OpenGL Matrices – the missing bits

While generally the available documentation on how the OpenGL matrices work is quite good, there are some missing bits. Although not necessary for your everyday rendering, they give one some insight on how rasterization in general and OpenGL in special works.

W coordinate after perspective divide

After conversion to normalized device coordinates(ndc) one might think each vertex looks like

$$ \vec{v}_{ndc} = \frac{1}{w} \begin{pmatrix} x \\ y \\ z \\ w \end{pmatrix} = \begin{pmatrix} \frac{x}{w} \\ \frac{y}{w} \\ \frac{z}{w} \\ 1 \end{pmatrix} $$

however it looks more like

$$ \vec{v}_{ndc} = \begin{pmatrix} \frac{x}{w} \\ \frac{y}{w} \\ \frac{z}{w} \\ \frac{1}{w} \end{pmatrix} $$

the $w$ coordinate is not divided by itself, but is inverted instead. This is done because the interpolation between vertices still needs to take place and for perspective correct interpolation one needs the camera space depth $z = -w_{cam}$.

\vec{v}_{\alpha} &=& \frac{(1-\alpha)\frac{\vec{v}_0}{-z_0}  + \alpha\frac{\vec{v}_1}{-z_1}}
{(1-\alpha)\frac{1}{-z_0} + \alpha \frac{1}{-z_1}} \\
&=& \frac{(1-\alpha)\vec{v}_0 w_{0_{ndc}} + \alpha\vec{v}_1 w_{1_{ndc}}}
{(1-\alpha) w_{0_{ndc}} + \alpha w_{1_{ndc}}}
\end{eqnarray} $$

instead of dividing by $-z$ we can multiply with $w_{ndc}$ as multiplication is faster than division.

Note that for brevity the given formula assumes a scanline based rasterizer as it interpolates only between two vertices. The general approach is to use barycentric coordinates to interpolate between all three vertices simultaneously.

Row major or column major

Even though even Wikipedia says OpenGL is column major, it is actually storage agnostic. However by default it interprets your 16 element array as:

$$ \begin{bmatrix}
m_0 & m_4& m_8 & m_{12}\\
m_1 & m_5& m_9  & m_{13}\\
m_2 & m_6& m_{10}  & m_{14}\\
m_3 & m_7& m_{11}  & m_{15}

Yet most OpenGL functions dealing with matrices offer a transpose parameter which you can use to specify the used order. For a comparison of storage orders see the Eigen documentation.

Using the XBox Controller with Ubuntu (the modern way)

If you want to get your Xbox One/ Xbox 360 running on ubuntu you basically have the choice between the in-kernel xpad driver and the userspace xboxdrv driver.

Most of the guides recommend using xboxdrv as xpad has been stagnating. However using xboxdrv has some disadvantages; as it runs as a daemon in userspace you have to manually take care of starting/ stopping it and giving your user access to the virtual devices it creates.
Xpad on the other hand just works as any other linux driver directly inside the kernel which is more  efficient and hassle free.

Fortunately while pushing SteamOS Valve has updated the xpad driver bringing it on par with xboxdrv:

  • they added support for Xbox One Controller
  • they fixed the communication protocol – no more blinking controller light

Update July 22, 2015

Unfortunately there are still several issues with the SteamOS driver. This follow-up post discusses them and the solutions in detail.

The bottom line is that I updated the official linux driver with chunks found in the SteamOS driver, as well as in several patches floating around the internet. Code and install instructions are available at Github.

How to draw a line interpolating 2 colors with opencv

The build-in opencv line drawing function allows to draw a variety of lines. Unfortunately it does not allow drawing a gradient line interpolating the colors at its start and end.

However implementing this on our own is quite easy:

using namespace cv;

void line2(Mat& img, const Point& start, const Point& end, 
                     const Scalar& c1,   const Scalar& c2) {
    LineIterator iter(img, start, end, LINE_8);

    for (int i = 0; i < iter.count; i++, iter++) {
       double alpha = double(i) / iter.count;
       // note: using<T>(iter.pos()) is faster, but 
       // then you have to deal with mat type and channel number yourself
       img(Rect(iter.pos(), Size(1, 1))) = c1 * (1.0 - alpha) + c2 * alpha;

Introducing Sensors Unity

Sensors-Unity is a new lm-sensors GUI for the Unity Desktop. It allows monitoring the output of the sensors CLI utility while integrating with the Unity desktop. This means there is no GPU/ HDD support and no plotting.
If you need those you are probably better suited with psensor. However if you just need a overview of the sensor readings and if you appreciate a clean UI you should give it a shot.

