Lazy Evaluation in Microsound Synthesis

The microsound synthesis framework in the LC computer music programing language integrates objects and library functions that can directly represent microsounds and related manipulations for microsound synthesis. Together with the mechanism that enables seamless collaboration with the unit-generator-based sound synthesis framework, such abstraction can help provide a simpler and terser programing model for various microsound synthesis techniques. However, while the microsound synthesis framework can achieve practical real-time sound synthesis performance in general, it was observed that temporal suspension in sound synthesis can occur, when a very large microsound object beyond microsound time-scale is manipulated, missing the deadline for real-time sound synthesis. In this paper, we describe our solution to this problem. By lazily evaluating microsound objects, computation is delayed until when the samples are actually needed (e.g., for the DAC output), and, when performing the computation, only the amount of samples required at the point is computed; thus, temporal suspension in real-time sound synthesis can be avoided by distributing the computational cost among the DSP cycles. Such a solution is beneficial to extend the application domains of the sound synthesis framework design beyond microsound synthesis towards more general sound synthesis techniques.