perf(es/parser): micro-optimize parser hot paths

[hardfist]

Description:

Micro-optimize swc_ecma_parser hot paths using local Valgrind/Cachegrind simulation data. This builds on the prior parser optimization work and focuses on the TypeScript parser benchmark input.

Changes in this PR:

• Avoid repeated comment-buffer scans and reserve comment maps before publishing buffered comments.
• Reuse static atoms for known keyword/identifier tokens instead of slicing and interning source text on hot identifier paths.
• Fast-path the common primary-expression identifier path and outline rarer arrow-expression recovery.
• Avoid moving boxed expressions for simple variant checks in hot expression parser paths.
• Fast-path plain primary identifiers before the slower rest parser path when Flow pattern matching is not enabled.
• Use a direct token-set check for unary-prefix expression starts.
• Skip and narrow the plain ( subscript path when no_call already forbids consuming calls.
• Keep token/input accessor hot paths inline where measurements showed the combined parser path benefited.

Micro-Optimization Progress

Target benchmark: swc_ecma_parser TypeScript parser input (crates/swc_ecma_parser/benches/files/typescript.js)
Measurement mode: isolated Valgrind Cachegrind run with parser-harness client requests around the parse loop
Primary metric: Cachegrind I refs
Baseline commit: 25b2b1fb5c
Optimized commit: ae6ab6d2eb

Commit	Benchmark	Tool	Metric	Before	After	Delta	Notes
ef0a84219e	typescript.js parse, 1 iter	Cachegrind	I refs	688,371,852	653,923,388	-34,448,464 (-5.005%)	First optimized pass
621283573d	typescript.js parse, 1 iter	Cachegrind	I refs	688,371,852	653,872,357	-34,499,495 (-5.012%)	Adds boxed expression ref checks
d10b5e49b6	typescript.js parse, 1 iter	Cachegrind	I refs	688,371,852	653,859,937	-34,511,915 (-5.014%)	Adds unary-prefix token-set check
5016fa25fa	typescript.js parse, 1 iter	Cachegrind	I refs	688,371,852	650,545,675	-37,826,177 (-5.495%)	Adds primary identifier fast path
e2a78b422c	typescript.js parse, 1 iter	Cachegrind	I refs	688,371,852	650,245,133	-38,126,719 (-5.539%)	Skips plain calls in no-call subscript parse
ae6ab6d2eb	typescript.js parse, 1 iter	Cachegrind	I refs	688,371,852	649,450,821	-38,921,031 (-5.654%)	Narrows the no-call subscript start check
ae6ab6d2eb	typescript.js parse, 1 iter	Cachegrind	I refs	650,245,133	649,450,821	-794,312 (-0.122%)	Incremental vs previous PR commit
ae6ab6d2eb	typescript.js parse, 1 iter	Cachegrind	I1 misses	8,294,027	7,444,699	-849,328 (-10.24%)	Cache simulation enabled; +118,950 (+1.62%) vs previous commit
ae6ab6d2eb	typescript.js parse, 1 iter	Cachegrind	LL misses	1,946,681	1,804,610	-142,071 (-7.30%)	Cache simulation enabled; +75 (+0.004%) vs previous commit

Artifacts kept locally under ignored optimization-artifacts/valgrind/cachegrind/:

• Baseline: 20260516T045236Z-baseline-isolated-typescript-cache-1iter
• Previous final: 20260516T082221Z-no-call-lparen-fast-return-typescript-cache-1iter
• Latest final: 20260516T082607Z-subscript-lparen-branch-typescript-cache-1iter

Validation:

• Latest commit ae6ab6d2eb: git submodule update --init --recursive
• Latest commit ae6ab6d2eb: cargo fmt --all
• Latest commit ae6ab6d2eb: cargo test -p swc_ecma_parser --features verify
• Latest commit ae6ab6d2eb: cargo test -p swc_ecma_parser --features flow,verify
• Latest commit ae6ab6d2eb: cargo clippy --all --all-targets -- -D warnings
• Earlier in this PR: cargo test -p swc_common

Known local baseline issue:

• cargo test -p swc_ecma_parser without verify fails errors_tests__typescript_errors__type_only_import_specifier__invalid_type_only__input_ts; the same one-test command fails on the clean branch HEAD without the latest local diff and passes with --features verify. CI uses cargo test -p swc_ecma_parser --features verify for this crate.

BREAKING CHANGE:

None.

Related issue (if exists):

None.

[changeset-bot]

🦋 Changeset detected

Latest commit: ae6ab6d

The changes in this PR will be included in the next version bump.

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[codspeed-hq]

Merging this PR will improve performance by 2.83%

⚠️ Different runtime environments detected

Some benchmarks with significant performance changes were compared across different runtime environments,
which may affect the accuracy of the results.

Open the report in CodSpeed to investigate

⚡ 11 improved benchmarks
✅ 208 untouched benchmarks
⏩ 31 skipped benchmarks¹

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
⚡	es/large/parser	36.3 ms	35 ms	+3.61%
⚡	typescript/fast-strip	403 µs	389.5 µs	+3.47%
⚡	es/parser/angular	14.4 ms	14 ms	+2.89%
⚡	es/parser/jquery mobile	17.9 ms	17.4 ms	+2.92%
⚡	es/parser/mootools	9 ms	8.8 ms	+2.51%
⚡	es/parser/three	56 ms	54.6 ms	+2.51%
⚡	es/parser/typescript	317.8 ms	308.2 ms	+3.09%
⚡	es/parser/underscore	2 ms	1.9 ms	+2.59%
⚡	es/parser/backbone	2.3 ms	2.2 ms	+2.24%
⚡	es/parser/jquery	11.6 ms	11.3 ms	+2.94%
⚡	es/parser/yui	8.9 ms	8.7 ms	+2.35%

Tip

Curious why this is faster? Comment @codspeedbot explain why this is faster on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing hardfist:codex/ecma-parser-micro-optimize (ae6ab6d) with main (2ab1b8a)²}

Footnotes

1. 31 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

2. No successful run was found on main (4c44d5d) during the generation of this report, so 2ab1b8a was used instead as the comparison base. There might be some changes unrelated to this pull request in this report. ↩