the outcome of these tests si that tcpserver is not suitable for my installation(s).
You might try latest iemnet: https://git.iem.at/pd/iemnet/-/jobs/17401/artifacts/download?file_type=archi...
[iemnet/tcpserver] didn't have immediate sending (TCP_NODELAY) enabled.
BTW, if you always broadcast to all clients, you can just as well use [netreceive -b]. But make sure to use Pd 0.51 which includes many networking bugfixes. Sending is not multithreaded and might block, but this is only relevant if you do audio.
Finally, if you want messages to get out as fast as possible, then TCP might be the wrong approach anyway and you might want to use UDP instead. Of course, UDP is unreliable, but for things like continous streams of data this is often not a problem.
Christof
On 23.06.2020 22:04, rolfm@dds.nl wrote:
it took me some time but here's what i found out:
(i could only work with the mrpeach library on deken which is the one from extended ).
the tests were with 3 clients: NodeMCU (Wifi ESP8266), Arduino Wifi Rev2 (Wifi NINA), old macbook (Pd tcpclient).
laptop Windows 10 Pd-50.0 : [metro 20] [message of 25 bytes( - SLIPENC - broadcast - tcpserver (iemnet or mrpeach).
for the Node and the Arduino there's not a real difference between mrpeach or iemnet. both clients report getting a burst every 240msec average, so mostly 12 messages in one go.
the pd-client on the macbook reports for iemnet 63-77% messages arriving in 20msec after the one before, the rest in bursts. the scores for mrpeach were varying more, from 75% to a low of 40%. (but i did not test for days on end, so it's nothing more than an educated impression).
i guess the difference between the Pd-client and the others is due to the specific Wifi libraries.
the outcome of these tests si that tcpserver is not suitable for my installation(s).
thanks everyone for your contributions in this thread.
rolf
Pd-list@lists.iem.at mailing list UNSUBSCRIBE and account-management -> https://lists.puredata.info/listinfo/pd-list