[PD] Send/Receive topology design

[IOhannes m zmoelnig]

On 2017-01-05 11:18, Jérôme Abel wrote:
>> what are "synchro issues"?
> 
> Latency between messages.

there is none. (Pd messages use "ideal" timing).

> 
>> the former is easier to patch. the latter will perform slightly better.
> 
> That's the point.

what's the point?

> 
>> still end up using [s]/[r]
> 
> I preferred use send/receive objects to visually encapsulate
> functionnalities/objects (see parent/child classes), naming them to be
> more clear when sending messages, and avoid to connect/disconnect cords ...
> 
> But I understand your tips and it will help me to make choices, thanks.
> 
> The discussion could be also extend to dealing with a certain number of
> times the same abstraction. In other programming languages, we use
> arrays/vectors of objects, it is very clear and powerful. May be the new
> [clone] could be a good way to implement those features into my patches.
> 
> Do you know if the performances are the same (wireless and with the
> first inlet of clone) ?

[send] should perform slightly worse than a direct connection. (even one
via [clone]), as it has to consult the symbol table to find all receivers.

the difference is small though.
a quick test shows, that sending a [bang( message to 1000 objects via
[clone]s inlet takes ~0.01ms, whereas sending it via [send] takes ~0.2ms.

but i'm afraid that those performance considerations will mostly lead to
premature optimisation.

mdfas
IOhannes

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: OpenPGP digital signature
URL: <http://lists.puredata.info/pipermail/pd-list/attachments/20170105/095a3a8d/attachment.sig>