Re: [Emc-developers] improvement for rs274
On Sat, 6 Jun 2020 at 19:33, Reinhard wrote: > current version from master: > real0m20,118s > > improved variant: > real0m0,186s That seems like a suspiciously large performance increase. Do runtests still pass? Do you fancy making a pull-request on github? -- atp "A motorcycle is a bicycle with a pandemonium attachment and is designed for the especial use of mechanical geniuses, daredevils and lunatics." — George Fitch, Atlanta Constitution Newspaper, 1912 ___ Emc-developers mailing list Emc-developers@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-developers
Re: [Emc-developers] improvement for rs274
On Sonntag, 7. Juni 2020, 17:36:17 CEST andy pugh wrote: > That seems like a suspiciously large performance increase. > Do runtests still pass? What's the sense of runtests? My debug sessions showed, that stored G5x could never have been tested. Same is true for tool length compensation. Or if there exists tests for that, that tests have never been validated by manual calculations. The bug I found was used unconditionally on every run. And I'm quite sure, that the bug affects imperial systems too. linuxcnc supports the most ridiculous functionality but has no tests for the very basics? What's the sense of all that testing then? I of cause did only test the basics, that I use. The diffs I sent where captured by "diff -Naur" which means, you can apply the code changes by running "patch" automatically. But that diffs are the improvement only. To fix the bug, you should follow my description. Its just a one liner. cheers Reinhard ___ Emc-developers mailing list Emc-developers@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-developers
Re: [Emc-developers] improvement for rs274
On Mon, 8 Jun 2020 at 02:56, Reinhard wrote: > > What's the sense of runtests? > The tests are largely to check that fixed bugs have stayed fixed, many are based on specific problems previously encountered. It's a ringle command to run them and a 20 minute wait. It doesn't seem that much to ask. linuxcnc supports the most ridiculous functionality but has no tests for > the > very basics? > If you have identified gaps in the tests, then tests to fill those gaps would be welcome. > The diffs I sent where captured by "diff -Naur" which means, you can apply > the > code changes by running "patch" automatically. But that diffs are the > improvement only. > I suggested a pull request so as to document the change in the history and to correctly apply authorship. > To fix the bug, you should follow my description. Its just a one liner. Which bug are you talking about here? (The patch seems to be an enhancement rather than bug fix, so are you talking about the "G5x" item that you mentioned earlier?) -- atp "A motorcycle is a bicycle with a pandemonium attachment and is designed for the especial use of mechanical geniuses, daredevils and lunatics." — George Fitch, Atlanta Constitution Newspaper, 1912 ___ Emc-developers mailing list Emc-developers@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-developers
Re: [Emc-developers] improvement for rs274
Hi Andy, On Montag, 8. Juni 2020, 10:10:42 CEST andy pugh wrote: > Which bug are you talking about here? I suppose, that for tests you use interpreter in commandline-variant? or said call "bin/rs274" ? If not, then please tell me how to use the interpreter from commandline. But if so, store a G54 from axis and then run the interpreter to see, what it does. You don't need to execute a single gcode command. Startup sequence fails already. Or just use G43 with tool length > 0 of a tool read from tool.tbl Then validate the values by manual calculations. These tests will fail, no matter whether you have imperial setup or mm setup. Reinhard ___ Emc-developers mailing list Emc-developers@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-developers