XU XUv0.13.1
中文

Blog

Xu 0.12.3: MIR optimization, hardware crypto, GPU passes, and window events

2026-11-13 · Xu v0.12.3

Xu 0.12.3 restores compiler inlining, tail-call recursion and loop folding for typed MIR, adds native instruction acceleration for hardware cryptography, and introduces standard-library GPU render/compute passes and window input events. Upgrade the compiler, standard library, and SDK tools together.

Compiler & Runtime Optimization

Restores typed-MIR inlining, integer tail recursion, and periodic loop folding. String repetition allocates its output buffer once. Hardware AES, ARM64 SHA-256, bit operations, and hex encoding use native instructions on supported architectures. Linux and macOS I/O waiters share each idle poll round.

Correctness & Safety Boundary

Fixes implicit shared borrows in bindings and assignments, restores native linking and resource cleanup, and bounds DEFLATE compressor output growth. Adds explicit-endian binary32/binary64 storage and bit conversions, and a WASI host for compiler tooling.

GPU & Windowing Foundations

Adds standard-library GPU render and compute passes, validated non-owning resource references, typed window surfaces, and per-window input event queues. Native GPU hardware support depends on the platform backend.

Downloads and Release Assets

Download the complete SDK, verify the SHA256SUMS checksums, and inspect the complete release notes on GitHub.