perf: add a mimalloc allocation fast path

[LingyuCoder]

Motivation

Avoid the overhead of mimalloc's aligned allocation API for allocations with ordinary alignment requirements, and make the cached-source hash benchmarks measure actual lookups.

Changes

Implement the shared mimalloc adapter in rspack_allocator. Route requests with alignment up to align_of::<usize>() and no larger than the allocation size through mi_malloc / mi_zalloc, retaining the aligned fallback and existing deallocation and reallocation behavior. Register the binding's global allocator in rspack_binding_api, preserving its allocator selection, while benchmarks register their own wrapper around the shared implementation.

Separate cached-source lookups from single and batched HashMap lifecycles. Execute both lookups and include allocation and destruction in the lifecycle measurements, using new benchmark names for the changed workloads.

[github-actions]

📦 Binary Size-limit

Comparing 91bc1f3 to fix: revert "feat(core): expose NormalModule beforeLoaders hook to JS" (#15660) by AsyncIter

🎉 Size decreased by 31.95KB from 68.85MB to 68.82MB (⬇️0.05%)

[github-actions]

Rsdoctor Bundle Diff Analysis

Found 5 projects in monorepo, 0 projects with changes.

📊 Quick Summary

Project	Total Size	Gzip Size	Change	Gzip Change
popular-libs	1.7 MB	523.6 KB	0	0
react-10k	5.3 MB	1.3 MB	0	0
react-1k	799.2 KB	217.5 KB	0	0
react-5k	2.6 MB	662.0 KB	0	0
ui-components	4.9 MB	1.4 MB	0	0

Generated by Rsdoctor GitHub Action

[codspeed]

Merging this PR will improve performance by 3.47%

⚠️ 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

⚡ 27 improved benchmarks
✅ 22 untouched benchmarks
🆕 3 new benchmarks
⏩ 48 skipped benchmarks¹

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
⚡	rust@create_module_hashes	7.6 ms	7.1 ms	+7.82%
⚡	rust@mangle_exports	3.1 ms	3 ms	+5.68%
⚡	rust@create_concatenate_module_unsupported_syntax	17.6 ms	16.7 ms	+4.85%
⚡	rust@concatenate_module_code_generation	43.7 ms	41.7 ms	+4.76%
⚡	rust@create_concatenate_module	13.7 ms	13.1 ms	+4.44%
⚡	rust@build_chunk_graph	26.1 ms	25 ms	+4.37%
⚡	rust@build_module_graph	448 ms	429.3 ms	+4.34%
⚡	rust@create_concatenate_module_bailouts	97.7 ms	93.7 ms	+4.28%
⚡	rust@create_concatenate_module_shared_roots	17.4 ms	16.8 ms	+4.04%
⚡	rust@flag_dependency_exports	5.1 ms	4.9 ms	+3.54%
⚡	rust@create_chunk_assets	1.1 ms	1.1 ms	+3.5%
⚡	sources@concat_source_add_few	583.9 ns	564.2 ns	+3.5%
⚡	rust@create_named_module_ids	3.2 ms	3.1 ms	+3.45%
⚡	bundle@css-modules-development	1.9 s	1.8 s	+3.32%
⚡	rust@create_module_assets	169.6 µs	164.5 µs	+3.12%
⚡	bundle@misc-development	2.8 s	2.7 s	+3.1%
⚡	rust@build_swc-loader	1,003.5 ms	974.7 ms	+2.96%
⚡	rust@flag_dependency_usage	10 ms	9.7 ms	+2.93%
⚡	bundle@basic-react-development	151 ms	146.7 ms	+2.87%
⚡	rust@create_full_hash	2.3 ms	2.3 ms	+2.49%
...	...	...	...	...

ℹ️ Only the first 20 benchmarks are displayed. Go to the app to view all benchmarks.

Tip

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

---

_{Comparing seal/try-mimalloc-fastpath (91bc1f3) with main (22f1dd8)}

Footnotes

1. 48 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. ↩