USB activity causes network error

Ok sorted that out. Now how do I know that linuxcnc is running on the last two cpu's, or, how do I force it to run there?

forum.linuxcnc.org/38-general-linuxcnc-q...-irq-affinity#292969

Far more advice than I could give.

Once you have isolated cores, Real time applications (like Linuxcnc) will run on one of those cores. More particularly, the RT servo thread will run there. I think user space parts will run on non isolated cores where there is no need to be in time...

Thanks Rob. So it will make sense then to force the eth0 irq to the last core?

mariusl wrote:

Thanks Rob. So it will make sense then to force the eth0 irq to the last core?

Yes, be guided by PCW there. Thats what he says but a lot of RT stuff I have read indicates it should be on a seperate core. I have wondered about setting the affinity for eth0 to core 6 and the linuxcnc rtapi thread irq to core 7 (given isolcpus=6,7) The latency test is really irrelevant with mesa hardware and you need to focus on network latency.

rodw wrote:

mariusl wrote:

Thanks Rob. So it will make sense then to force the eth0 irq to the last core?
Yes, be guided by PCW there. Thats what he says but a lot of RT stuff I have read indicates it should be on a seperate core. I have wondered about setting the affinity for eth0 to core 6 and the linuxcnc rtapi thread irq to core 7 (given isolcpus=6,7) The latency test is really irrelevant with mesa hardware and you need to focus on network latency.



 

I am testing on a machine with 8 cores but the units that are giving me uphill all have only 4 cores. I will isolate 2,3 and put the eth0 irq on 2 to test. I will report back once I have tested it. Unfortunately all the machines that are problematic are a couple of hours away from me. Is there a way to test the network latency?