Base64 Decoding with Ignorable Characters using SIMD instructions

The base64 format is ubiquitous: it is used in emails and Web documents to encode binary data as text. By convention, base64 text often includes ignorable characters such as spaces. We introduce the first base64-decoding algorithms with support for ignorable characters designed for single-instruction-multiple-data (SIMD) instructions. Our novel functions multiply the base64 decoding speed compared to traditional methods. Our algorithms have been implemented for different hardware architectures (ARM and x64). Our approach has been widely adopted in JavaScript runtimes and Web browsers.