NumRS2 0.3.1 Released — Pure Rust NumPy Replacement with 222k+ SLoC

The numerical computing foundation of the COOLJAPAN scientific computing ecosystem just leveled up again.

Today we released NumRS2 0.3.1 — a complete, production-grade pure Rust numerical computing library that serves as a high-performance, memory-safe alternative to NumPy.

No C. No Fortran. No system BLAS/LAPACK.
No Python interpreter overhead. No FFI.
Just clean, blazing-fast N-dimensional arrays and mathematical operations that compile to a single static binary (or WASM) and run everywhere — from laptops to browsers to edge devices to cloud clusters.

Why NumRS2 0.3.1 is a game changer

For decades, high-performance numerical computing in the Python world meant depending on NumPy (built on C/Fortran) with all its system dependencies and safety trade-offs.

These tools are powerful but suffer from:

• Complex installation and platform-specific binaries
• Memory unsafety and data races in native extensions
• GIL bottlenecks and Python loop overhead
• Poor WASM/embedded/no_std support
• Difficulty achieving reproducible, production-grade performance

NumRS2 0.3.1 ends all of that.

It delivers near or superior performance to OpenBLAS while being 100% memory-safe and portable.
Notable results:

• 128+ SIMD-vectorized functions (AVX2 / AVX-512 / ARM NEON)
• 80–172% of OpenBLAS performance via OxiBLAS
• 1,143× scheduler speedup in enhanced parallel execution
• 4,704+ comprehensive tests with property-based and reference validation

Technical Deep Dive: How We Built a Production-Grade NumPy Replacement in Pure Rust

NumRS2 is built directly on the SciRS2 ecosystem (v0.3.4) and follows a trait-based, extensible architecture:

1. Core Layer
   N-dimensional arrays with cache-friendly layout, broadcasting, fancy indexing, boolean masking, and expression templates for lazy evaluation + operation fusion.

2. Linear Algebra & Sparse
   Full integration with OxiBLAS (pure Rust BLAS/LAPACK), matrix decompositions (SVD, QR, LU, Cholesky), sparse formats (COO/CSR/CSC/DIA), and iterative solvers (CG, GMRES, BiCGSTAB).

3. Advanced Numerics
   Optimization (BFGS, L-BFGS, Trust Region, Nelder-Mead, Levenberg-Marquardt), root-finding (Brent, Newton-Raphson, Halley), automatic differentiation (forward/reverse + higher-order), FFT (1D/2D/real), polynomial interpolation.

4. Hardware Acceleration & Interop
   Automatic SIMD dispatch via SciRS2-Core, GPU acceleration (wgpu: Vulkan/Metal/DX12/WebGPU), parallel execution with work-stealing scheduler, Apache Arrow + Feather + PyO3 Python bindings.

Key Rust advantages:

• 100% Pure Rust (zero external C/Fortran dependencies)
• 222,275+ lines of production code with 2,709+ fully implemented public API items
• Zero stubs, zero Clippy warnings, fail0 enforced
• Memory-efficient chunking, arena/pool/NUMA-aware allocators
• SharedArray with O(1) cloning and SharedExpr for lifetime-free lazy evaluation

What’s inside 0.3.1 (released March 21)

• Stability improvements and full Clippy0 cleanup (Color trait ambiguity, counter loops, manual checked division fixed)
• Enhanced scheduler with 1,143× speedup and deadlock prevention
• Expanded statistical distributions via SciRS2 (noncentral Chi-square, noncentral F, Von Mises, Maxwell-Boltzmann, truncated normal, multivariate normal with rotation)
• Production readiness confirmed with 4,704+ passing tests and comprehensive benchmarks
• 222,275+ lines of pure Rust — zero warnings, modular and extensible

This is the foundation

NumRS2 is now the official numerical computing backend for the entire COOLJAPAN stack (total ecosystem: 21M+ SLoC Rust, 597 crates, 40+ production-grade libraries):

• SciRS2 — all core scientific computing primitives
• OxiBLAS — linear algebra acceleration
• OptiRS — training and optimization loops
• OxiMedia / VoiRS — signal and audio processing pipelines
• OxiGDAL — geospatial numerical analysis
• Spintronics / OxiHuman — physics and biomechanical simulations
• ToRSh / OxiRAG — high-throughput tensor and RAG operations
• Future integration with OxiLean for formally verified numerics

Repository: https://github.com/cool-japan/numrs

Star the repo if you want high-performance numerical computing without NumPy’s native dependencies or Python overhead.

The era of “just pip install numpy” with all its C/Fortran baggage is coming to an end.

Pure Rust numerical computing is here — fast, safe, and sovereign.

— KitaSan at COOLJAPAN OÜ March 21, 2026