Back: 1.2. Where is the parallelism in sample generation?

Up: 1. The Problem

1.3. Evaluation Criteria

• How much useful parallelism does this model facilitate? Regardless of any other advantages, a parallel programming model must facilitate parallel computation of samples. Furthermore, this parallelism must be useful, meaning that a) pr ocessors should compute values that will be used, rather than scheduling unnecessary tasks for the sake of increased utilization, and b) the parallelism provided by the model should reasonably match the number of processors on the target architecture.

• Is there a natural mapping between the model and the problem structure? Sample generation is almost universally understood and explained in terms of instruments, scores, and unit generators. A programming model which accommodates this structure promotes elegant implementations and ease of explanation. Natural does not imply familiar; a programming model which seems alien at first may, upon reflection, match the problem structure in a novel way.

• Can this model also accommodate real-time sample generation? Synthesis algorithms, like other algorithms, will constantly grow more complicated, requiring more resources than available. Despite this trend, many sample generation tasks can now be performed in real-time. Some programming models factor the problem in a way that precludes real-time computations.