[PD] [PD-announce] pd 0.46-0test1 released

[Cyrille Henry]

Le 21/08/2014 18:42, Miller Puckette a écrit :
> On Thu, Aug 21, 2014 at 06:06:52PM +0200, Cyrille Henry wrote:
>>
>> yes, i prefer fudi to.
>> but on small hardware, like the one i was developing 12 years ago, OSC was faster.
>> not because of the reduced bandwidth, but because of the conversion between ascii character to float.
>> but that time is over now, and this conversion should be quicker even on "small" hardware.
>>
>> cheers
>> c
>>
>
> I agree - this is still a problem for pd~, which uses FUDI.  I've been thinking
> for years about making a binary form of FUDI:
>
> f <4 byte float>
> s <zero-terminated character string>
> ;
> ,
>
yes, pd~ communication is very heavy :
sending 8 different white noise to pd~ and sending them back to pd use more than 50% cpu on a decent computer.
sending silence is much better, since their are less data to convert, but it's far from perfect.
I think having no conversion at all will not be enough optimisation.
That's why there is now a share memory external.
for audio transfer : 1 synchronisation int is needed at every audio block and it add only 1 more block delay.
since it's very efficient, the curent implementation of fudi in pd~ is no longer a problem for me.

cheers
c



> One thing I keep wondering is how, in netsend/netreceive, to name this so that
> it won't get confused with raw binary.  Maybe it justneeds a name (BUDI?)
>
> cheers
> M
>
>
> _______________________________________________
> Pd-list at lists.iem.at mailing list
> UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list
>