And switch to a code-flow that's a bit more self-explanatory than the current
min/max combinations.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Jonas Ådahl <jadahl@gmail.com>
This makes it more obvious where we're using units/us and units/ms as input
variable and what the output is. Clutters up the code, but still better than
dealing with us/ms differently per function, and still better than carrying
all the 1000.0 multiplications/divisions manually.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Jonas Ådahl <jadahl@gmail.com>
The return value of a profile is a unitless factor, not a speed.
Same applies for s1/s2, these are factors, not speeds.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Jonas Ådahl <jadahl@gmail.com>
This reverts commit 8a6825f160.
Aside from introducing bugs, this doesn't really help with anything, it adds a
requirement to rename everything to make clear where we're using µs and where
we're using ms and that just clutters up the code.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Acked-by: Jonas Ådahl <jadahl@gmail.com>
The new values were in units/us and didn't make the switch back to ms in
8a6825f160.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Jonas Ådahl <jadahl@gmail.com>
There is no need here to use µs since we're just handling speeds/thresholds,
not actual events where a ms granularity can be too high.
Moving back to ms lets us drop a bunch of zeroes that clutter up the code, and
since the acceleration functions are a bit magic anyway, having the various
1000.0 factors in there makes it even less obvious.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Jonas Ådahl <jadahl@gmail.com>
In order to provide higher precision event time stamps, change the
internal time measuring from milliseconds to microseconds.
Microseconds are chosen because it is the most fine grained time stamp
we can get from evdev.
The API is extended with high precision getters whenever the given
information is available.
Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Motion normalization does not work well for devices below the default 1000dpi
rate. A 400dpi mouse's minimum movement generates a 2.5 normalized motion,
causing it to skip pixels at low speeds even when unaccelerated.
Likewise, we don't want 1000dpi mice to be normalized to a 400dpi mouse, it
feels sluggish even at higher acceleration speeds.
Instead, add a custom acceleration method for lower-dpi mice. At low-speeds,
one device unit results in a one-pixel movement. Depending on the DPI factor,
the acceleration kicks in earlier and goes to higher acceleration so faster
movements with a low-dpi mouse feel approximately the same as the same
movement on a higher-dpi mouse.
https://bugzilla.redhat.com/show_bug.cgi?id=1231304
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Currently unused, but store the ratio of DPI:default DPI for later use.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Deceleration at low speeds is intended to enhance precision when moving the
pointer slowly. However, the adaptive deceleration we used was badly
calibrated, at slow-but-normal speeds the pointer became too slow to manouver.
We don't want to drop deceleration completely, the subpixel precision it
provides is useful. And it also helps those that can't move a 1000dpi mouse by
exactly one unit.
Make the adaptive deceleration steeper so it only kicks in at extremely slow
motions and defaults to 1 at anything resembling normal movement (i.e. pointer
moves like the physical device does).
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
To quote Bryce Harrington from [1]:
"MIT has released software under several slightly different licenses,
including the old 'X11 License' or 'MIT License'. Some code under this
license was in fact included in X.org's Xserver in the past. However,
X.org now prefers the MIT Expat License as the standard (which,
confusingly, is also referred to as the 'MIT License'). See
http://cgit.freedesktop.org/xorg/xserver/tree/COPYING
When Wayland started, it was Kristian Høgsberg's intent to license it
compatibly with X.org. "I wanted Wayland to be usable (license-wise)
whereever X was usable." But, the text of the older X11 License was
taken for Wayland, rather than X11's current standard. This patch
corrects this by swapping in the intended text."
libinput is a fork of weston and thus inherited the original license intent
and the license boilerplate itself.
See this thread on wayland-devel here for a discussion:
http://lists.freedesktop.org/archives/wayland-devel/2015-May/022301.html
[1] http://lists.freedesktop.org/archives/wayland-devel/2015-June/022552.html
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Acked-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
Acked-by: Jonas Ådahl <jadahl@gmail.com>
For really slow motions, the previous acceleration factor would go down to
effectively zero. So the slower the mouse motion was, the more it would be
slowed down which made the mouse at low speeds almost unusable.
Cap the minimum acceleration at 0.3 which provides a predictable slow motion
for the cursor when high precision is required.
New/old acceleration functions comparison:
^
| /
| /
ty| _________/
| / /
| / /
| / /
|/ / <----- new minimum accel factor
| /
|/___________________>
tx
i.e. the general shape is maintained, but it doesn't go to zero anymore. The
functions aren't parallel, the new shape is slightly flatter than the previous
one and they meet at the point where the functions flatten for the threshold
(tx/ty). ascii art has its limits...
https://bugzilla.redhat.com/show_bug.cgi?id=1227039
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Our motion filter takes the last couple of vectors to calculate speed,
provided the direction stays the same and it is within a certain timeout. It
does not take into account lifting the finger, so the velocity on the first
event is off.
