Brief Description of HiFill Accelerator and Comparison with OpenGL

OpenGL is a system used primarily in CAD and scientific visualization of data, where picture quality and accuracy is not as critical as in the virtual reality games and multimedia. Due to its low price and high performance, HiFill can replace OpenGL in PCs.

In OpenGL the drawing speed is limited even at costly hardware.  Moreover, this hardware is specialized and tailored for curremt specs of the system. In order for the system to use its full potential, the OpenGL hardware has to be upgraded with ever increasing speed  of processors. Performance of the HiFill accelerator increases with higher speed processor, with no need for modifications. Where standard accelerators are used central processor has to wait after cache memory is filled. That limits performance of the entire system.

Graphic accelerator HiFill works without delay, the speed of the system is always equal to the speed of the processor. Acceleration is achieved primarily through reduction of data on several levels. There is no need to calculate the entire area of elementary objects, transfer them over the bus, and save them in the video memeory, since HiFill works with line drawings of polygons. Speed of standard graphic systems depends not only on video memory but on bus throughput. HiFill transfers less data over the bus never reaching the transfer capacity and acceleration can be increased by simply using a faster video memory .
For instance, the speed can be increased four times using MDRAM in video memory than using standard DRAM. HiFill is currently the only simple choice for PC under Unified Memory Architecture. Due to the Extended Color Mode (mode where in 256 colors HiFill provides 4M colors) the entire videopage with resolution of 800x600 pixels can be placed in secondary cache and increase the speed even more .

HiFill is considerably more open for use with a variety of (visibility) algorithms than other graphic accelerators and their set of accelerated functions.  They reduce volume of data too, but programmers are required to use certain type of algorithms. These then do not accelerate more sphisticated scanline and raytracing algorithms. On top of that, there are a number of sets of accelerated functions released by manufacturers and varying from version to version, which makes programming much more complex.

HiFill allows programmers to use those fractal textures with higher zoom factor, those where Level of Detail management can be used, and where memory requirements are low. HiFill together with textured LOD allows for change in picture quality depending on the camera movement (faster movement = less detail = more pictures per second) and speed of processor. Description of fractal textures allows for certain objects to enter into a model , after the camera gets closer. That is of real advantage in scenery modelling applications with a large number of  details on the horizon (small model for fast 3D transformation).
Fractal textures can be very well animated in real time - shaders.

HiFill is designed primarily for scanline alghoritms, where Phong shading, Radiosity and Nurbs in rasterization (model is small) can be used and the picture quality equals RayTracer (in this time is know realtime scanline renders - RendrWare - it is more speed then SW OpenGL ). HiFill is a relatively simple circuit (2Kgates) and if integrated in RAMDAC or VGA or UMA architecture on the chipset, does not significantly increase the price of the circuits.

Next version of HiFill, Turbo HiFill (10Kgates) will have a reduced videobuffer so that it can be all placed in the first level cache in the main processor. The main goal will be to use maximum compression of pictures for use that is compatibel with HDTV (from 1:100 to 1:10 000 without compromising picture quality). Future version (100Kgates) will rasterize in real time by beam, directly from the model, without dealy. 
