Joel Jacobs wrote: > Ok, found something interesting. I left the isolcpus=1 in grub and > re-enabled hyper-threading and latency topped out around 11.2us with > acceptable performance. I thought it should be fixed so ran EMC2 and > after a few hours the error tripped again. Here is the details from > dmesg: > > > It looks to me that the 1058805 is the problem child here and it's not > an Unexpected realtime delay at all. Looks like an Unexpected > premature execution of the thread! Any idea what could cause this > Right, but the code that monitors execution time doesn't have specific limit values, it just looks at recent history and looks for ANY variance. If the process is normally not being retained in CPU cache, but on occasion it IS retained in cache, then it will execute faster that time. Otherwise, due to some configs doing more work on particular thread executions (such as when the trajectory planner is running at a sub-multiple of the servo thread) you can get these variances.
Anyway, I don't think a 111 us fluctuation should be a problem for a 700 us thread. A tool that monitors "cushion", ie how much non-real time is left over when the real time thread runs might be useful. As long as even a LITTLE time is left over on ever thread, then there is no problem. If even once in a blue moon there is zero time left, or an overrun, then it IS a problem. Jon ------------------------------------------------------------------------------ Enable your software for Intel(R) Active Management Technology to meet the growing manageability and security demands of your customers. Businesses are taking advantage of Intel(R) vPro (TM) technology - will your software be a part of the solution? Download the Intel(R) Manageability Checker today! http://p.sf.net/sfu/intel-dev2devmar _______________________________________________ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
