http://bluebugs.github.io/tags/data-parallelism/Recent content in Data-Parallelism on Cedric BailHugoen-usThu, 19 Jun 2025 18:48:59 -0700http://bluebugs.github.io/blogs/go-data-parallelism/Thu, 19 Jun 2025 18:48:59 -0700http://bluebugs.github.io/blogs/go-data-parallelism/<h2 id="why-data-parallelism-matters-in-go">Why Data Parallelism Matters in Go</h2> <p>Go is a fast language, but it lacks easy ways to express data parallelism and does not provide direct access to low-level <strong>SIMD</strong> (Single Instruction Multiple Data) instructions. As a result, standard libraries like <a href="https://github.com/golang/go/issues/19636">base64</a>, <a href="https://github.com/golang/go/issues/68188">hex</a>, <a href="https://github.com/golang/go/issues/63347">utf8</a>, <a href="https://github.com/golang/go/issues/53178">json</a>, <a href="https://github.com/golang/go/issues/24499">jpeg</a>, and <a href="https://github.com/golang/go/issues/71255">map</a> are slower than expected. Other ecosystems are more likely to adopt specialized, high-performance libraries, which is why, in some cases, for example, Node.js can outperform Go.</p> <p>The solution to this bottleneck is for the Go compiler to generate SIMD instructions. There are three main approaches to enabling SIMD in programming languages:</p>