I see, are you running this sim config on RT builds as well? I have an atom-based PC and a P4-based PC that I can use, so I'll run the same test on both of them to see what happens. If you can run on a faster PC too, it would be a good sanity check. More data is definitely good at this point.
Thanks! Rob On Thu, Dec 5, 2013 at 10:03 PM, sam sokolik <[email protected]> wrote: > everything is here.. > > http://electronicsam.com/images/KandT/testing/circblendlatest/ > > It even has the cycle timer (last thing I added for timing programs...) > > ignore the stellabeeerror.ngc - that was showing and issue in master > that your tp didn't show > > I can have faster computers I can try it on. Do you have slower ones? :) > > sam > > > On 12/05/2013 08:45 PM, Robert Ellenberg wrote: > > Hi Sam, > > > > I ran a test on a Core2 2.6Ghz machine with an RT build and didn't see an > > error, which is a bit of a head scratcher. This machine is running ubuntu > > 10.04 from the recent 2.5.3 LinuxCNC CD, so it should be a standard > setup. > > It could be something about my config that isn't pushing it hard enough. > Do > > you mind sharing your config folder? I'd like to run the same test under > > identical conditions if possible. > > > > Thanks, > > Rob > > > > > > On Thu, Dec 5, 2013 at 7:39 PM, Robert Ellenberg <[email protected]> > wrote: > > > >> Thanks for isolating those bits of code! I'll run those snippets on my 2 > >> machines with RT builds tonight to see if the same failure occurs. > Logging > >> kindof works in RT builds but it dumps the text to a system log instead > of > >> stdout. Also, floats don't print correctly because of a limitation in > >> rtapi_print. Given the annoyances with RT builds, don't worry about > >> capturing debug logs for now. It's plenty helpful just to have a > snippet of > >> code that fails. > >> > >> I took a look at steve.ngc, and it looks like there are a bunch of > >> "almost" tangent arcs, so there are still some slowdowns here. The > latest > >> commit should handle arcs and lines that are exactly tangent. I could > cheat > >> and make the tolerance looser, but it will likely introduce small > >> acceleration spikes that will be hard to control. Still, this type of > >> benchmark is really useful to see what the new code can and cannot do. > >> > >> I'll post again soon when I figure out the cause of the > shoot-to-infinity > >> error. My guess at this point is that a function is implemented > differently > >> on the RT side that is not catching a NAN or divide by zero error. > >> > >> -Rob > >> > > > ------------------------------------------------------------------------------ > > Sponsored by Intel(R) XDK > > Develop, test and display web and hybrid apps with a single code base. > > Download it for free now! > > > http://pubads.g.doubleclick.net/gampad/clk?id=111408631&iu=/4140/ostg.clktrk > > _______________________________________________ > > Emc-developers mailing list > > [email protected] > > https://lists.sourceforge.net/lists/listinfo/emc-developers > > > > > > ------------------------------------------------------------------------------ > Sponsored by Intel(R) XDK > Develop, test and display web and hybrid apps with a single code base. > Download it for free now! > > http://pubads.g.doubleclick.net/gampad/clk?id=111408631&iu=/4140/ostg.clktrk > _______________________________________________ > Emc-developers mailing list > [email protected] > https://lists.sourceforge.net/lists/listinfo/emc-developers > ------------------------------------------------------------------------------ Sponsored by Intel(R) XDK Develop, test and display web and hybrid apps with a single code base. Download it for free now! http://pubads.g.doubleclick.net/gampad/clk?id=111408631&iu=/4140/ostg.clktrk _______________________________________________ Emc-developers mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/emc-developers
