Part Number: AM6442
Hi, there is a problem when I test igh ethercat on AM6442 rt kernel.
1. The rt kernel is from ti-processor-sdk-linux-rt-am64xx-evm-08.01.00.39，the configuration CONFIG_PREEMPT_RT is opened
2. Test rt kernel with cyclictest
The real-time performance of CPU0、CPU1 is not very good.
Isolate CPU1，real-time performance of CPU1 is better than CPU0.
3. Igh ethercat test
I connect motor control module(hardware) to cpsw network port or pru network port.
Use the ethercat generic driver which is base on cpsw network port driver or pru network port driver.
I run a app testing Igh ethercat：first clock_nanosleep(1ms), then get and change motor state through ethercat api in each period.
This app will always calculate time consumed per period and record the min/max value，then print the result to uart console every 1s.
3.1 Do not Isolate CPU1，then run Igh ethercat test app
The time consumed of some periods is much larger than 1ms, whether using cpsw network port or pru network port.
3.2 Isolate CPU1，then run Igh ethercat test app on CPU1
The time consumed of most period is better，around 1.2ms.
But some period consume much larger than 1ms, with skipping unmatched datagram by ethercat.
becasue of isolating CPU1, the warning about skipping unmatched datagram by ethercat appear frequently.
This problem will appear whether use cpsw network port or pru network port to test igh ethercat.
Use usb ethernet port(AX88772C) to test
The time consumed of most period is better，around 1.2ms.
Even if isolate CPU1, the warning about skipping unmatched datagram by ethercat appear rarely.
If the cpsw network port driver or pru network port driver limit the real-time performance of the ethercat?
3.3 If there are some ethercat tests result on AM6442 rt kernel which can be provided?
4. How to improve real-time performance of AM6442 rt kernel
If there is another way to improve real-time performance of AM6442 rt kernel except isolating CPU1?
