Trackpoint (Dualpoint Stick) too slow, confused

[Simon Thum]

Am 28.04.2010 13:20, schrieb Tobias Wolf:
> On Wed, 2010-04-28 at 11:44 +0200, Simon Thum wrote:
>> This intends to preserve precision. I'd bet any speed gains one may have
>> under another OS is driver magic.
> 
> I never used this machine under another OS and am not sure how
> responsive it is there. It’s just too slow, when it crosses the screen
> diagonal in more than a second at full force.
I think that's normal for an unaccelerated trackpoint.

> How do I find out where the kernel bits that used to be there went?
If they ever worked, probably in kernel [commit] lists.
> I guess that is where the real resolution should be set?
> 
> I have Linux 2.6.32.
It's possible there is a kernel-side solution, but just as well not.

> 
>> Since sysfs fails, probably acceleration is your friend. The velocity
>> code wasn't really written with trackpoints in mind but it should work
>> fine, maybe with tweaks applied.
>>
>> If the knob is really slow, I'd recommend a profile which ascends early
>> (like the smooth (soft-knee) limited or linear profile, 6 or 7). You may
>> also find it helpful to increase "velocity scaling", which adjusts the
>> profile responses. This can be done with xinput and xorg.conf[.d]
>>
>> It may take some time to find proper settings; they're documented here:
>> http://www.x.org/wiki/Development/Documentation/PointerAcceleration
> 
> I revisited that page. Maybe I shouldn’t have just skimmed over it the
> last time.
> 
>> In any case, I'd be interested in how it worked out.
> 
> The only profile that seems natural to me is 6 (linear) when I crank up
> the feedback. In the others I get odd »staircase« motion (see the
> animated gif under [1]).
That sure looks awkward! I hope at least you don't need to push hard in
those cases.

This is odd, you should get acceleration only if the pointer is moving
fast on its own. Maybe you need to decrease (not increase as I
suggested) the velocity scaling and try again.

BTW, what's 'crank up'? The settings are designed to be reasonable in
the 1..5 range, roughly.

> 
> It has completely different sensorimotor contingencies now, so I have to
> relearn to control the cursor using this.
> 
> Is there a difference between 30/10 and 3/1? Why is this a fraction and
No.
> not a float? How does it interact with the acceleration setting that
Because of a design decision done back when many different float formats
were competing. There was no safe way to handle it, so a integer
rational was used.
> Gnome exposes in the Mouse Properties? Should I reset that one to zero?
I don't know gnome well, but I guess it will just set it's values using
"xset m" at startup. So I'd say you should set them as you want them to
be. There are values representing default (-1 I think, see man page),
but no values for no modification.

You can always see the device's state using
   xinput get-feedbacks <device>
and
   xinput list-props <device>

"xset m" is modifying the feedbacks only, but on _all_ pointers. Maybe
you can convince gnome not to call it by uninstalling it, or setting
some option*. It's not deprecated AFAIK, but at least questionable for
the cases I know it's used for.

Cheers,

Simon


[*] preferably the latter, if possible
> 
> Thanks,
> Tobias
> 
> [1] i.imgur.com/v3Tud.gif
>