Latency test discrepancy after ethercat feature adding

[feralMaster]

Hello gentelmen,

@BsAtHome I am facing weird latency issue from moment of installation last RIP of linuxcnc 2.10.0 where is implemented the initf lcec.activate servo-thread...

This issue created real mess in the ethercat setup mainly noticeable during running beckhoff ax5yzz SoE which is pretty sensitive DC at latency at all.

The difference we noticed with the @grandixximo during testing between custom ec master from "community branch" and "grandixximo branch"...

We can replicate it if it will be necessary.. time of duration take with spare, its mainly for visualisation

Visualization at pictures

GRUB_CMDLINE_LINUX_DEFAULT="quiet isolcpus=domain,managed_irq,2-3 irqaffinity=0-1 nohz_full=2-3 rcu_nocbs=2-3 rcu_nocb_poll rcutree.kthread_prio=5 preempt=full workqueue.cpumask=1 processor.max_cstate=1 intel_idle.max_cstate=1 intel_pstate=active cpufreq.default_governor=performance ipv6.disable=1 i915.enable_dc=0 i915.enable_fbc=0 i915.enable_psr=0 mitigations=off nmi_watchdog=0 nosoftlockup tsc=reliable clocksource=tsc transparent_hugepage=never skew_tick=1 acpi_irq_nobalance irqpoll noirqbalance pcie_aspm=off module_blacklist=evbug,intel_cstate,intel_rapl"

[BsAtHome]

You probably want to talk to @grandixximo and @rodw-au who have been more into ethercat. I'm probably just a bystander here.

[grandixximo]

I asked him to post this for the latency-histogram, the new method seems to be getting wildly different readings

[grandixximo]

But now that I look again, even --legacy mode is getting weird reading on the latest master, what could be effecting the latency that much?