Having developed entire subsystems (graphical, compression, etc.) for several OSs, we have thorough knowledge of how system software works. Among completed projects:

Implementation of POSIX-compliant APIs

This following is a brief CyNIX specification — for this operating system, we've developed API, SDK, graphics and compression subsystems, and the Java machine:

• POSIX-compatible; EL/IX specification (Embedded Linux, levels 1, 2, and 3) supported as well
• Less than 1M footprint with Java2 Micro Edition (CLDC/MIDP) included
• Modular structure, highly configurable to any particular needs
• Runs on ARM7/ARM9 CPUs, although portable to other hardware
• Features compressed modules and archives
• Dynamic libraries (shared objects), supporting symbolic linkage
• Both Unicode and multi-byte characters supported
• Extensive C/C++ graphics library; its low-level subset made somewhat similar to respective J2 Micro Edition classes for increased usability
• Easy to use SDK featuring C and C++ compilers, support tools, various converters and utilities, artwork (4300+ files) for royalty-free use in applications, and extensive documentation
• PIM-style sample applications (like phone book/dialer), in both C/C++ and Java; full source code included with the SDK
  Gated open source distribution
  

Programming Intel, ARM, Motorola, Hitachi, etc. CPUs

We've written programs for Intel, ARM, Hitachi, Motorola, and other CPUs, using both high-level languages and respective processors' assemblers.

Some of those applications were time-critical, requiring thorough knowledge of processor features to maximize their potential - such were real-time compressors/decompressors (some built into OSs), low level graphics procedures, as well as bytecode interpreters.
For our development tools, we implemented back-ends, employing high level of optimization, for Intel, Motorola, and Hitachi CPUs.

Having developed entire subsystems (graphical, compression, etc.) for several OSs, we have intimate knowledge of how system software works. Among completed projects:

Fast real-time compressors-decompressors

We have proprietary compressor working with a 4k sliding window dictionary, and having outstanding speed performance, which makes it exceptionally well suited for real-time applications such as on-the-fly "transparent" compression/decompression.

Its compression speed is slightly better than of all of the other compressor belonging to the LZ77 family of compressors, while decompression speed is amazingly high, being very close to RLE. Which, in turn, means that decompression is almost as fast as reading uncompressed data.

This compressor, therefore, is even better suited for real-time applications than the patented (by Unisys) LZW compressor that is used for GIF files. But, unlike LZW, its use is completely free.

Customized compressors with optimum ratios for specific data sets

We have developed a special file statistics-gathering application that, if run on a substantial amount of data, is capable of semi-automatic building of compressors optimal for such data.

It works by compiling tables of relative frequencies of particular lengths of any repeating sequences of bytes, as well as distances to such sequences within compressor's sliding window dictionary. The result is a set of Huffman code tables that is optimal for a given particular type of data.

This technology had been successfully used for the built-in compressor of an operating system used in PDAs.

Powerful compressors beating ZIP and matching RAR

This particular compressor uses block sorting technique and does best when used on large volumes of data. It has great compression strength, almost always beating ZIP, sometimes by a huge margin, and often beating RAR, but is not as light-weight as our other compressors.

We guarantee that the underlying technology is 100% royalty-free.

Image optimizers

We have developed a number of image optimization techniques. All are effective, but some are truly unique.
For instance, for use with Abot, we have developed a dithering (a.k.a. color approximation error propagation) algorithm working on just individual lines of a high-color image, with great results.

Another interesting algorithm worth mentioning here is HLS-based (Hue-Luminance-Saturation) color reduction algorithm, capable of preserving "overall" impression made by the image. As an example, suppose you reduced number of colors in an image to just one; that one color would then be equivalent to the color you would "perceive" if you looked at the original, unmodified image from such a long distance that the entire image would appear to be a single colored spot, or even dot).

Such algorithms, along with animation optimizers and other such tools, are instrumental in building high-performance graphical OS subsystems for limited devices.