Re: AT+CIPSENDxxx performance

2015-04-28T11:50:07+08:00

I think the AT firmware only needs one AT+CIPSEND command that works for TCP and UDP, is always bufferred (however the new CIPSENDBUF works), nagle algorithm disabled, and can be termunated like CIPSENDEX. Although I think a better termination character sequence is $$$ (or better yet, user definable) to return to command mode.

Statistics: Posted by doughboy — Tue Apr 28, 2015 11:50 am

AT+CIPSENDxxx performance

2015-04-27T12:53:17+08:00

I did a load test using the different AT+CIPSEND commands, and this is what I found out.

v1.0.1b2 using any of the CIPSEND command, I get 1.8tps (555ms) sending 2.5kb of data (one 2k and second .5k).

using v1.0.1, CIPSEND and CIPSENDEX performance is reduced to half (transmit time DOUBLED!!!).
CIPSENDBUF is marginally faster. I get 1.86tps. (530ms to complete transmission).

So I don't see what the purpose of having so many different CIPSEND commands.
All you really need to one kind of CIPSEND command, and make it work really well.

Statistics: Posted by doughboy — Mon Apr 27, 2015 12:53 pm

ESP8266 Developer Zone

Re: AT+CIPSENDxxx performance

AT+CIPSENDxxx performance