[parsnips]

What

Adds an opt-in mimalloc feature to the bin crate. With it, mimalloc::MiMalloc is the process's #[global_allocator] in place of the system allocator.

• crates/bin: optional mimalloc dependency, a mimalloc = ["dep:mimalloc"] feature, and a #[cfg(feature = "mimalloc")] #[global_allocator] in main.rs.
• Root Cargo.toml: mimalloc = { version = "0.1", default-features = false } in [workspace.dependencies]. The crate's secure feature stays off.
• docs/adr/0030-optional-mimalloc-allocator.md: the measurements and why the feature is off by default.
• docs/manuals/04-quickstart-setup-guide.md: the build command.
• .github/workflows/clippy.yml: extenddb/mimalloc joins the test-hooks clippy run, so CI builds the feature.

Default builds do not change.

Why

Closes #953

Measured on the PostgreSQL build of main at 9c3fbfb plus this feature only, on glibc 2.36. The load is conditional inserts with unique keys, sent as PutItem with attribute_not_exists and as PartiQL INSERT, with 32 requests in flight:

Operation	Allocator	Throughput (ops/s), per measurement	Server CPU per request (µs)
PutItem	glibc malloc	15,092, 15,099, 14,992	257, 257, 251
PutItem	mimalloc	15,679, 15,065, 15,415	226, 222, 227
INSERT	glibc malloc	14,613, 6,616*, 14,058	297, 319*, 287
INSERT	mimalloc	14,909, 14,595, 15,134	259, 263, 253

* The host also ran other work, which disturbed this window.

• Server CPU for each request fell by about 12% for both operations.
• Throughput rose by about 2% to 4%, which is close to the variation between measurements.
• Peak resident memory of the server was higher with mimalloc: 398 to 428 MB against 309 to 359 MB after about 600,000 inserts.
• In a perf profile under the PutItem load, libc.so.6 took 16% of the server's samples with glibc malloc and 5% with mimalloc.

An earlier revision of this description gave figures from a fork build that carried two unrelated write-path changes (28% of CPU in libc.so.6, 8.6% more throughput, 23% less CPU). The figures above replace them. The ADR describes the workload and the host.

Testing done

• cargo fmt --check: clean.
• cargo build -p extenddb and cargo build -p extenddb --features mimalloc: both build. nm shows mi_malloc_aligned and mi_free in the feature build only.
• cargo test -p extenddb --features mimalloc: 3 passed.
• cargo clippy -p extenddb --features mimalloc -- -W clippy::pedantic -W clippy::unwrap_used -W clippy::expect_used: no new warnings in crates/bin. The two warnings there are missing backticks on doc lines 4 and 10 of main.rs, and they are on main too.
• I did not run cargo test --workspace. The library crates and their test binaries do not see the feature, because the allocator is set only in the bin crate.
• Benchmark: the table above. Both builds came from this branch, cargo build --release --locked -p extenddb with and without --features mimalloc, in a Debian 12 container with frame pointers on.
• The checks above ran on the first revision. Since then the changes are a code comment, the ADR and the clippy workflow line. I have not run the new clippy command locally.

Checklist

• I have read CONTRIBUTING.md
• All tests pass (cargo test --workspace): not run; only the bin crate changes (see above)
• Code is formatted (cargo fmt --check)
• Clippy is clean (cargo clippy -- -W clippy::pedantic): no new warnings in the changed crate
• I have added or updated tests for new functionality: none; a global allocator has no unit-testable behaviour here
• I have updated documentation if behavior changed
• Breaking changes are noted below (if any)
• If this changes the wire protocol, Storage trait, auth model, on-disk
  format, or public CLI surface, an RFC has been accepted or is linked
  below. Otherwise, an ADR captures the decision (link below).

ADR / RFC: ADR-0030. The number 0030 is the next unused one after the index's 0029. Renumber it if 0028 or another number is reserved.

---

By submitting this pull request, I confirm that my contribution is made under
the terms of the Apache License 2.0 and I agree to the Developer Certificate of
Origin (DCO). See CONTRIBUTING.md for details.

🤖 Generated with Claude Code

[LeeroyHannigan]

Thanks for the measurements and the ADR. I read the change against main at 9c3fbfb. Default builds do not change, and the allocator applies only to the bin crate. The CI runs for this PR wait for a maintainer to approve them, so no check has run yet.

[LeeroyHannigan]

I read 83320bb. The clippy run at clippy.yml:39 now builds extenddb/mimalloc, and the comment in main.rs points at ADR-0030. The ADR now measures main at 9c3fbfb. It reports about 12% less server CPU per request and a throughput change close to the noise. It also reports higher peak resident memory with mimalloc. Clippy, with the new feature run, and Licenses pass at this head.