ecmascript-math-variadic-arena
branch
from
July 5, 2026 13:09
d65f2ee to
4d29d1e
Compare
Boshen
added a commit
that referenced
this pull request
Jul 6, 2026
### 🚀 Features - 260425f semantic/examples: Include unresolved references (#24214) (camc314) - 2d9b0b3 minifier: Fold boolean-literal ternary branches in value contexts (#24110) (Dunqing) - 61fbf10 ast: Implement `ReplaceWith` on all AST types (#24013) (overlookmotel) - 7db7a29 allocator: Add `ReplaceWith` trait (#24012) (overlookmotel) - 4eb074e mangler: Add `reserved` option for names that must not be mangled (#24041) (Dunqing) - 2e62012 data_structures: Add `StringExt` trait (#24006) (overlookmotel) - 60e7160 minifier: Drop side-effect-free IIFEs whose result is unused (#23967) (Dunqing) - 26dd9e2 ast: Add method to widen inherited enum ref to parent ref (#23961) (overlookmotel) ### 🐛 Bug Fixes - e8b50ee transformer: Clean up semantics for stripped TypeScript syntax (#24180) (camc314) - d966d0b react_compiler: Remove clippy allows (#24168) (Boshen) - 854ef8d react_compiler: Compile generic functions instead of over-bailing on type-param hoisting (#24158) (Boshen) - 093586c react_compiler: Align memoization cache-slot allocation with Babel (#24157) (Boshen) - 09c8f59 react_compiler: Normalize snapshot fixture paths (#24142) (camc314) - f13df97 react_compiler: Drop stray empty statement from catch bindings (#24133) (Boshen) - cb2a505 react_compiler: Codegen destructuring reassignment targets (#24131) (Boshen) - b82c394 react_compiler: Propagate codegen invariants instead of emitting empty bodies (#24128) (Boshen) - 5771982 react_compiler: Render unchanged programs as source in fixture snapshots (#24129) (Boshen) - 4b16e1a transformer/async-to-generator: Preserve direct eval scope flags (#24136) (camc314) - 4e9194f react_compiler: Lower `delete obj.prop` to Property/ComputedDelete (#24123) (Boshen) - 0b25582 ast: Type binding node `typeAnnotation` as `TSTypeAnnotation | null` (#23113) (Boshen) - 018c0e5 transformer: Hoist lowered async declarations (#22770) (camc314) - 652fbaf mangler: Keep names of destructured exported bindings (#24036) (Dunqing) - e274415 minifier: Don't drop global calls that throw despite pure arguments (#23917) (Dunqing) - 59abb30 minifier: Only merge string literals in `try_fold_add` when the inner operator is `+` (#23622) (Jerry Zhao) ### ⚡ Performance - c5ca77b transformer: Avoid cloning refresh options (#24191) (camc314) - bf1a151 react_compiler: Compile out debug printers (#24184) (Boshen) - abb44a0 transformer: Build fixed object-rest arguments (#24190) (camc314) - a4db731 isolated_declarations: Use `ReplaceWith` instead of `TakeIn` (#24016) (overlookmotel) - ff10855 transformer: Use `ReplaceWith` instead of `TakeIn` (#24015) (overlookmotel) - bd49aff ecmascript: Avoid heap-allocating Math.min/max/imul operands (#23941) (Lawrence Lin) - e4b708b react_compiler: Skip compiled files before prefilters (#24171) (Boshen) - c59f2fe rust: Return impl ExactSizeIterator from slice-backed accessors (#24144) (Boshen) - 5d6d04a codegen: SWAR-skip boring byte runs in sourcemap line/column scan (#24023) (Boshen) - a55e0be traverse: Reduce string operations in `get_var_name_from_node` (#24007) (overlookmotel) - e6d48e1 transformer/nullish_coalescing: Move cold path into separate function (#23989) (overlookmotel) - c4e35b5 transformer/object_rest_spread: Pre-allocate capacity in `Vec` (#23988) (overlookmotel) - 527b8e5 transformer/decorators: Narrow type earlier (#23987) (overlookmotel) ### 📚 Documentation - 30d17f5 allocator: Clarify docs for `TakeIn::take_in_box` (#24093) (overlookmotel) - 675e6a8 ast: Correct doc comment for `PrivateFieldExpression` (#24008) (overlookmotel) - e4c30e6 minifier: Explain what `dce` mode means (#23994) (Dunqing) - 37cbf88 ast_macros: Document fields of `StructDetails` (#23959) (overlookmotel) - 4de3e54 ast: Correct doc comment (#23948) (overlookmotel) Co-authored-by: Boshen <[email protected]>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
try_fold_math_variadicconstant-foldsMath.min/Math.max/Math.imul(...)bycollecting the operand values into a
Vec<f64>. That is a system allocation on everyfoldable call, on the minifier's constant-folding hot path.
The evaluation context already exposes the arena (
ConstantEvaluationCtx: GetAllocator,and the module already imports
ArenaVec), so collect into anArenaVecinstead. Thelist is transient and only iterated / indexed (
.iter(),.first(),.get()), so thechange is behaviour-identical — the system allocation is simply gone.
Results —
cargo allocs(minifier system allocations)Tests
cargo test -p oxc_ecmascript --lib --tests— passcargo test -p oxc_minifier --lib --tests— 539 passAI disclosure
Prepared with AI assistance (Claude). I located the hotspot with a per-call-site allocation
profile across real files, confirmed the list is transient/iterated-only, and verified the
reduction with
cargo allocsplus the test suites.