Re: [Machinekit] Re: RPI gipo help

[schoone...@gmail.com]

If you run a dummy base thread, but don't actually connect anything to it, it can improve latency with rt-preempt on multi-core processors. I suspect it just prevents cache hopping, by keeping all the code in one cache whilst the other processor deals with the base thread. Experiment using 'latency-test 50us 1ms' as opposed to just 'latency-test' which will start just a 1ms servo thread by default. Linuxcnc is still oriented towards software step generation, so always runs a base thread on a latency test unless specifically told otherwise. On 29/08/18 09:18, KL Chin wrote: Hi Tim, TQVM.. With latency-histogram —nobase, the latency look good under LinuxCNC.  Regards, KL Chin On Wednesday, 29 August 2018 01:02:30 UTC+8, Timothy March wrote: KL Chin I do believe the difference between latency test are LinuxCnc has both Base thread and Servo thread running. Machinekit just has Servo thread. With LinuxCnc you might want to try command line: latency-histogram —nobase Lantecy test then may be comparable Tim -- website: http://www.machinekit.io blog: http://blog.machinekit.io github: https://github.com/machinekit --- You received this message because you are subscribed to the Google Groups "Machinekit" group. To unsubscribe from this group and stop receiving emails from it, send an email to machinekit+unsubscr...@googlegroups.com. Visit this group at https://groups.google.com/group/machinekit. For more options, visit https://groups.google.com/d/optout. -- website: http://www.machinekit.io blog: http://blog.machinekit.io github: https://github.com/machinekit --- You received this message because you are subscribed to the Google Groups "Machinekit" group. To unsubscribe from this group and stop receiving emails from it, send an email to machinekit+unsubscr...@googlegroups.com. Visit this group at https://groups.google.com/group/machinekit. For more options, visit https://groups.google.com/d/optout.