On Tuesday, 30 April 2013, Mathias Gaunard wrote:
A lot of audio processing happens in the frequency domain.
How does your library address this?
Yes handling domains other than the time domain is very important and I will think more about this and make sure it is supported by the library. I have not yet looked yet into how GIL handles different domains or even if it does. The library right now is just a concept and I am starting to flesh the idea out some more. I have been thinking about representation in different domains but have not yet decided which way to go. My plan is to work on various use cases of the library and compare different designs before I decide. Conceptually I like the idea of the domain being encoded in the data format type but I have not really explored the consequences of this yet. The extra dimension of Domain if we add it to the data format type will have no effect on the layout in memory unlike all other parameters in the type which is one reason against doing this, but I can see places where it is useful to know so the algorithms can be defined in the correct way. So it needs some more thought. Additionally the library will provide a few basic common algorithms and may be expanded over time. A fft will certainly be one of those algorithms, but primarily GIL and GAL initially deal with efficient use/access of variable memory layouts such that algorithms can be defined to operate on them generically and efficiently.