Real-world impact is mainly on scrolling. Before commit 289e4675
filter: enforce minimum velocity
the first motion on a scroll was accelerated by a factor of 0 and swallowed.
After 289e4675 the motion was calculated based on the timeout and a fraction
of the expected effect. Now the first scroll motion is based on the real
finger motion since setting the finger down and thus feels a bit more
responsive.
It also makes a couple of test cases using litest_assert_scroll() work again
since the miniumum motion is now as expected.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
In the current code, a timeout or direction change on the first tracker will
result in a velocity of 0. Really slow movements will thus always be zero, and
the first event after a direction is swallowed.
Enforce a minimum velocity:
In the case of a timeout, assume the current velocity is that of
distance/timeout. In the case of a direction change, the velocity is simply
that since the last tracker.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
This timeout defines how far back in the events we search for velocity
calculations. For really slow movements, 300ms is not enough. It causes the
velocity to be 0 -> accel factor of 0 -> no movement.
As a result, really slow movement does not move the cursor.
Up the timeout to 1 second instead.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Let the caller set the various fields, here we just calculate stuff.
No functional changes.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
This code was largely lifted from the X server in
bb25b2ad29 but didn't take the copyright messages that
applied to that code.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Acked-by: Jonas Ådahl <jadahl@gmail.com>
Acked-by: Simon Thum <simon.thum@gmx.de>
Those touchpads presents an actual lower resolution that what is
advertised.
We see some jumps from the cursor due to the big steps in X and Y
when we are receiving data.
For instance, we receive:
E: 13.471932 0003 0000 16366 # EV_ABS / ABS_X 16366
E: 13.471932 0003 0001 9591 # EV_ABS / ABS_Y 9591
E: 13.471932 0000 0000 0000 # ------------ SYN_REPORT (0) ----------
E: 13.479924 0003 0000 16316 # EV_ABS / ABS_X 16316
E: 13.479924 0003 0001 9491 # EV_ABS / ABS_Y 9491
E: 13.479924 0000 0000 0000 # ------------ SYN_REPORT (0) ----------
E: 13.487939 0003 0000 16271 # EV_ABS / ABS_X 16271
E: 13.487939 0003 0001 9403 # EV_ABS / ABS_Y 9403
E: 13.487939 0000 0000 0000 # ------------ SYN_REPORT (0) ----------
-> jumps of ~50 in X in each report, and ~100 for Y.
Apply a factor to minimize those jumps at low speed, and try
keeping the same feeling as regular touchpads at high speed.
It still feels slower but it is usable at least
Signed-off-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Tested-by: Vasily Khoruzhick <anarsoul@gmail.com>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
The main purpose of this patch is to allow the user to actually slow
down pointer movement using libinput_device_config_accel_set_speed, this
is achieved by changing the max-accel setting from "2.0 - speed" to
"2.0 - speed * 1.5", resulting in a max-accel of 0.5 when the user configures
speed at -1.0, the other accel profile parameters are adjusted by the same
factor to keep the curve the same.
This means that the user can get the exact same behavior as before by
multiplying the old setting by 0.6667 (2/3), this also means that this
change not only allows the user to select a slower speed, but to keep
things balanced the same as before, also a higher speed.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Change vector_get_direction input to a normalized_coords type rather than
passing in a separate x,y pair, and rename it normalized_get_direction to
match. Since it now depends on the normalized_coords type which gets declared
in libinput-private.h also move it to libinput-private.h .
Note this commit also contains a functional change wrt the get_direction
usuage in the palm detection. The palm-detection code was calling get_direction
on non normalized coordinates, this commits changes the code to normalize
the coordinates first. This is the right thing to do as calling get_direction
on non normalized coordinates may result in a wrong direction getting returned
when the x and y resolution of the touchpad are not identical.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Add a normalized_length helper function and use this where applicable,
just a minor cleanup.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
If the delta is 0, the distance is the number of units (within this ms). Delta
1 means velocity across 2 ms, etc.
Bonus: this doesn't return infinite speed anymore if we get more than one
event per ms. This can happen on any device approaching 1000Hz poll rate, but
definitely got triggered by the test suite.
Actual effect was limited, since we cap out acceleration at max_accel we just
hit this earlier and it stayed there.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
This way the unaccelerated deltas returned by libinput are correct.
To maintain the current behavior we slow down the input speed by the magic
factor and likewise the accelerated output speed. This produces virtually the
same accelerated deltas as the previous code.
