Mesa hm2/hm2_7i96s.0: error finishing read

More
16 Oct 2022 13:27 - 16 Oct 2022 13:30 #254218 by PCW
How large is the latency spike?
What are your kernel command line parameters?
Are all power management options,speed switching/C states >C1 disabled in the BIOS?
(ms latencies are often related to power management/clock speed switching)
Last edit: 16 Oct 2022 13:30 by PCW.

Please Log in or Create an account to join the conversation.

More
16 Oct 2022 13:39 #254221 by bensttech
i will need to measure it again but its something in 1432312`s
i have tried combinations of "intel_idle.max_cstate=0 processor.max_cstate=0 idle=poll" and also isolcpus=2,3. it seems system load does change the how periodicaly it happens. power management is off in the bios.

Please Log in or Create an account to join the conversation.

More
16 Oct 2022 14:57 #254231 by PCW
That large of a latency spike is likely power management related

Are you sure all CPU frequency switching is disabled in the BIOS?

Please Log in or Create an account to join the conversation.

More
16 Oct 2022 22:04 #254271 by bensttech
Yep double and tripple checked. Running normally the read.time is between 370k and 520k when it blips it goes to around 1.6M

I just booted up my other box deb10 cnc 2.8 and 4.19.0-17-rt-amd64 and its rock solid at about 500k ish. it is different hardware though so not really a fair test tomorrow i will try the exact hardware

Please Log in or Create an account to join the conversation.

More
16 Oct 2022 22:13 #254276 by snowgoer540

model name      : Intel(R) Celeron(R) CPU  J1900  @ 1.99GHz


FWIW if you search the forum for J1900 you'll see people have mixed results with that processor. Some people swear by them, and others (myself included) couldn't get it to work reliably. Not saying that is your issue, but figured it was worth mentioning nonetheless.

Please Log in or Create an account to join the conversation.

More
16 Oct 2022 22:19 #254280 by bensttech
Thanks for the insight the strange thing is i have another machine same mainboard same FPGA card on 2.8 and its dead stable. i suspect its kernel driver related but it v frustrating

Please Log in or Create an account to join the conversation.

More
16 Oct 2022 22:20 #254281 by PCW
1.6M is close to 1 ms, you might try increasing the servo thread period to 1.5 or 2 ms

Please Log in or Create an account to join the conversation.

More
16 Oct 2022 22:24 #254282 by bensttech
That would be a work around but i would like to get to the bottom of it as i know i have another machine running same mb and fpga card without issues on 2.8.

Please Log in or Create an account to join the conversation.

More
16 Oct 2022 22:32 #254283 by bensttech
increasing the servo period to 1.5ms instantly made my worse case blip 2.4M, not exactly sure what that means.

Please Log in or Create an account to join the conversation.

More
16 Oct 2022 22:53 #254288 by PCW
it probably means that packets were being dropped (timed-out)
at a 1 ms servo thread rate (default timeout is 80% of servo period)

Please Log in or Create an account to join the conversation.

Moderators: PCWjmelson
Time to create page: 0.179 seconds
Powered by Kunena Forum