I think I implemented an uglier version of this fix independently in my code. Thanks for pointing this out, I'll apply the patch and see if I can simplify anything.
On Mon, Dec 16, 2013 at 12:13 PM, sam sokolik <sa...@empirescreen.com>wrote: > Ah - that makes sense. > > As far as the acceleration constraint violation below - I think you > also still have this problem that was fixed recently. > > > http://git.linuxcnc.org/gitweb?p=linuxcnc.git;a=commit;h=bfb837ef97113a311b9e431855a7390ddcc34c80 > > sam > > > On 12/16/2013 10:56 AM, Robert Ellenberg wrote: > > That's by design, though it's not optimal. I abort optimization if a > > segment is already in progress , to prevent stepping on tp 's toes. Also, > > the optimization doesn't run on the last segment in the queue, because > that > > segment has the potential to be shortened by a future blend arc. > > > > Both of these problems are fixable. I was looking into how to detect "end > > if file" in the code, and it seems that the interp_waiting state > > corresponds to the interpreter running out of valid moves (rather than > just > > having a full queue). If this is the case, then a simple check should be > > enough to flag the last move. Similarly, I could make blend arc creation > > work on the unchanged portion of a segment in progress. > > > > Rob > > On Dec 16, 2013 11:34 AM, "sam sokolik" <sa...@empirescreen.com> wrote: > > > >> This program (which should be tangent arcs) seems to slow down to a stop > >> at the beginning of the first and last arc. > >> > >> % > >> (1" square with rounded corners) > >> G90 G54 G20 > >> G64 > >> G0 X0 Y.25 Z0 > >> G1 Y.75 F50 > >> G2 X.25 Y1 I.25 > >> G1 X.75 > >> G2 X1 Y.75 J-.25 > >> G1 Y.25 > >> G2 X.75 Y0 I-.25 > >> G1 X.25 > >> G2 X0 Y.25 J.25 > >> M30 > >> > >> > >> > >> On 12/15/2013 9:40 PM, sam sokolik wrote: > >>> Arcspiral.ngc now runs at a peak of 400+ipm - only switches to > parabolic > >>> blends at the very end of the program (probably where the arcs are not > >>> 'tangent' enough). Awesome! > >>> > >>> steve.ngc runs in 7 seconds and seems to slow down down in 2 spots > >>> (switches to parabolic blends..) Awesome! :) > >>> > >>> stellabee1 runs a lot faster but.. When running G64P.005 - I get a > >>> overage around line 44545. > >>> > >>> Here is a short example that shows the errror - if you run the program > >>> farther - it will have a hand full more of these. says line 23 > >>> > >>> http://imagebin.org/282452 > >>> > >>> G20 > >>> G0 Z0.0120 > >>> G17 G40 > >>> G80 G90 G94 > >>> S1000 M3 > >>> (G04 P3) > >>> G64P.005 > >>> F999.0000 > >>> G18 > >>> F999.0000 > >>> G0 X3.0560 Z-0.1065 > >>> X3.0660 Z-0.0958 > >>> G2 X3.0800 Y3.3700 Z-0.0853 I-0.0126 K0.0314 > >>> G1 X3.0940 Z-0.0824 > >>> X3.1040 Z-0.0889 > >>> G3 X3.1160 Y3.3700 Z-0.0814 I0.0122 K-0.0061 > >>> G1 X3.1220 Z-0.0830 > >>> X3.1260 Z-0.0824 > >>> X3.1340 Z-0.0735 > >>> X3.1380 Z-0.0745 > >>> X3.1420 Z-0.0791 > >>> X3.1500 Z-0.0941 > >>> X3.1600 Z-0.1020 > >>> X3.1660 Z-0.0990 > >>> X3.1740 Z-0.1000 > >>> G3 X3.1880 Y3.3700 Z-0.0922 I0.0188 K-0.0172 > >>> G3 X3.1960 Y3.3700 Z-0.0736 I0.0697 K-0.0189 > >>> G1 X3.1980 Z-0.0706 > >>> M2 > >>> > >>> Great work!! > >>> sam > >>> > >>> > >>> > >>> > >>> > >>> On 12/15/2013 03:37 PM, Robert Ellenberg wrote: > >>>> I think I fixed the issue in the latest commit. It turned out that I > >> wasn't > >>>> checking the type of segment properly in an error check, so it would > >> flag > >>>> some arcs as not tangent even when they were. Along with some other > >> fixes > >>>> to get acceleration right, it should be running at nearly full speed > >> now. > >>>> Since my last push, I also added a few optimization fixes that should > >> run > >>>> code like stellabee a bit faster. In particular transitions from > >> tangent to > >>>> parabolic blending in some cases forced two unnecessary exact stops, > >> which > >>>> caused more ripple and slowdown than would otherwise be expected. > >>>> > >>>> I also ran the 3D chips code on my Sherline, off of an Atom CPU. I > >> didn't > >>>> see any RT interruptions even with the base thread running. My plan > for > >>>> testing next is to set up a version of the test config with a base > >> thread, > >>>> and spoof an e-stop. This way, the parallel port can be disconnected > to > >> do > >>>> build tests. > >>>> > >>>> -Rob > >>>> > >>>> > >>>> On Fri, Dec 13, 2013 at 2:44 PM, sam sokolik <sa...