Mold Linker 3.0 Released: Rui Ueyama Rewrites Ultra-Fast Linker from C++ to Rust, Drops oneTBB

The Three-Question Framework

What Happened?

On October 5, 2026, legendary systems programmer Rui Ueyama (creator of LLVM's lld and mold) officially tagged the mold 3.0.0 release. The milestone release represents a total, ground-up rewrite of mold from 60,000 lines of modern C++20 into pure Rust. Alongside the language migration, mold 3.0 completely removes Intel's Threading Building Blocks (oneTBB) dependency, switching to Rust's native Rayon parallel execution runtime, and transitions its primary build system from CMake to Cargo.

Why It Matters

Mold is the fastest ELF and Mach-O linker in software development, widely utilized by systems engineers and large infrastructure teams to slash binary linking times from minutes to fractions of a second. However, maintaining high-throughput multithreaded object code processing in C++ introduced persistent concurrency synchronization edge cases and complex platform-specific threading dependencies. The Rust rewrite eliminates memory-safety and data-race hazards during complex symbol resolution, reduces peak RSS memory consumption by 22% on massive 10GB binary link targets, and makes cross-compiling the linker trivial across Linux, macOS, and BSD systems.

What Could Happen Next

With mold 3.0 stabilized in Rust, Ueyama plans to leverage Rust's ecosystem to implement first-class WebAssembly (Wasm) linking targets and experimental Mach-O ARM64 optimization pipelines. Major Linux distributions and enterprise CI/CD toolchains are updating their packaging scripts to incorporate the Rust-based mold binary into standard developer toolchains.

Executive Key Takeaways

Complete Rust Migration: 60,000 lines of C++20 rewritten into idiomatic, memory-safe Rust in mold 3.0.

Removal of oneTBB: Intel's heavy threading runtime replaced with lightweight Rust Rayon work-stealing parallelism.

22% Peak Memory Reduction: Rust memory layouts and string interning cut RSS overhead on massive executables.

Sub-Second Link Performance Preserved: Maintains sub-second link times, linking Chromium in 1.4 seconds on 64-core systems.

Streamlined Cargo Toolchain: Simplifies building, vendoring, and cross-compilation across diverse target platforms.

In one of the most notable infrastructure modernizations of the year, systems engineer Rui Ueyama officially released mold 3.0.0 on October 5, 2026, completing an exhaustive ground-up rewrite of the high-speed Unix linker from C++20 into Rust.

Since its initial release in 2021, mold gained international acclaim by dramatically outperforming traditional GNU ld, GNU gold, and LLVM lld linkers. On multi-core development workstations, mold reduced the time required to link massive codebases—such as the Chromium browser or the Linux kernel—from tens of seconds down to barely one second, fundamentally transforming the developer edit-compile-test feedback loop.

However, achieving that extreme performance in C++ required intricate lock-free data structures, manual memory mapping, and reliance on Intel's Threading Building Blocks (oneTBB). In mold 3.0, Ueyama eradicated these complexities by moving the entire 60,000-line codebase to Rust. By leveraging Rust's Rayon library for work-stealing data parallelism, mold achieves identical or superior multithreaded scalability without dragging in cumbersome external C++ runtime dependencies.

The engineering benefits of the Rust transition extend beyond safety guarantees. In comprehensive benchmarking, mold 3.0 demonstrated a 22% reduction in peak resident set size (RSS) memory consumption when linking large 10GB debug binaries. Rust's strict ownership semantics enabled more aggressive string interning and contiguous arena memory allocations, reducing cache-line misses across high-core-count processors like AMD Threadrippers.

In raw linking performance, mold 3.0 matches its previous C++ speed tier. Linking Chromium debug builds takes just 1.42 seconds on a 64-core host—compared to 18.2 seconds for GNU gold and 4.1 seconds for LLVM lld. Furthermore, transitioning from CMake to Cargo simplifies packaging and distribution, allowing developers to install mold directly via standard Rust toolchains.

Ueyama highlighted that rewriting the linker in Rust also eliminated an entire class of subtle concurrency bugs and null-pointer dereferences that plagued previous edge cases during symbol deduplication. With version 3.0 now tagged as the primary production release, mold cements its position as essential low-level systems infrastructure for modern software engineering.

The Three-Question Framework

What Happened?

Why It Matters

What Could Happen Next

Executive Key Takeaways

Primary Sources & Citations