On Wed, 2012-02-01 at 10:59 -0500, Tom Easterday wrote: > run the latency-test on the idle core AND run glxgears there (using > taskset to move it too), my latency is very bad.
That makes perfect sense: the video involved in glxgears locks out interrupts for protracted periods, so running it on the same core as the real-time handler should dramatically increase interrupt latency. It seems the AXIS UI is much better behaved, so running it on the real-time core doesn't affect latency all that much. Of course, running AXIS on an otherwise idle core will vastly improve overall performance, but that's not really the point of fencing off that core. I'd want to study the whole latency thing a lot more closely, with steppers whining away and the interpreter chewing through G-Code, before concluding that running *anything* other than real-time tasks on that core was a Good Idea. Now, if you had a four-core CPU, you could put the real-time stuff on one core, AXIS on another, have two left for everything else, and get wonderful performance... but, then, that's not a cheap Atom box. [grin] -- Ed http://softsolder.com ------------------------------------------------------------------------------ Keep Your Developer Skills Current with LearnDevNow! The most comprehensive online learning library for Microsoft developers is just $99.99! Visual Studio, SharePoint, SQL - plus HTML5, CSS3, MVC3, Metro Style Apps, more. Free future releases when you subscribe now! http://p.sf.net/sfu/learndevnow-d2d _______________________________________________ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
