Portable Real-Time Applications
In today's visual world of data processing, many people think solving problems with computers means implementing graphical user interfaces (GUIs). From this point of view, writing real-time applications means writing GUIs while at the same time mastering the system-dependent functions that ensure predictable response times, often in conjunction with arcane hardware features. This mixture of system-dependent GUIs and hardware-dependent real-time functions usually leads to complex, expensive and non-portable applications. To attack the general problem of writing portable real-time applications, I will first look back at the roots of the UNIX operating system. Then, I will apply the lessons learned to a simple multimedia application that runs on three very different platforms: Linux, IRIX and Win32 (MS Windows 95/98/NT/2000).
Most people know how a sine or a square wave sounds. They can be heard as a beep or a test signal from a PC speaker, a telephone or a common musical instrument such as a flute. The portable application we will implement produces such sounds. Sine, triangle and square waves are just special cases of a more general wave produced by the Duffing oscillator (see Resources). Depending on the parameters, which can be adjusted by two sliders in the GUI, this oscillator is also able to emit chaotic sounds. (Chaotic, in this sense, gets its meaning from nonlinear dynamics or chaos theory.) When starting the application, you will actually be able to put some research results to the test by listening to the sound and by watching the graphical behaviour of the Duffing oscillator. Nonetheless, such an application has to produce sound continuously, otherwise the sound will be distorted by clicks or even silence. Thus, this application is an example of a real-time application.
Before starting our implementation, we must think about the design of such a portable application, forgetting for the moment that we want to implement our multimedia application with a GUI. Most UNIX programmers know what is meant by the UNIX philosophy. In the good old days, when all user interfaces were textual, not graphical, Kernighan and Pike explained this notion in the epilogue of their book The Unix Programming Environment. They emphasized the importance of breaking a problem into separate sub-problems with a simple interface between them, usually a pipe. Putting a pipe between the processes means having the textual output of one process read as textual input by the next process. Each sub-problem was then implemented and tested stepwise on its own, preferably by applying existing tools.
This design philosophy allows for writing portable applications and is a sharp contrast to today's development environments. Today, many programmers use some visual development environment to build monolithic applications bound to one platform.
The crucial question is: can a GUI-based real-time application be implemented as an old-fashioned UNIX pipeline? It can—you just have to choose the right tools. A GUI-based application which allows for adjustment of parameters, emission of sound and visualization of results can be broken down into a pipeline of processes: GUI-->Sound Generation-->Sound Output-->Graph.
Stage 1 (GUI) provides a way to interactively adjust the parameters of a mathematical model. These parameters may be adjusted with knobs, slide bars or with a point in a two-dimensional plane, whichever is most intuitive. It can easily be replaced without affecting other stages, provided it produces the same kind of data on its standard output.
Stage 2 (generation) is invisible to the user, so it does not need a GUI. It takes the parameters from stage 1 and processes them using a mathematical model of a physical process. The only challenge at this stage is doing both parameter input and continuous computation at the same time, but at different unsynchronized rates.
Stage 3 (sound emission) reads the resulting waveform and hands it over to the sound system. Because of significant differences in the implementation of sound systems on platforms like Linux, IRIX and Win32, we need to have some experience with encapsulating platform-dependent code. Fortunately, this is the only stage written for a particular platform.
Stage 4 (graphical output), just like stage 1, has contact with the user and will therefore be implemented as a GUI. Just like stage 1, the results can be represented in many different ways without affecting other stages. Examples are tables, simple amplitude diagrams, trajectories in phase space or Poincaré sections.
Each stage can also be used on its own or as a building block in a completely different application. Stage 3 is the most interesting building block.
- Readers' Choice Awards 2013
- Advanced Hard Drive Caching Techniques
- Linux Kernel News - November 2013
- December 2013 Issue of Linux Journal: Readers' Choice
- Mars Needs Women
- Sublime Text: One Editor to Rule Them All?
- Raspberry Pi: the Perfect Home Server
- RSS Feeds
- Web Administration Scripts
- New Products
- animal pajamas
2 hours 28 min ago
- thanks for you post.
2 hours 35 min ago
- thanks for share, great
19 hours 49 min ago
- There are factors which are
1 day 48 min ago
- Gnome 3 ?
1 day 1 hour ago
- Reply to comment | Linux Journal
1 day 5 hours ago
- "Redis RethinkDB 4.5%" on Best NoSQL Databases
1 day 15 hours ago
- on the ground
1 day 22 hours ago
- I was able to read the whole
1 day 23 hours ago
- since i have read the title i
2 days 2 hours ago