@empirescreen.com> > >> wrote: > >>>>> I forgot to say - I just pulled your latest changes before testing... > >>>>> > >>>>> sam > >>>>> On 12/13/2013 1:37 PM, Robert Ellenberg wrote: > >>>>>> Can you send a link to that G code? I'd like to run it myself and > see > >> if > >>>>> I > >>>>>> can pinpoint the slowdown. I've identified and fixed a bunch of > little > >>>>>> slowdowns over the last week, so the problem may already be solved. > >>>>>> > >>>>>> Thanks! > >>>>>> Rob > >>>>>> > >>>>>> > >>>>>> On Fri, Dec 13, 2013 at 2:00 PM, sam sokolik < > sa...@empirescreen.com> > >>>>> wrote: > >>>>>>> Couple of things. > >>>>>>> > >>>>>>> arc-arc blends are slower than current TP... current tp does > >> arcspiral > >>>>>>> at a peak of 100ipm while the new TP does it at about 70. > >>>>>>> So current TP http://imagebin.org/282155 1:19 minutes > >>>>>>> New TP http://imagebin.org/282156 1:51 (you can see it does > >> parabolic > >>>>>>> blends) > >>>>>>> > >>>>>>> The neat thing is.. If you do a G64P.003 the naive cam detector > >>>>>>> converts the short arcs to line segments and you get... > >>>>>>> > >>>>>>> http://imagebin.org/282156 21 seconds > >>>>>>> > >>>>>>> that peaks at almost 400ipm > >>>>>>> > >>>>>>> sam > >>>>>>> > >>>>>>> > >>>>>>> > >>>>>>> > >>>>>>> > >>>>>>> > >>>>>>> > >>>>>>> > >>>>>>> On 12/12/2013 2:49 PM, Robert Ellenberg wrote: > >>>>>>>> I agree with Sam that we're ready for testing from other users. > >>>>> However, > >>>>>>>> there are a few features that we may want to add before a merged > >> into a > >>>>>>>> main build. One limitation now is that there are a bunch of > >>>>> configuration > >>>>>>>> variables that are hard coded: > >>>>>>>> > >>>>>>>> - Optimization depth (currently set to ~40 segments) > >>>>>>>> - Enable / disable arc blend flag > >>>>>>>> - Max feed override (for sizing blend arcs, assumed to be > >> 200%) > >>>>>>>> Given these extra settings, I'm curious what the best way is get > >> them > >>>>>>> from > >>>>>>>> an INI file to the trajectory planner. For the experts, how are > INI > >>>>>>>> settings propagated into motion currently? Is there an easy way to > >> tack > >>>>>>> on > >>>>>>>> a few more config options? > >>>>>>>> > >>>>>>>> Thanks, > >>>>>>>> Rob > >>>>>>>> > >> > ------------------------------------------------------------------------------ > >>>>>>>> Rapidly troubleshoot problems before they affect your business. > >> Most IT > >>>>>>>> organizations don't have a clear picture of how application > >> performance > >>>>>>>> affects their revenue. With AppDynamics, you get 100% visibility > >> into > >>>>>>> your > >>>>>>>> Java,.NET, & PHP application. Start your 15-day FREE TRIAL of > >>>>>>> AppDynamics Pro! > >>>>>>> > >> > http://pubads.g.doubleclick.net/gampad/clk?id=84349831&iu=/4140/ostg.clktrk > >>>>>>>> _______________________________________________ > >>>>>>>> Emc-developers mailing list > >>>>>>>> Emc-developers@lists.sourceforge.net > >>>>>>>> https://lists.sourceforge.net/lists/listinfo/emc-developers > >>>>>>>> > >>>>>>>> > >> > ------------------------------------------------------------------------------ > >>>>>>> Rapidly troubleshoot problems before they affect your business. > Most > >> IT > >>>>>>> organizations don't have a clear picture of how application > >> performance > >>>>>>> affects their revenue. With AppDynamics, you get 100% visibility > into > >>>>> your > >>>>>>> Java,.NET, & PHP application. Start your 15-day FREE TRIAL of > >>>>> AppDynamics > >>>>>>> Pro! > >>>>>>> > >> > http://pubads.g.doubleclick.net/gampad/clk?id=84349831&iu=/4140/ostg.clktrk > >>>>>>> _______________________________________________ > >>>>>>> Emc-developers mailing list > >>>>>>> Emc-developers@lists.sourceforge.net > >>>>>>> https://lists.sourceforge.net/lists/listinfo/emc-developers > >>>>>>> > >> > ------------------------------------------------------------------------------ > >>>>>> Rapidly troubleshoot problems before they affect your business. Most > >> IT > >>>>>> organizations don't have a clear picture of how application > >> performance > >>>>>> affects their revenue. With AppDynamics, you get 100% visibility > into > >>>>> your > >>>>>> Java,.NET, & PHP application. Start your 15-day FREE TRIAL of > >>>>> AppDynamics Pro! > >>>>> > >> > http://pubads.g.doubleclick.net/gampad/clk?id=84349831&iu=/4140/ostg.clktrk > >>>>>> _______________________________________________ > >>>>>> Emc-developers mailing list > >>>>>> Emc-developers@lists.sourceforge.net > >>>>>> https://lists.sourceforge.net/lists/listinfo/emc-developers > >>>>>> > >>>>>> > >>>>> > >> > ------------------------------------------------------------------------------ > >>>>> Rapidly troubleshoot problems before they affect your business. Most > IT > >>>>> organizations don't