Sensors Unity is available from this PPA

It is written in Python3 / GTK3 and uses You can contribute code or help translating via launchpad.


In contrast to other applications the interface is designed around being a application. Instead of getting another indicator in the top-right, you get an icon in the launcher:

The user interface
The user interface

The idea is that you do not need the sensor information all the time. Instead you launch the app when you do. If you want to passively monitor some value you can minimize the app while selecting the value to display in the launcher icon.

To get the data libsensors is used which means that you need to get lm-sensors running before you will see anything.

However once the sensors command line utility works you will see the same results in Sensors-Unity as it shares the configuration in /etc/sensors3.conf.


Unfortunately configuring lm-sensors via /etc/sensors3.conf is quite poorly documented, so lets quickly recap the usage.

  • /etc/sensors3.conf contains the configuration for all sensors known by lm-sensors
  • however every mainboard can use each chip in a slightly different way
  • therefore you can override /etc/sensors3.conf by placing your specific configuration in /etc/sensors.d/ (see this for details)
  • you can find a list of these board specific configurations in the lm-sensors wiki
  • to disable a sensor use the ignore statement
  • #ignore everything from this chip
    chip "acpitz-virtual-0"
     ignore temp1
     ignore temp2
  • to change the label use the label statement
  • chip "coretemp-*"
     label temp1 "CPU Package"

    Sensors-Unity Specific Configuration

Sensors-Unity allows using the Pango Markup Language for sensor labels. For instance if you want “VAXG” instead of “CPU Graphics” to be displayed, you would write:

label in4 "V<sub>AXG</sub>"

In order not to interfere with other utilities and to allow per-user configuration of the labels/ sensors Sensors-Unity first tries to read ~/.config/sensors3.conf before continuing with the lm-sensors config lookup described above.

introducing is a new python wrapper for libsensors of the lm-sensors project. libsensors is what you want to use to programmatically read the sensor values of your PC with Linux instead of parsing the output of the sensors utility. is not the first wrapper – there are two alternatives, confusingly both named PySensors.

PySensors (ctypes) follows a similar approach to by using ctypes. However instead of exposing the C API it tries to be a object-oriented(OO) abstraction, which unfortunately lacks many features and makes the mapping to the underlying API hard. Furthermore it does not support Python3.

PySensors (extension module)  does not use ctypes and thus is more efficient, but if you write a python script probably compiling a extension module is worse than losing some performance when reading the values.
Additionally there is python3 support and also some OO abstraction. The latter is somewhere in between the C API and proper OO: sensors_get_label(chip_name, feature) is ChipName.get_label(feature) instead of feature.get_label().

So what makes immediately different is that it does not try to do any OO abstraction but instead gives you access to the raw C API. It only adds minor pythonification: you dont need to mess with pointers, errors are converted to exceptions and strings are correctly converted from/ to utf-8 for you.

However working with the C API directly is tiresome at times – therefore there is also an optional iterator API, which is best shown by a demo:

import sensors


for chip in sensors.ChipIterator("coretemp-*"):
    print(sensors.chip_snprintf_name(chip)+" ("+sensors.get_adapter_name(chip.bus)+")")
    for feature in sensors.FeatureIterator(chip):
        sfi = sensors.SubFeatureIterator(chip, feature)
        vals = [sensors.get_value(chip, sf.number) for sf in sfi]
        label = sensors.get_label(chip, feature)
        print("\t"+label+": "+str(vals))



coretemp-isa-0000 (ISA adapter)
	Physical id 0: [38.0, 80.0, 100.0, 0.0]
	Core 0: [37.0, 80.0, 100.0, 0.0]
	Core 1: [35.0, 80.0, 100.0, 0.0]
	Core 2: [38.0, 80.0, 100.0, 0.0]
	Core 3: [36.0, 80.0, 100.0, 0.0]

for a more sophisticated example see the in the repository.

Replacing your desktop laptop with a ITX workstation

If you use your laptop as a desktop replacement, you will at some point get an external display and a mouse/ keyboard for more convenient usage.
At this point the laptop becomes only a small case of non-upgradable components.

Now you could as well replace your laptop by a real case of comparable size.  This will make your PC not only easily upgradable, but allow higher-end components while being more silent at the same time.

Continue reading Replacing your desktop laptop with a ITX workstation

Streaming the Screen on Android

In this post I want to discuss way of getting the screen content of your Android device to the TV or monitor. If you wonder why one might want to do such a thing – just think about playing some Android games with a bluetooth gamepad or watching a movie where your PC is not available.

