DSPkern

DSPkern is an open-source digital signal processing kernel designed to provide a compact, real-time DSP runtime for audio and other streaming data. The project aims to deliver predictable low-latency performance across a range of platforms, from microcontrollers to desktop CPUs, while offering a modular framework for building DSP workflows.

Architecture and design: DSPkern centers on a processing graph composed of nodes that implement DSP operators such as filters, delays, saturators, and mixers. A lightweight scheduler drives block-based processing with deterministic timing, allowing operators to process fixed-size blocks in a contiguous pipeline. The kernel supports both fixed-point and floating-point arithmetic and includes a small memory management layer suitable for embedded environments.

Features: portable C/C++ API, SIMD-accelerated paths on supported architectures (ARM NEON, x86 SSE/AVX), pluggable operators, and a minimal runtime footprint. It offers a plug-in interface for adding custom DSP modules without changing kernel code. A cross-platform build system (CMake) and examples facilitate integration into various projects.

Platform compatibility and usage: Targeted at ARM Cortex-M and equivalent embedded CPUs as well as x86-64 desktops. It emphasizes deterministic latency and efficient memory use, with a focus on real-time constraints common in audio processing. Applications include embedded audio effects, synthesizers, and educational demonstrations of DSP pipelines.

Development and licensing: DSPkern is maintained by a community of contributors under a permissive license. The project uses an issue tracker, contribution guidelines, and code-review processes to maintain code quality. Documentation covers API references, tutorials, and migration notes.