AXIS UI lags

Hello,

As you might know, linuxCNC slows AXIS down when it needs to: the display is less important than the actual NC.

However, I have made changes to AXISUI (more specificly to the Update() function of LivePlotter in axis.py).
One of those changes is when I press a button (which is a halpin), the machine moves in X+ while it is pressed.
Another change looks at a potentiometer (which is also a halpin), performs a non-linear conversion and outputs it to another hal pin.

With this, I can press a button and let the axis move with a feedrate override determined by the potentiometer.
But because AXIS sometimes slows down, the changes to the button and potentiometer can be delayed.
These delays can easily exceed 1 second.

When moving the axes, this is unacceptable as it might cause a crash.

Does anyone know how I can prevent the delay?
Or maybe how I can execute code parallel to axis?

I'm already planning on writing my own hal component to do the conversion for the potentiometer.
But I still have to figure out the details of the other changes I've made.

IMHO for serious machines, no time critical controls should be implemented in the GUI
Jog controls will have only 1 ms lags if implemented entirely in real time hal functions
with real time hardware

In other words. Instead of using a soft button on a UI screen. Use a real physical button on a control panel.

But I am not sure how much that may help, because I am not sure that the Halui interface pins you are likely using are real time.

Yes, I am assuming that the button and potentiometer signals go directly (or via real time hal comps) to motion (which is real time)

I have accepted it indeed isn't a good plan to implement these things in the GUI.
I have managed to make my first own hal component (hurrah!) for the potentiometer conversion.
For the jog controls I think I have gotten it down only using hal components and hal pins.
By the way, all the buttons were physical.
The slow performance/reaction time of the buttons only became apparent after I had to switch computers (because the first one had broken down).
I used the GUI because at first because at first I didn't know how to make hal components and also because I didn't know about the existence of some of hal pins.

The only thing I'm figuring out now has to do with twopass processing of the hal configuration files.
I have split my hal code up in different segments/files and each one uses loadrt functions for loading the hal components required in its file.

Like:
jog_buttons.hal: loadrt and2 names=hmi.jog.jog-criteria-accepted
on_off_buttons.hal: loadrt and2 names=hmi.on_off.and-off

There are some more. However, I get an error when starting linuxcnc:

It seems there is something wrong with one of my "loadrt and2" statements.
I have looked and at first there doesn't seem to be anything wrong.

I have tested it in a VM and it seems to be legit to use names such as "and2.0".

The looks suspicious -- as if there was an inadvertent blank
character in the name intended to be and2.2

Wow. Thanks for spotting that.
I can't imagine I glossed over that.

However, the "hal loadrt" line a couple of lines above the "orig_loadrt" line doesn't show the space.
And in my hal code the space isn't present either:

Is it possible the maximum number of and2 components with twopass is 17?
In my VM, when I use 17 components, everything loads fine.
But when I add an 18th component, it gives the error.

EDIT: I might have miscounted as a dmesg revealed: "names: can only take 16 arguments".
So it seems "and2 names=" is limited.
Any idea how I get around this while still using names?

SECOND EDIT:
I think I might've found something: forum.linuxcnc.org/forum/24-hal-componen...omponents-with-names

I had a similar issue. Although the latency test was ok, problem was solved with new PC hardware.
Having a real button or a soft button did not made any change.

My original post was here:
forum.linuxcnc.org/forum/38-general-linu...sh-large-delays-axis

My last post there says that with a new graphics card the problem was solved, but it is incorrect. The issue was only solved with a complete new pc.