Specifically I want to introduce SlimPort. SlimPort is a feature of Nexus devices which is unfortunately not covered much in reviews.
Basically SlimPort is DisplayPort over the Micro-USB connection of your device allowing you to mirror its display.

But the future has arrived: we got Miracast!

One might wonder why one should go through the hassle of using a old-school HDMI cable.
You can get a Chromecast Stick for 35$ and nowadays it also supports Miracast so you can simply stream the images over WiFi.

Well Miracast is all nice if all you need to do is to put up some slides without carrying all possible adapters with you. But as soon as you try to stream a movie or a game you will reach its limitations.

Remember that Miracast works by grabbing the Framebuffer and compressing it with H.264. While encoding happens in hardware it still takes some time and it inevitably introduces compression artifacts. This means:

  • in games you get a noticeable lag – especially in FullHD
  • in movies you get noticeable artifacts – especially in FullHD
  • in both cases your battery will get drained for heavy WiFi and Encoder usage

Going old-school

Going with the old-school cable on the other hand you get HDMI 1.4 transfer rates for up to 1080p at 60Hz while saving the battery.

Configuring the second screen is quite straightforward in android. As Mirroring is your only option, there is actually nothing to configure. Once you connect the adapter android will set up your monitor based on its EDID information and transfer image and audio over HDMI.
In case you only want to have the image over HDMI, simply attach your speakers to the phone and android will re-route the audio.
The days where you had to manually set up everything are over.

Furthermore most adapters have an micro-USB port allowing to still charge your phone while using SlimPort.

Device Support

The downside is that most of the devices do not support SlimPort. The device list more or less boils down to

  • Google Nexus 4/ 5
  • Google Nexus 7 (2013)
  • LG G2/ G3

Samsung devices go with the alternative MHL. Comparing these two SlimPort has the bandwidth advantage of 5Gb/s vs. 3Gb/s of MHL so it does not have to compress that much. However both are clearly better than going wireless.


Secure Owncloud Server

This article is about how to securely configure the machine where your Owncloud instance will be running.
Even if you set-up your connection with Owncloud in a secure way,  your data still can be compromised by exploiting security flaws in the underlying architecture.

In the following we specifically will cover the underlying software stack and brute-force password hacking attempts.

Continue reading Secure Owncloud Server

How to manually update a deb package from source

Probably everyone has encountered a package in Ubuntu which was not the newest released version while one for some reason needed the newest one. The first step is to search for a PPA with the desired version. But what if there is no such PPA or you want to build the version yourself? This is where this guide comes in. Note however that this is not aimed at ordinary users – you need some experience with programming/ compiling to successfully build a package.

Before you start

Before you start make sure that you have source packages enabled in your software sources.
Next you obviously need the upstream source tar-ball of the new program which should look something like <packagename><version>.tar.gz.
Download this tar-ball to a new directory <somedir> and extract it there.

Updating Package info

For the following commands I assume you are in the previously created directory <somedir>.

First we need to get the old version of the source package

apt-get source <packagename>

This will download and extract the old source package into <packagename><oldversion>.

Now we need some helper scripts to perform the upgrading as well as the build-time dependencies of the package

sudo apt-get install dpkg-dev devscripts fakeroot
sudo apt-get build-dep <packagename>

Next change into the extracted sources of the old package and update the packaging

cd <packagename>-<oldversion>
uupdate -v <newversion> ../<packagename>-<newversion>.tar.gz

# change into the extracted new package
cd ../<packagename>-<newversion>

# update version info
dch -l ~ppa -D $(lsb_release -sc)

For more information see the Debian New Maintainers Guide.

Building the program

To trigger a rebuild of the program simply execute


Uploading your version to a PPA

To upload a package to a PPA you first need to sign it to prove that you are the author. To do this you have to execute the following in the <packagename><newversion> directory

debuild -S

Furthermore you need the upload tool dput to actually perform the uploading

sudo apt-get install dput

Now change to <somedir> and execute

dput ppa:<your_username>/<repository> <source.changes>

You can find more information at Launchpad.

Secure Owncloud setup

While the Owncloud Manual suggests enabling SSL, it unfortunately does not go into detail how to get a secure setup. The core problem is that the default SSL settings of Apache are not sane as in they do not enforce strong encryption. Furthermore the used default certificate will not match your server name and produce errors in the browser.

In the following a short guide in how to set-up a secure Apache 2.4 server for Owncloud will be presented.

Continue reading Secure Owncloud setup