Raspberry Pi 4

Hi All,

It looks like the new raspberry pi 4 fix's a lot of problems I have read about for using one for linuxcnc. Mainly the addition of real time ethernet. A few questions...

1. Will the raspberry pi 4 now we a suitable choice for connecting a 7i76E or 7i92?

2. Mesa looks to have a 7C80 in development specifically for the raspberry pi, is that board going to be available anytime soon? Does this new raspberry pi effect the development?

1. It should but Ethernet performance is not known yet.

2. The RPI4 should just work with the 7C80/7C81 assuming its
SOC SPI interface is the same as the RPI3

The 7C81 _is_ available. the 7C80 still needs its inmux firmware
module support added to the hostmot2 driver

Made a quick and dirty test with 7i76E on the RPI4 4GB. It works, yes.
Can jog and run axis, didn't test more than that.
Got sveral messages that main loop took up to 0.25 seconds.
So I guess it doesn't work (so well) after all.

There are some things that can be tuned, but you should leave it running some gcode for at least an hour to see if it drops the connection to 7i76E.
If it does not drop it, there's still hope.

I seem not to have RT preempt patches, only the PREEMPT.

I'll go hunting for that.

Wait....

geneslinuxbox.net:6309/gene/Genes-os9-st...ebian-stretch-armhf/

But it is for the 3b, so might not work.
Thanks to Andy for the link.

Did not have time nor a microsd card to test that image yet.

If you start LinuxCnc from a terminal, do you see

Running POSIX realtime

Or does it say non realtime ?

I am rather sure it said POSIX non realtime. This one says while my "production" machine says
Not a very good test I am afraid. I'll start googling later in the evening.

Edit: There are actually brances with rt patches at github.com/raspberrypi/linux , called rpi-4.19.y-rt, I am using rpi-4.19.y. I will compile the rt version and give it a go.

Well, the route with "rpi-4.19.y-rt" didn't work. What works is to start with the "rpi-4.19.y" branch and apply the usual rt patches.
I now got this
and a latency_histogram that looks like this, after some random surfing on the net.


I will have to wait until tomorrow with actual test. I am pretty sure it will run (somewhat).

And to newbynobi, you have a line "import vte" in gmoccapy line 39, is that actually used? "buster" doesn't have a suitable package python-vte.

A servo thread only plot would be interesting