Ok - here are some random thoughts. Here is an example of a program that people have complained about.
http://electronicsam.com/images/KandT/testing/steve.ngc This is made up of what looks like tangent arcs and line segments (I have not done the math) Here is the reason why I think it is. Mach runs tort.ngc in almost the same amount of time as linuxcnc (pathological non tangent arc/line - arc arc type paths) Mach though seems to run the above program at maximum (F3600) through the whole path. Linuxcnc currently runs the above program pretty hurky jurky. So I am thinking mach has a special case for tangent arcs. Could the new tp atleast check for tangent line-arcs, arc-arcs and then couldn't they be added similarly to line segments (not as complicated maths to figure out) that might satisfy 90% of the people out there... (maybe) Am I talking out of my a$$? (or - did we already discuss this?) thanks sam On 12/4/2013 1:08 PM, sam sokolik wrote: > BTW - there was an issue with arc acceleration constraint in master that > Chris fixed yesterday.. > > http://git.linuxcnc.org/gitweb?p=linuxcnc.git;a=commitdiff;h=bfb837e > > sam > > > > On 12/4/2013 1:01 PM, sam sokolik wrote: >> I ran this program >> >> http://electronicsam.com/images/KandT/testing/internet.ngc (named >> because I found it on the internet... - didn't even try to figure out >> what it was..) >> >> 533228 line program running G64P.005 (this program also crashes running >> strait G64 - Robert - do you still want me to debug? I see you think >> you may have a handle on it..) >> >> Old TP 2:37:42 >> New TP 1:38:49 >> >> Quite an improvement!! >> >> sam >> On 12/3/2013 3:20 PM, Robert Ellenberg wrote: >>> On Dec 3, 2013 2:42 PM, "sam sokolik" <sa...@empirescreen.com> wrote: >>>> I have not seen the blip.. (but surprisingly I may have found a blip in >>>> master...) >>>> Ok - spiral - wow. Running at almost 400ipm at the largest diameter. >>>> Damn impressive. >>>> >>>> I am a bit perplexed though.. If I run stellabee1.ngc ( >>>> http://electronicsam.com/images/KandT/testing/stellabee1.ngc ) with >>>> g64P.005 - it actually runs faster with parabolic blends (master). >>>> (40min vs 50 minutes about) I was expecting a 2 fold at least >>> improvement. >>> >>> The reason for that might be the arcs in the program. I reduced the >>> acceleration of arcs to fix the overage in arc-arc.ngc, but the tradeoff is >>> lower performance when there is a lot of acceleration and deceleration. >>> >>>> While the penguin - (metric so .1 is about .004") >>>> parabolic blend P.1=2:21 Strait G64=4:12 >>>> circular blend P.1=1:41 Strait G64=2:27 >>>> >>>> So - I don't know if that is correct - or there is some issue. Is the >>>> stellabee file just too wacky for the readahead to help? >>>> >>>> I could not run the stellabee file with strait g64 because it fails. I >>>> will try to get more info. (large jump in position/acc) >>>> Ok - I did a halscope with program line as part of it.. With stellabee >>>> and strait G64 - looks like line 32142. >>> Could you build the program with my debug flags enabled? It will spit out a >>> log that will be useful to diagnose what's happening. Run this command: >>> >>> make EXTRA_DEBUG = '-DTP_DEBUG -DTC_DEBUG' >>> >>> then run Linuxcnc like this: >>> >>> linuxcnc > test.log >>> >>> This works in simulation mode to dump a log of what's happening in tp. I'd >>> like to see if these spikes are visible in the planned motion. >>> >>> Thanks! >>> 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 >>> Emc-developers@lists.sourceforge.net >>> 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 >> Emc-developers@lists.sourceforge.net >> 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 > Emc-developers@lists.sourceforge.net > 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 Emc-developers@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-developers
