GoKawiilGo 1.26 and 1.27 add experimental SIMD support, including a portable API
Go 1.26 introduced architecture-specific SIMD APIs for amd64, and Go 1.27 extends this with support for arm64 (NEON) and wasm. Go 1.27 also adds an experimental, platform-agnostic 'simd' package modeled loosely on the C++ Highway library, letting developers write vector code once that runs across AVX, AVX2, AVX512, NEON, and wasm SIMD instructions, with software emulation on unsupported platforms.
GoKawiil's interpretation of the reporting above, not reported fact.
Previously, tapping into SIMD required hand-written Go assembly, which was only practical for narrow, performance-critical code paths, leaving much software unable to exploit available CPU vector hardware. By offering a portable interface, Go's designers appear to be aiming to make SIMD-accelerated code more accessible to ordinary Go developers without sacrificing near-native performance, which could broaden its use in areas like cryptography, data processing, and AI workloads.
- Go 1.26 added an amd64-specific SIMD API; Go 1.27 extends architecture-specific support to arm64 (NEON) and wasm.
- Go 1.27 introduces an experimental cross-platform 'simd' package for write-once vector code across supported architectures.
- The new APIs aim to make SIMD acceleration accessible beyond hand-tuned assembly, following Go's earlier use of SIMD in its Green Tea garbage collector.
Source: go.dev — David Chase, 2026-09-25
Published there as: “Platform-Independent SIMD in Go”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.