have a clear picture of how application > performance > >>>>> affects their revenue. With AppDynamics, you get 100% visibility into > >> your > >>>>> Java,.NET, & PHP application. Start your 15-day FREE TRIAL of > >> AppDynamics > >>>>> Pro! > >>>>> > >> > http://pubads.g.doubleclick.net/gampad/clk?id=84349831&iu=/4140/ostg.clktrk > >>>>> _______________________________________________ > >>>>> Emc-developers mailing list > >>>>> Emc-developers@lists.sourceforge.net > >>>>> https://lists.sourceforge.net/lists/listinfo/emc-developers > >>>>> > >> > ------------------------------------------------------------------------------ > >>>> Rapidly troubleshoot problems before they affect your business. Most > IT > >>>> organizations don't have a clear picture of how application > performance > >>>> affects their revenue. With AppDynamics, you get 100% visibility into > >> your > >>>> Java,.NET, & PHP application. Start your 15-day FREE TRIAL of > >> AppDynamics Pro! > >> > http://pubads.g.doubleclick.net/gampad/clk?id=84349831&iu=/4140/ostg.clktrk > >>>> _______________________________________________ > >>>> Emc-developers mailing list > >>>> Emc-developers@lists.sourceforge.net > >>>> https://lists.sourceforge.net/lists/listinfo/emc-developers > >>>> > >>> > >> > ------------------------------------------------------------------------------ > >>> Rapidly troubleshoot problems before they affect your business. Most IT > >>> organizations don't have a clear picture of how application performance > >>> affects their revenue. With AppDynamics, you get 100% visibility into > >> your > >>> Java,.NET, & PHP application. Start your 15-day FREE TRIAL of > >> AppDynamics Pro! > >> > http://pubads.g.doubleclick.net/gampad/clk?id=84349831&iu=/4140/ostg.clktrk > >>> _______________________________________________ > >>> Emc-developers mailing list > >>> Emc-developers@lists.sourceforge.net > >>> https://lists.sourceforge.net/lists/listinfo/emc-developers > >>> > >>> > >> > >> > >> > ------------------------------------------------------------------------------ > >> Rapidly troubleshoot problems before they affect your business. Most IT > >> organizations don't have a clear picture of how application performance > >> affects their revenue. With AppDynamics, you get 100% visibility into > your > >> Java,.NET, & PHP application. Start your 15-day FREE TRIAL of > AppDynamics > >> Pro! > >> > http://pubads.g.doubleclick.net/gampad/clk?id=84349831&iu=/4140/ostg.clktrk > >> _______________________________________________ > >> Emc-developers mailing list > >> Emc-developers@lists.sourceforge.net > >> https://lists.sourceforge.net/lists/listinfo/emc-developers > >> > > > ------------------------------------------------------------------------------ > > Rapidly troubleshoot problems before they affect your business. Most IT > > organizations don't have a clear picture of how application performance > > affects their revenue. With AppDynamics, you get 100% visibility into > your > > Java,.NET, & PHP application. Start your 15-day FREE TRIAL of > AppDynamics Pro! > > > http://pubads.g.doubleclick.net/gampad/clk?id=84349831&iu=/4140/ostg.clktrk > > _______________________________________________ > > Emc-developers mailing list > > Emc-developers@lists.sourceforge.net > > https://lists.sourceforge.net/lists/listinfo/emc-developers > > > > > > > > ------------------------------------------------------------------------------ > Rapidly troubleshoot problems before they affect your business. Most IT > organizations don't have a clear picture of how application performance > affects their revenue. With AppDynamics, you get 100% visibility into your > Java,.NET, & PHP application. Start your 15-day FREE TRIAL of AppDynamics > Pro! > http://pubads.g.doubleclick.net/gampad/clk?id=84349831&iu=/4140/ostg.clktrk > _______________________________________________ > Emc-developers mailing list > Emc-developers@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/emc-developers > ------------------------------------------------------------------------------ Rapidly troubleshoot problems before they affect your business. Most IT organizations don't have a clear picture of how application performance affects their revenue. With AppDynamics, you get 100% visibility into your Java,.NET, & PHP application. Start your 15-day FREE TRIAL of AppDynamics Pro! http://pubads.g.doubleclick.net/gampad/clk?id=84349831&iu=/4140/ostg.clktrk _______________________________________________ Emc-developers mailing list Emc-developers@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-developers