Tag
A deep-dive blog post exploring how GCC lowers switch statements, showing its transformation into jump tables (including offset handling and a ~8-entries-per-label sparsity threshold) and falling back to decision trees of comparisons for sparser cases.
nesformer 是一个开源项目，通过编译器将 NES 的 6502 CPU、2KB RAM 与卡带 ROM 直接编译进一个未经训练的 Qwen3.5 模型权重中，让模型逐 token 预测 CPU 执行日志来运行《超级马力欧》等游戏，且帧输出与传统模拟器完全一致。
AMD released the ROCm HRX Loom SDK v0.1, a specialized compiler and IR stack that targets AMD CPUs, GPUs, and NPUs, with support for AMDGPU HSA, SPIR-V/Vulkan, and emerging x86/WebAssembly backends.
Nim 2.4.0 stable release ships small-string optimization strings, incremental compilation (--ic:on), a new cross-thread cycle collector (--mm:yrc), improved sum types, and support for cyclic module dependencies.
An in-depth look at the underlying implementation of Go's empty interface (any): the two-pointer structures of eface and iface, the meaning of the type word and data word, and how the compiler avoids unnecessary heap allocations through constants, zeroVal, staticuint64s, and other tricks.
Spritely Institute 发布了 Hoot 0.10.0，这是一个将 Scheme 编译到 WebAssembly 的 Guile 编译器后端及通用 WebAssembly 工具链。本次更新包含大量编译器、运行时和 Scheme 标准库（如 SRFI-13、SRFI-60、多个 R6RS 模块）的新增功能与修复，并发布了用于 Lisp Game Jam 的游戏开发模板。
Discusses Filian, the Fil-C toolchain, and building Fil-C binaries on Debian, marking progress toward memory-safe C adoption in mainstream Linux distributions.
Syl is a new systems programming language that encodes compile-time computation in a modal type system, using static/dynamic and erased/unerased mode axes to enable monomorphization and compile-time dependent types, inspired by two-level type theory and OxCaml.
A new rustc PR introduces adaptive sparse encoding for metadata tables, achieving an average 20.18% reduction in crate metadata size according to rustc-perf results.
Modular 宣布 MAX 的完整源代码现已公开，并推出简化的 MAX License，允许开发者免费使用、修改和发布商业产品，同时通过少量限制防止生态碎片化。此举是在 Qualcomm 完成收购 Modular 后，向统一异构计算软件平台迈出的关键一步。
Modular announced that its graph compiler, runtime, and hardware backends are now open-sourced alongside kernels, models, serving infrastructure, and the Mojo compiler and standard library, letting developers build and deploy AI across hardware platforms without vendor lock-in.
Modular 宣布将 MAX 随 Modular 26.7 源码开放，以打破 AI 对单一硬件厂商或芯片的锁定，推动 AI 算力民主化。
Crystal 1.22.0 正式发布，包含 251 项变更和 33 位贡献者的参与，主要亮点包括标准流模拟 SIGPIPE 行为、全面重构的 DWARF 解析以提升性能、所有目标平台支持原生事件循环、支持 LLVM 23/24，以及 aarch64-linux 晋升为 Tier 1 平台。
The Rust compiler team announced two new groups: a GPU target notification group staffed by experts from AMD, Nvidia, and Intel to help diagnose GPU-related issues, and a Linker working area focused on integrating alternative linkers such as Wild into the compiler.
NetWasm is an independent compiler, CoreLib, runtime and SDK that compiles C# and .NET CIL directly to WebAssembly, producing binaries smaller than size-optimized Rust (47,895 vs 53,635 bytes) while outperforming Mono AOT on GC benchmarks.
EDG's C++ front end went open source on September 30, 2026, with The C++ Alliance becoming its nonprofit home to provide fiscal sponsorship, technical continuity, and community-driven contributions under a three-track development model.
PyTorch's blog details the open-source MLIR-based compiler backend (KTIR to DataflowIR to SpyreCode) that translates PyTorch Inductor's decisions into executable dataflow pipelines for IBM's Spyre AI accelerator.
Cloudflare and VoidZero team up with Max Schwenk to incubate tsrs, a faithful Rust port of the TypeScript compiler within the Oxc toolchain, claiming up to 8x faster speeds and ~70% less memory than TypeScript 7.0 while leaving language design fully under the TypeScript team's control.
AWS 与 PyTorch 宣布推出 TorchNeuron——一个面向 AWS Trainium 芯片的原生 PyTorch 开源后端，支持 eager 执行、torch.compile 与 NKI 自定义内核，模型代码只需 `.to('neuron')` 即可运行，将于 10 月 20 日以 Apache 2.0 公开 Beta。
A blog post highlighting why 2026 is a landmark year for the Rust compiler, with the next-gen trait solver, the Polonius borrow checker, the parallel frontend nearing stabilization, and significant LLVM performance gains all converging at once.