LinuxCNC S-Curve Accelerations

Thank you, I'll keep that in mind.

I think there is much more to do and to finish because there are great jitters during running cloud of point... it can not approximate it and the computations last too much time(sometime mills and more) to finish ...

for signle G2 G3 it is working well ... 

I have talked with Michel, he said, he ends with lcnc and tp at all ... 

Now it is on us ...

my s-curve blending is comparable to the current TP, there is some loss, but it is unavoidable, but not in mills or more range.

if calculations are fitted in the servo thread time(which should be 250u as industrial standard) it should be nice to see or try something new ... for 7 phase generator is lcnc waiting many years and it is step to its adulthood

Michel told me to update his code but from my point of view If anybody on anyhardware will want to use it .. it will me close to imposible create a rt stuff for calculations ... I think it has to be splitted to rt and non rt part which will buffering each other for reaching as small as possible servo thread times

I am not expert in that at all...

github.com/grandixximo/linuxcnc

I have not build it yet, but it should work, test and let me know

Edit:
There were build errors, should all be fixed, built on debian 13 no issues, have a simple xyz test configuration, it works with more joints/axis no problem, xyz just quick and simple.

wow it works pretty well!

great job

there is some peaking but much less then I expected ... servo period is 250 000ns

I reduced BASE_PERIOD down to  50000 ns and peaking is weaker ... 

Testing at real setup isolated last core is with cold system latency 12.37us 

Wow! Thank you for testing, I actually never looked at the effect on the servo thread, I was more concentrated on velocity and acceleration changes, it is an interesting testing method, there maybe further possible optimizations, possibly some c++ tricks I'm not aware of, but in current state is the best I could come up with, and it is IMO respectably decent, nice to get some validation...

your job is amazing ... i am watching the scope curves without the breathing ....

conculations of polynomes of 5th grade probably take some power ... for more axis at same time can introduce jitter and at cloud of point it will be most visible ...

I am actually testing it at 1000us 500us and 250us which is most commom industrial scan time ... intel core duo, i5 and i7 ... every has different computation capacity but jittering are present during clouds 

I will take some more time and I will post outcome 

thank you for your job! is very interesting 

Thank you, by the way all coding was done by 杨阳 aka YangYang, most of the credits go to him, not sure what account he has here on the forum, I did the on field testing feedback, and I now ported to master.
Excited to see your testing results, if you show me your testing methods, I will try to reproduce on my end.

多谢杨阳

thanks a lot, Yang Yang!

Testing is at different architectures(because there is some differences at hardware level and in instruction set) with different G code types... It will influence calculations and produce jitter 

Most of us is running 3axis therefore i started there.. 

Then I will move to 6axis setup and do same for multiaxis movement

Then I will move to hardware benchtop setup and then to real iron... 

I am not in the shop and I can not share more or any code.. then later