@cm said:
alex, it is too early for assumptions regarding *distributed processing* or similar. basically we encounter the same problems here for audio-rendering than with video-rendering - only arithmetical processes which are absolutely independant can be performed on different processors and therefore parallel. it might be possible, if an abstraction layer can be found which presents several processors as a single CPU to the process - the other way around is much easier and already common practice.
christian
Interesting. Given, as you say CM, the requirement for completely independent arithmetical processess, and the challenge of developing an abstraction layer to present multiple CPU's as a single entity, (I gather this is the problem with software like Logic, it can't use more than one CPU for the difficulty of sync)
then wouldn't a master/slave setup change that relationship and forego the need to find a way of 'mirroring' the algorithmic process?
i.e. Mir runs on the master until a percentage (75%/) of resource peaks, then the slave 'kicks in', indepndent of CPU alignment, and only synced with a MIR built in 'midi' type process?
Output, i understand, will ultimately come through the master, as we would need a 'single' source, (unless two machines are synced output to an external mixer, but that would defeat the purpose of the whole MIR construct anyway) but if, in the process of managing the 'throughput' from input source (i.e. 4 pc's generating audio signal) we use two machies instead of one, with the slave only as a means of using more RAM/CPU etc., then doesn't this effectively increase the chance of mixing and placing all input live, without bounce, frezze, or some other suspended stasis inducing method? I'm thinking about an orchestral soundscape here.
I don't know what you know about cars CM, but the analogy of the slave as a 'turbo charger' seems to match what i'm trying to say. As we want more power, the turbo kicks in, in harmony with the original 'engine'.
I'm wondering as well, if a slave 'node' for MIR would be worth thinking about.
I don't know enough about the mathematics of computer process to present this is a more intelligent fashion, so please forgive me the ignorance i show, in the lack of relevant terminology
Regards,
Alex.