mimalloc

Microsoft's mimalloc is a compact, high-performance memory allocator that delivers up to 5.3x faster allocation than glibc malloc under heavy multithreaded workloads, with ~50% less RSS. It excels at allocations <= 4 KB. On allocation-heavy workloads, mimalloc delivers a measured 13–22% overall speedup versus system malloc — making it, alongside jemalloc and tcmalloc, the highest-leverage one-line performance change available to most projects.

The allocator dominates the container

The single most important framing the fastest-dynamic-arrays research established: switching from glibc to mimalloc moves the needle more than switching from C to Rust. On a 100M uint64_t push_back benchmark, the language ranking (C, Rust, C++) had all entries within ~10% of each other; the allocator swap caused a 2–3× swing. This reorders almost every received opinion about container performance — std::vector with mimalloc is typically faster than folly-fbvector with system malloc, despite the latter being a far more sophisticated container.

Usage

#[global_allocator]
static GLOBAL: mimalloc::MiMalloc = mimalloc::MiMalloc;

One line in main.rs and you're done. This is arguably the single lowest-effort, highest-impact optimization available to any Rust project.

mimalloc vs jemalloc vs tcmalloc

Use mimalloc for cross-platform builds and small-allocation-heavy code. Use jemalloc (via tikv-jemallocator in Rust) for servers with mixed allocation sizes and long-running workloads where fragmentation resistance matters — and when integrating with folly-fbvector or folly-f14, which depend on jemalloc's xallocx. Use tcmalloc when 4 MB+ allocations dominate; tcmalloc hits 50× system libmalloc throughput at that size. On musl targets (Alpine/static builds), any of the three is nearly mandatory — musl's default causes a 7x slowdown.

Cross-language benchmarks confirm mimalloc's dominance: 13% faster than tcmalloc on the Lean compiler, 15% lower P99 latency than jemalloc for small allocations. Arena/bump allocators like bumpalo deliver 2–5x speedup over general malloc for batch allocations. See rust-memory-allocators for the full Rust comparison and fastest-dynamic-arrays for why this single choice dominates dynamic-array performance.