The magic factor is applied to the default denominator for guessing a
resolution based on the touchpad diagonal. We can't really get around this
without having a resolution from the touchpad; meanwhile this produces
virtually the same coordinates before/after.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Converting to integer before the sqrt calculation can cause loss of
motion at low speed.
Signed-off-by: Derek Foreman <derekf@osg.samsung.com>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
The acceleration curve consists of four parts, in ascii-art like this:
_____________
/
____/
/
/
where the x axis is the speed, y is the acceleration factor.
The first plateau is at the acceleration factor 1 (i.e. unaccelerated
movement), the second plateau is at the max acceleration factor. The threshold
in the code defines where and how long the plateau is.
This patch adjusts the curve based on a [-1, 1] range. For anything below 0,
the plateau is longer (i.e. accel kicks in at a higher speed), the second
incline is flatter (i.e. accel kicks in slower) and the max accel factor is
lower (i.e. maximum speed is slower). For anything above 0, the inverse is
true, acceleration kicks in earlier, harder and is faster in general. So the
default/min/max curves overlaid look something like this:
________ max
| _______ default
| / _____ min
_|_/_/
/
/
Note that there's a limit to what ascii art can do...
Note that there are additional tweaks we can introduce later, such as
decreaseing the unaccelerated speed of the device (i.e. lowering the first
plateau).
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
No functional changes, prep work for the config interface.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
To keep the implementation of a filter separate from the users of a filter.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
We ran a userstudy, evaluating three different accel methods. Detailed results are
available at:
http://www.who-t.net/publications/hutterer2014_libinput_ptraccel_study.pdf
We found that there was little difference between the method we had in
libinput 0.6 and this three-line function. Users didn't really notice a
difference, but measured data suggests that it has slight advantages in some
use-cases.
The method proposed here is the one labeled "linear" in the paper, its profile
looks roughly like this:
_____________
/
____/
/
/
where the x axis is the speed, y is the acceleration factor.
The first plateau is at the acceleration factor 1 (i.e. unaccelerated
movement), the second plateau is at the max acceleration factor. The threshold
in the code defines where and how long the plateau is.
Differences to the previous accel function:
- both inclines are linear rather than curved
- the second incline is less steep than the current method
From a maintainer's point-of-view, this function is significantly easier to
understand and manipulate than the previous one.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Makes it possible to use from the touchpad code.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
In an attempt to bring method into the madness, normalize the touchpad deltas
to those of a USB mouse with 400 dpi. This way the data we're dealing with in
the acceleration code is of a known quantity.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
This just moves a decimal point around, at the expense of making the approach
harder to understand. The only time the const acceleration matters is when
applied to the velocity but it only matters in relation to the threshold which
is a fixed number.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
velocity is in unit/ms, the threshold is in units/ms. Once we divide
velocity/threshold, we're not in units/ms anymore but have a unitless factor.
Use a separate variable to avoid confusion.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Instead of always going from an accel of 0.0 to 1.0 between a velocity of
0.0 and 1.0, make the lead-in length of the curve depend on the threshold
setting, using half of the threshold for the lead-in.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
There was an error in the value passed to the second calc_penumbral_gradient
call causing a jump in the acceleration curve. This commit fixes this.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Cleanup the code a bit, and make sure accel is at least 1.0 .
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
I doubt this does what we think it does. It doesn't soften the delta changes,
rather it introduces bumps in the smooth processing of the changes.
abs(delta) below 1.0 is untouched, and abs(delta) beyond 3 or 4 isn't
noticable much. But in the slow range around the 1/-1 mark there is a bump.
For example, if our last_delta is 1.0 and delta is 1.1, the "softened"
delta is set to 0.6. That is stored as last delta, so an input sequence of:
0.8, 0.9, 1.0, 1.1, 1.2, 1.0, 0.8, 1.1
results in "softened" deltas that don't match the input:
0.8, 0.9, 1.0, 0.6, 0.7, 1.0, 0.8, 0.6
A better approach at smoothing this out would be to calculate the softened as:
current = current ± diff(last, current) * 0.5
or even weighted towards the new delta
current = current ± diff(last, current) * 0.25
In tests, this makes little difference. Dropping this function altogether is
sufficient to make the pointer pointer behave slightly better at low speeds
though the increase is small enough to attribute to confirmation bias.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
For better consistency with filter_dispatch(). And move the things around to keep
the consumable API together.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
calculate_velocity() didn't skip pointer trackers far away in time when
calculating the initial velocity. This check was done later when
iterating the rest, so while at it, simplify the function by doing both
iterations in one single loop.
Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
