I have a huge application that runs on EMX and 4.2 SDK. I have hand tune the code to have the best performance in terms of runtime. The code is composed of a while loop that checks button states, analog/digital inputs, many UARTs, can bus…
This loops was runing at 300ms per iteration on 4.2. Now on 4.3 it runs at 1s per iteration…
Do you observe the same on your application and platforms? Does anyone have solved that?
Here’s a comparison of practically the “same” code in terms of runtime:
RTC in,24634, micro seconds RTC in,24679, micro seconds
BPC in,5227, micro seconds BPC in,9852, micro seconds
OW detection in,350, micro seconds OW detection in,626, micro seconds
OW in,22174, micro seconds OW in,39539, micro seconds
BUZ in,290, micro seconds BUZ in,466, micro seconds
BUZEX in,189, micro seconds BUZEX in,547, micro seconds
LED in,189, micro seconds LED in,222, micro seconds
ANS in,7787, micro seconds ANS in,9772, micro seconds
DNS in,5138, micro seconds DNS in,6693, micro seconds
REL in,237, micro seconds REL in,224, micro seconds
NMEA in,1915, micro seconds NMEA in,3185, micro seconds
IMU in,4572, micro seconds IMU in,5370, micro seconds
buzoo in,189, micro seconds buzoo in,224, micro seconds
Computed data in: 21434, micro seconds Computed data in: 17722, micro seconds
Can sentence inputs in: 617, micro seconds Can sentence inputs in: 857, micro seconds
Can ana input,289, micro seconds Can ana input,558, micro seconds
Candig input,270, micro seconds Candig input,336, micro seconds
Evaluates instructions in : 136812, micro seconds Evaluates instructions in : 229199, micro seconds
powoo in,807, micro seconds powoo in,1087, micro seconds
rellevel in,400, micro seconds rellevel in,645, micro seconds
Evaluates buzoo in: 449, micro seconds Evaluates buzoo in: 729, micro seconds
Evaluates logs variable in: 1330, micro seconds Evaluates logs variable in: 2735, micro seconds
send data to can and CC in: 11955, micro seconds send data to can and CC in: 37665, micro seconds
Manage CC in: 20715, micro seconds Delay to manage each line sent by the CC: 1153, micro seconds
On left the code running on 4.2, on the right side, the code running on 4.3
@ leforban -
Don’t use USB power without attaching debug.
First run: 25 min 50s for the 4.3SDK vs 24min 51s for the 4.2 version.
The new sdk seems slower but not at the same proportion that I observe. However the bench is computation intensive while my application is more control intensive (even if there’s string analysis… and some computation tasks…)
I am comparing results using the debugger attached.