SIMDsynth

This is a SIMD-accelerated audio synthesizer written in C++ with the JUCE framework, using vectorized math to simulate polyphonic audio synthesis with filters, envelopes, LFOs, and sine and saw wavetable oscillators.

It's optimized for both x86-64 (SSE) and ARM64 (NEON) platforms, in order to highlight the SIMD techniques in a JUCE context.

Some rudimentary factory presets are included in the plugin - they will be created automatically on first execution of the plugin.

High-Level Overview

• Multiple synthesis formats (AU, VST3, Standalone)
• Up to 16-voice polyphony
• Multiple waveforms (sine, saw, square) using wavetables
• Sub-oscillator with keyboard tracking
• Unison feature with detune
• ADSR envelopes
• LFO modulation
• Filter per voice
• 4x oversampling to reduce aliasing
• Preset management system
• Includes a basic set of Factory Presets for testing purposes.

Technical Implementation:

• Built using the JUCE framework
• Uses SIMD (Single Instruction Multiple Data) optimization for efficient processing
• Supports both x86 (SSE/SSE2/SSE4.1) and ARM (NEON) architectures
• Implements wavetable synthesis with 2048-point tables
• Uses modern C++17 features
• Note: An initial attempt at integrating Tony Hardy-Bicks' DFM1 Filter is available in the "dfm1-filter-integration" development branch!

TODO:

• Move file operations to a background thread to prevent audio glitches.

• Add a modulation matrix for more flexible routing

• Implement additional LFO waveforms

• Add envelope curves/shapes

• Add filter types (currently has one filter type)

• Implement additional oscillator waveforms

• Add effects section (reverb, delay, etc.)

• Add MIDI learn functionality for parameters

• Implement undo/redo for parameter changes

• Add parameter smoothing for all controls

• Consider splitting the Voice struct into smaller components (Oscillator, Envelope, etc.)

• Move DSP-related code into separate classes

• Create a dedicated parameter management class

• Add more robust error handling for file operations

• Implement graceful fallbacks for missing presets

• Add parameter validation

• Add unit tests for DSP algorithms

• Implement automated testing for preset loading/saving

• Add performance benchmarks

• Use more C++17 features:

	// Instead of raw pointer management:
	std::optional<Filter> filter;
	std::variant<float, double> sampleType;

• Consider implementing the PIMPL idiom to reduce compilation dependencies and improve build times.
• Add detailed parameter documentation
• Include performance considerations and CPU usage guidelines
• Document the SIMD optimization strategy
• Add code comments explaining DSP algorithms

(c) 2025, seclorum