Can the OPI5 be Configured to Run LCNC?

I'll have a look at it (hopefully) soon when the mainline kernel is ready for the rk3588(s). At the moment it isn't and the Rockchip kernel is a mess. I've been updating the gpu driver and some other things, that took me a day, it's selectively patched which I continued. Creating patches with the new build system doesn't seem to work yet, for that I use sometimes the old build system to do a lot of different files at different locations. 

For now the patches will add the dtb for the opi5 and update gpu/hdmi/usb drivers. I expect that Panfork will work with this kernel but need to test it first.
- Compiled kernel is with the WiFi drivers
- Config without (you could copy the config from the deb)
- I've placed the patches in the patch folder instead of userpatches because some patches has to be at the first place.

mega.nz/folder/ypZxEAra#6NI9tibbdtoE_mAf8gdvKQ

So I reflashed the sd card with the ubuntu images I managed to compile instead of trying to reinstall the rt kernel. I did in fact do a apt upgrade which changed the kernel. I applied the hold and ran a few latency tests, one with youtube playing and one without with 6 glxgears running for at least 1000 seconds. It looks like the preempt rt patch is doing its job as you can see with the screenshots. I will also play around with the debian image as well. But I think I have enough to start playing with my Mesa 7196s board and see if I can't get a few stepper motors moving.

Great you've got it finally running, but there are some strange spikes on the ends I'm not sure what impact that will have, but they shouldn't be there. It could be the WiFi driver or power saving settings? Or did the temp rise too high?

You could try to see what causes the interrupts (probably on core 7).
cat /proc/interrupts. The updated drivers might help too.

could be temps, I live in Los Angeles and it got a little toasty this weekend. Currently the ambient temp in my Apt is 82F/27.8C. One of the things I wanted to machine is a nice beefy heat sink for the the OPi5 as there doesn't seem to be a commercial option as far as I know. The OPi5 had been running for at least an hour before I ran those tests so that could be a factor.

Also not excluding the other possibilities. I could hook up the Ethernet cable and run the test again to see if it is wifi.

As far as power savings I only disabled the suspend mode and screen dimming as well. I had forgotten to disable those and it ended up going into a power save mid test. Like you said it wouldn't wake up from it. So I had to start the test over after a forced shutdown. On that note I noticed it will not reboot upon a restart. It gets gets stuck at the boot process with a black screen. I have to power down and power back on in order for it to boot.

Also for some reason my last post didn't have the other image of the test without youtube playing in the background. Here is the one without a web browser open.

It seems it doesn't downscale the cpu, even at 85C while I was reading that it's it max temp. It was hard to reach that temp with 2 cores isolated in a heated room 20C (rainy Netherlands), had to take off the heatsink and run cpu stress and a few glxgears, still no spikes. I actually don't know if it really will cause issues, especially with a mesa and still a fair low latency.

A good chance that the wifi adapter plays a role, if that's the case you could see if the latest kernel I'd uploaded will help, it has updated usb drivers. Perhaps it does fix the reboot issue too. The power adapter could also be the source. You did disable screen blank too?

btw the heatsink I've got, is not great, but it does something:
aliexpress.com/item/1005005142012365.html

The fan doesn't do much which still is necessary with full cpu use (without the 2 cores isolated) so probably gonna make a bracket for a bigger heat sink.

kanuboy13b wrote:

 I applied the hold and ran a few latency tests, one with youtube playing and one without with 6 glxgears running for at least 1000 seconds. It looks like the preempt rt patch is doing its job as you can see with the screenshots. .

Thats actually quite good. You are only using 200 bins (columns) If you run:
latency-histogram --nobase --sbins 1000

You get rid of the uneeded base thread and with 1000 bins, it wil llook better!

rodw wrote:

Thats actually quite good. You are only using 200 bins (columns) If you run:
latency-histogram --nobase --sbins 1000

I will definitely give this a try, as well as run the test again with an ethernet connection instead of a wifi usb dongle. I will also try ryoka's updated kernel to see if that will resolve anything. Also going to try a different Wifi Dongle, the one's I have are rather old and do not have 5G, and for some reason, my wifi setup at home is really laggy with 2.4Ghz.

Wifi can be a latency killer for sure.

Shame the OPi5 doesn't have Dual Ethernet. One for the Mesa card and one for internet. However, I figure its a non-issue, if I was really concerned with spikes from the wireless I imagine I could just disable wifi during cnc operations

There is a leaked picture of an OPi5 plus, it looks like that one has dual Ethernet. On the github page there are made some preparations for it.
jamesachambers.com/wp-content/uploads/20...3/OrangePi5Plus.webp

Turning WiFi off is an option indeed, you could make a short-key in the keyboard settings for that if it's necessary. The laggy WiFi on the 2.4ghz band could be due to a lot of things of course. Some experiences I've had: DECT phone nearby the router, multi broadcast device like a chromecast/Apple airplay and a router without IGMP Snooping Enabled, different channel could help, when it's busy on all channels instead of 20/40mhz it could help to set it on 20mhz (less noise).