[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[sc-dev] SCe feasibility
Thanks everybody for your thoughts on this, SC embedded seemed to
have sparked at least some interesting conversation.
Despite all the issues mentioned in earlier emails, I am not willing
to give up that easily. Hence, here some thoughts and questions
regarding the feasibility of such a port:
* Fixed point conversion: there are some papers out there on the
automatic creation of DSP code converting floating point to fixed
point (e.g. http://r2d2.enssat.fr/bindocs/publications/
Menard02_CASES.pdf). Does anyone have experience with that? Defining
one precision for signals, one for timing and so forth might lead to
relatively simple conversion rules...
* The port of SC server itself (osc, nodes, timing etc) seems
manageable, what requires more effort is to get the algorithms in the
UGens done. Given the limited computational power of many target
devices, maybe a subset of UGens would be sufficient, so: What kind
of stuff would people think is the most important? Oscillators,
Filters (no Klank ;-) - One might be able to do quite a lot with all
sinus oscillators and a few filters...
* At computational power: most iPods for example use a dual 90MHz ARM
32bit processor. Any thoughts on how much one might be able to run on
such machines? Not sure how well threads can be distributed on two
cores, this might be another issue...